SCROLL NEWS / DISCOVERY
Search headlines
Find the stories shaping the conversation.
Results for " Loading " 101 found 🔀 AI-powered Shuffle
New downtown Salt Lake entertainment district plans raise hopes, questions in Japantown
Smith Entertainment Group’s new three-block downtown Salt Lake City district would replace Japantown Street loading docks with shops and offices.
Genesis Mission: SYNAPS-I Brings Agentic AI to Experimental Workflows
With support from ALCF computing resources and AI services, the SYNAPS-I team demonstrated a new agentic AI platform that uses natural-language instructions to automate complex experimental workflows as part of the DOE Genesis Mission. Oct. 8, 2026 — Science has increasingly used artificial intelligence (AI) as a kind of microscope — sorting data, analyzing images […]
Who Won August and September: Original Films or Franchises?
(Welcome to the Entertainment Strategy Guy, a newsletter on the entertainment industry and business strategy. I write a weekly Streaming Ratings Report and a bi-weekly strategy column, along with occasional deep dives into other topics, like today’s article. Please subscribe.) I just heard someone say that if you’re reading a pundit or data analyst, and they don’t tell you that they got anything wrong, they’re not a pundit but an “influencer”. I agree with that. You have to tell your audience when you get things wrong, or they won’t (or shouldn’t) trust you as much. Today, we’ve got another edition of “What I Got Right, What I Got Wrong”. In general, I’m patting myself on the back (and patting very hard) about a few things like IP at the box office, LIV golf, horror films, crowdfunding movies, superheroes, and old people going to the theaters. That said, I also made a couple of data mistakes on the streaming bubble popping and HBO’s datecdotes, so I’m not perfect. But first, I need your feedback... Subscribe Follow-Ups: What Should We Call Mid-Budget Movies Which Aren’t Cinematic Enough for Theaters But Are Too Expensive to Make Money on Streaming? After I asked for feedback on what we should call straight-to-streaming films that are too big to pencil out on streaming but not really big enough to resonate in theaters, I got some great suggestions from you all, including... “Moldilocks” from John Aboud “Little Big Indie” from Travis Frick1 “Midflicks” from Jonathan Funke. “Extra-Medium” from Jona Nwuke. (Read the explanation in the footnote.) Between these and Brandon Katz—who suggested “Bermuda Budget Triangle”, “The Platform Gap” and “The Distribution Deadzone”—we’ve got some excellent suggestions. And they’re better than my suggestion, the “Streaming Budget Dead Zone”, so let’s take a poll! The winning entry becomes the new term. Loading... RIGHT: LIV Golf is Fully Bankrupt…Is Anyone Else Next? A few years ago, I was always a bit perplexed at all the articles I’d read praising LIV Golf and their strategy to disrupt the PGA. LIV Golf, a brand new professional golf league, offered PGA stars ten times what they made on the PGA Tour to join their new league. And folks praised this strategy for its initial “success” in that a lot of big names did indeed leave the PGA. Yeah, of course the players came over. LIV Golf paid them so, so, so much more! But LIV Golf hadn’t discovered a way to increase potential revenue. So think about this in basic business terms… They paid much more in costs… …but had no real way to increase revenue. That’s not a strategy! That’s deficit financing. It won’t work unless you bankrupt the competition (and then turn around and pay those same golfers much, much less). It’s not a sustainable strategy and only lasts as long as the company/person/nation backing it decides they’re cool with losing money. For LIV Golf, that meant Middle East oil wealth, in particular Saudi Arabian money. Clearly, the current Iran War has hurt Middle East finances, so they need to trim their more exorbitant spending. And LIV Golf was part of that trimming. This should be a major warning for others. TGL’s parent company, TMRW Sports, just got a $1 billion valuation, despite TGL’s (the indoor golf league) horrible TV viewership of less than half a million viewers per match. Unrivaled, the women’s basketball league, just secured a $650 valuation, despite its horrible TV viewership and a new rival (Project B) entering the scene next year. To relate this to streaming, Hollywood should ask which streamers may have wealthy patrons funding their losses. (Let’s be clear: Google, Amazon and Apple.) Could those patrons lose their appetite? For Amazon, probably not. But Apple has a new boss, so maybe! WRONG: My Analysis of the Streaming Bubble Was Missing a Week I’m going to be congratulating myself a lot today, but I make mistakes too, like this data goof. When I last compiled the data on the decline in TV shows, I was missing one week at the end of June. If you look at the first image, you can see that one week was mislabelled as “July”. Mainly, this image got updated to show a 27% decline, not 29%: This change doesn’t really impact the overall analysis, but it is two percent better. By the way, through the third quarter, the decline increased and it’s now a 30% decrease since 2022. (I’ll write/visualize this in an upcoming article.) But I want you to trust me and my data. Especially these days, when many people are using LLMs that I know are inserting faulty data into their charts, I want to keep earning my audience’s faith. So that’s the accurate data. RIGHT: IP Remains Very, Very Popular In July, I wrote a giant (and I mean giant) article on Backrooms, Obsession, and the box office, going over what we know, what we don’t know about what works in theaters, looking at YouTubers, IP, the horror genre, comic book movies, and a whole lot more. The month of August really tested a lot of my theses and, being honest, mostly supported my arguments. Let’s start with IP. Looking at 8-Aug (the weekend after Spider-Man: Brand New Day came out) to 18-Sep (the weekend that Resident Evil came out, which I think provides a nice bookend to this time period), there were seven films based on pre-existing IP: Resident Evil (2026) ($126 million) Practical Magic 2 ($65 million) _Insidious: Out of the Furthe_r ($65 million) Coyote Vs. Acme ($59 million) Paw Patrol: The Dino Movie ($53 million) Tony ($15 million) Super Troopers 3 ($7 million) Compare those to the notable original films from the past month—I actually could have included more movies, but here are just thirteen, bringing us to an even twenty films—including.... The End of Oak Street ($54 million) Buddy ($26 million) Mutiny ($15 million) By Any Means ($15 million) The Dog Stars ($14 million) Runner ($14 million) One Night Only ($11 million) Hope ($8 million) Spa Weekend ($7 million) The Uprising ($6 million) Teenage Sex and Death at Camp Miasma ($6 million) Onslaught ($3 million) Eli Roth’s Ice Cream Man ($2.8 million) Here’s that in chart form: Five of the top six films in this time period were all based on IP. I made a big chart of films that grossed over $200 million at the box office before Spider-Man: Brand New Day and The Odyssey hit theaters. Let’s update that chart! By the way, if you want to see how I categorized each film—so another bar chart—here it is: I know that many of my fellow critics/pundits/analysts dislike films based on IP and how Hollywood is making so many of them. And I’m sympathetic to this point of view. As I’ve written many, many times before, you need a balance between existing franchises, new IP, and original films. And Hollywood clearly needs to make more films like Resident Evil (a well-made film from a visionary director) and fewer Practical Magic 2’s (which didn’t get critical or customer buzz). But at some point, critics and pundits need to contend with what audiences are telling them: Movie-goers aren’t showing up to original films. Audiences are speaking with their dollars, telling you they want more IP and franchises. You can try to convince studio heads to make fewer IP-based films and franchises, but the data and numbers aren’t there. Instead, critics need to work harder to convince audiences to show up for original films. Aim your ire/concern at the average person, not studio heads.2 Because they’re just making the films that audiences are telling them to make. WRONG: Original Horror Films Didn’t Break Out I’ll be honest, even though I wrote an article casting some skepticism on the horror genre in July, if you asked me to make a prediction, I would have predicted that, in August, a new, original horror film would have blown up. No, seriously, I just assumed that Obsession and Backrooms presaged a change in audience behavior. But none of the buzzy new original horror films from August—Teenage Sex and Death at Camp Miasma, Onslaught (not an action film in spite of the ads), The End of Oak Street, Eli Roth’s Ice Cream Man, or _Buddy—_broke out. To be clear, exactly one of those films (Buddy) had good “ROI”, but again—I try to be specific in my language—none were “popular” in any broad sense of the word. None of them will be “saving” movies theaters like Backrooms or _Obsession_helped save the summer. The new Insidious film and Resident Evil, both based on IP, were far and away the biggest horror films since July. (We’ll see if this changes in October/Halloween season.) RIGHT: Stay Skeptical about Crowdfunding... I’ve long been skeptical about crowd-investing platforms as one of Hollywood’s saviors, mainly because there’s so much hype/buzz. People need to stay more skeptical about more things, explaining both the potential upside but also the downsides. In particular, the media often hypes crowdfunding at the start and never checks in on the actual results after they’ve come in later. And August gave us our first update! _Ice Cream Man—_directed by Eli Roth—grossed $6 million off of a $5.5 million budget. This is a production of The Horror Section, which was one of the first “crowd investing” studios with 2,400 investors, which means that 2,400 investors probably lost money. They certainly aren’t getting as great of returns as if they had just invested their dollars in the stock market. Hopefully Stiletto (Tagline: “Someone’s Going to Make it Rain Blood!”) does better next month. WRONG: Another Data Goof Here’s another data error. When the first episode of House of the Dragon came out, HBO put out that it had 21.5 million viewers in the first three days, and I read that to mean in the US…but no, it was global. So my US-only datecdotes charts shouldn’t have included it. We never got US-only numbers for the first episode, but for the final episode, HBO put out that it had 11 million US viewers. (And that global dropped to 21 million viewers.) Here’s the updated chart (which I’ve since used in the Streaming Ratings Report): Still, this show is absolutely huge. RIGHT: Superhero Films Remain Very, Very Popular After Supergirl flopped, I read a few takes that “comic book movies are going the way of the Western”. Post-Spider-Man: Brand New Day, that take didn’t age well. To be fair, I have a very nuanced take on the superhero genre right now; it’s down right now, for a lot of reasons. But it’s not “dead”. Maybe _Spider-_Man is just a really popular character? I saw that take, and it’s a fair counter-argument. (But pundits arguing that superhero movies were dead should have mentioned this $1.5 billion counter-argument…) But is it just Spider-Man? The next Avengers film already has $50 million in pre-sales (and I was skeptical that that film would do well) and the Avengers: End Game re-release topped the box office two weekends ago (over three original films). And I wouldn’t bet against Batman or Superman. So maybe it’s just Spider-Man, Batman, Superman and the Avengers. Oh, and Deadpool, of course. And Black Panther. And Wolverine. And probably the X-Men. Plus a well-made Wonder Woman or the Hulk film could break out. But that’s it! It’s just those ten characters/teams. Oh, what’s that? Lanterns is also doing well on HBO? (See previous section…) To be fair, I’m actually pretty sympathetic to the argument that more popular characters—like Spider-Man and Batman—anchor more popular films. In fact, I made that exact argument three years ago when I first wrote about the “Marvel-cession”. In many ways, you can blame The Guardians of the Galaxy for fooling Marvel Studios (and the rest of us) into believing that any character could pop. It turns out, the list of iconic characters is probably smaller than most people think. But it’s probably too early to say that superhero films and comic book movies are dead unless “death” means a slight decline over a longtime. RIGHT: Who Killed Theaters? Old People I get frustrated whenever I see headlines or analysis about how young people are “returning” to theaters. As I’ve detailed(for years), young people have always powered the US box office, despite narratives about “kids these days” and their “phones”. Really, what’s changed post-2020/pandemic is that old people aren’t going to the movies nearly as much. This summer, I saw a movie (from an older director) in a theater near a retirement community, and multiple older people at the theater were talking about how this was their first time seeing a movie in years. I dislike personal anecdotes, so YouGov can fill in the data, best summarized by this headline: “Who killed movie theaters? Not the youths”. According to them, 64% of people aged 18-29 have seen a movie in the last year, but only 30% of 65-and-older. 20% of 18-29 have seen a movie in theaters in the last week and 42% in the last month. Here’s the polling data: Most concerning? Many Americans (17%) think theaters are a worse or much worse experience than watching films at home. Slight WRONG: Hadestown Opens Big A live theater capture of the Broadway musical, Hadestown, made $20 million at the US box office, which begs the question: was I wrong to be skeptical about musicals a few years ago? Yes and no. On the one hand, $20 million is a far cry from being “popular”, so yeah, the genre isn’t that popular overall and Hadestown is one of the more popular musicals from recent years (i.e. the “Taylor Swift Data Fallacy” in action). On the other, I doubt filming this cost all that much, and I don’t think that they spent much on marketing, so this is a good source of ancillary revenue. Smaller Updates WRONG: As I mentioned in a Streaming Ratings Report, I underestimated the budget for Enola Holmes 3. It probably cost more like $50 million, if not more. But... I’m not sure that it really matters? At sub-10 million hours, prices have to come down to make this work. WRONG: Netflix is giving Ink a 27-day in theaters! To quote the kids/YouTubers these days, let’s go! Now I might actually have a chance to see Danny Boyle’s latest in theaters. I’d complained about this in a “Coming Soon” section, but I was heartened to read that Netflix is giving multiple films longer theatrical windows this year. RIGHT: Netflix is sending 4-5 films per year to theaters. Netflix is slowly but surely sending more and more films to theaters, as I cautiously predicted earlier this year. For now, it’s just three big films and a number of awards contenders, but still, this is great news. And they’ll be releasing box office grosses! Just this week, Ted Sarandos confirmed that KPop Demon Hunters 2 will come to theaters (and my guess is it performs in the box office top ten at a minimum). WRONG: Angel has 3 million subscribers! How do I know this? Well, they told Deadline, who reported it. I marked this as “wrong”, since they’ve doubled their subscribers in one year but, you know, they don’t really have a hit film to speak of and they’re still losing money. WRONG: Furious was only renewed for one more season. I accidentally wrote “two more seasons” in my latest “Renewals, Cancellations, Un-Orders and Removals Update”. RIGHT: House of David is ending with its third season. In July, Prime Video renewed House of David for a third season, as I just wrote in my latest “Renewals and Cancellations” report. Well, now it’s ending after that third season. Why? As I’ve been writing, its viewership wasn’t great. I got feedback that this show didn’t cost very much, but it cost enough that its limited viewership didn’t save it. WRONG: Adults was a Hulu original! So I missed Adults when it first came out last year; I saw that it aired on FX and just assumed that it was a linear-first program. Turns out, it aired three episodes on FX, but binge-released the rest of its episodes the next day on Hulu. Huh. So I should have covered it last year! But I just wrote about it. 1 “When I was a kid in the 90s, if you wore a t-shirt to school that wasn’t too big and also wasn’t too small, but somehow didn’t quite fit, we’d say you were wearing an “extra-medium” shirt.” 2 As always, a huge exception is Disney, which barely makes anything original anymore.
$350 million Memorial Health project to bring road and parking changes
A $350 million Memorial Health project will add a 90-bed tower, cafeteria, loading dock and parking garage.
Football Frenzy 10/02
NORTHEAST KANSAS (KSNT) — Week 5 of the high school football season is in the books. Send final scores to Kole.Emplit@ksnt.com and Sydney.Clark@ksnt.com. 6A 5A 4A 3A 2A 1A 8-Man For more sports news, click here. Keep up with the latest breaking news in northeast Kansas by downloading our mobile app and by signing up for […]
Google’s new Android sideloading rules are live — and here are all the ways to already bypass them
Google’s developer verification rules are now live, but users can bypass new APK sideloading blocks using ADB, Shizuku, and Advanced Flow.
How to Listen to Your Favorite Music, Online or Off the Grid
Streaming services can introduce you to a wonderland of nearly unlimited audio, but downloading your digital collection has advantages.
Scientists Solve a Decades-Old Stick Insect Mystery
The discoveries could help manage stick insect outbreaks and reduce their impact on forestry and agriculture. Australia’s coastal landscapes are home to large stick insects belonging to the genus Anchiale, a group of closely related species. For decades, scientists and nature enthusiasts have documented them, uploading more than a thousand observations to databases like iNaturalist
Intel Panther Lake Teardown, 18A, BSPD, GAAFET, SemiAnalysis STEEL
Panther Lake debuts the first commercial implementation of backside power delivery (BSPDN), introduces Intel’s first iteration of gate-all-around (GAA) transistors, and showcases their advanced packaging capabilities with its Foveros-S assembly. With Panther Lake, Intel’s manufacturing arc has shifted from nebulous roadmaps to shipped silicon, a significant milestone on their long road back to competitive semiconductor manufacturing. To evaluate the extent of Intel’s comeback, we tore down Panther Lake. The SemiAnalysis STEEL teardown lab breaks down advanced datacenter and AI hardware. To learn more about our pipeline or to commission a teardown, contact sales@semianalysis.com. WE’RE HIRING: Architecture, floorplan, packaging, manufacturing, and labs experts. Opportunities from system to transistor and everywhere in between. Check out our Careers page. Our teardown traces 18A from its four-sheet RibbonFETs (Intel’s marketing name for GAAFETs) and gate stacks through contacts, frontside and backside wiring, and the bonded carrier. We explain how these material and integration choices improve gate control and reduce resistance, while adding capacitance, thermal resistance, and process complexity. Our measurements put Panther Lake’s 18A compute logic and TSMC N3E GPU logic at similar logic density. However, 18A does not lead TSMC N3P, N2 or Samsung SF2 in peak density. Panther Lake’s CPU cores are incremental updates, and the high-end GPU still uses TSMC N3E. Panther Lake assembles one compute tile, one GPU tile, and one I/O tile atop a passive base tile using Intel’s Foveros-S advanced packaging. Both compute tile variants use Intel 18A. The Xe3 GPU options are a 4-core GT1 tile on Intel 3 and a larger 12-core GT2 tile on TSMC N3E. Both I/O tile variants use TSMC N6. [1], [2] Our analysis centers on the PTL-U compute tile, both the 4-core and 12-core GPU tiles, as well as the 12-lane I/O tile. In conventional chips, power and signal are routed through the same frontside metal stack towards the device frontend. Power rails consume scarce routing resources near the transistors, while tall via stacks carry VDD and VSS from the coarse upper wires to local rails. Backside power delivery (BSPD) moves the main power network behind the transistor layer, to the backside, separating it from frontside signal routing. We covered BSPD and its impacts in 2024. [3], [4], [5] Intel’s BSPD implementation, branded as “PowerVia”, routes power through dedicated backside metals to nano-TSVs, which connect those rails to local source/drain (S/D) contacts. Implementing that separation requires Intel to build the interconnect stacks from both sides of the wafer. The frontside comprises the M0-M14 signal stack, while the backside comprises the BM0-BM5 power stack. M0 and BM0 are closest to the transistors. The nano-TSVs connect the two sides, but Intel patterns and etches each via from the front after forming the contacts. A narrow via runs from the side of the contact deep into the silicon substrate. Intel then completes the frontside signal metal stack, bonds the wafer to a carrier, flips it and removes the original substrate until the buried via tips are exposed. The backside metal stack is then deposited directly on the revealed vias. The nano-TSV and backside-via profiles taper in opposite directions because Intel forms them from opposite sides of the wafer. The transistor structures form the FEOL. Local contacts and nano-TSVs connect them to the wiring. M0 begins the frontside interconnect stack. The silicon carrier remains attached above the frontside interconnects. It supports the device wafer during substrate removal and backside processing and remains part of the finished chip’s thermal path. PowerVia removes the main power distribution from the congested frontside metals, routing supply through shorter and wider backside wires. Its lateral landing still occupies area in the standard cell, so it recovers less cell area than a direct backside contact. [3] Nano-TSVs beside the logic devices carry VDD or VSS from the backside power network, while signal connections continue upward through the frontside metals. Backside Interconnects Samsung SF2 data is included for comparison to Panther Lake’s within this article. SF2 is the incumbent GAA foundry node but lacks BSPD, serving as a useful reference to evaluate 18A. A full teardown of Samsung’s S26 products, processed on SF2, will be shared soon.Nanosheet-cut EDS comparison. The PowerVia supply path runs from the backside Cu rails through Mo-lined W nano-TSVs to the local transistor contacts. In this cross section, the tapered connection spans roughly 150 nm from the contact level to BM0. The Ta liner confines Cu and promotes adhesion to the surrounding stack; the AlOₓ etch stop controls the next dielectric etch above the rail. Dielectric beneath the ribbons electrically separates the devices from the backside wiring and removes the conducting silicon body below the channel. [6] AlOₓ serves as an etchstop (ES), enabling endpointing and protecting the underlying layers. Low-volatility aluminum fluoride reaction products resist the fluorinated plasma, allowing a thin AlOₓ film to protect the metal while the surrounding low-k dielectric is removed. [6], [7]. While the BM0 and layers above the M1 lines show double AlOx layers, Our SMIC N+3 teardown showed single AlOₓ layers. SMIC uses a simpler local AlOₓ substack, while the remaining cap and etch sequence provide the required landing protection. So why double layers? The closely spaced AlOₓ doublets provide two protected endpoints in the etch sequence. Intel documents an AlOₓ/SiN/AlOₓ stack that explains the benefit. The main dielectric plasma etch stops on the first AlOₓ film; a selective wet clear opens that film; a second plasma etch removes the intermediate SiN and stops on the second AlOₓ film. The final wet clear exposes the metal landing surface. SiN is the intermediate dielectric in Intel’s published example. [8] The second stop protects the metal through a cap breakthrough. Wide openings can etch faster than narrow ones, and etch depth varies across the wafer. Metal under an early-clearing opening would otherwise be exposed while other openings still need more etching. Staged protection widens the process window and reduces metal erosion, corrosion and void formation. [8] TSMC documents AlN/AlOₓ/SiOC/AlOₓ above Cu, with AlN blocking Cu diffusion, and a simpler AlN/SiOC/AlOₓ variant that omits one AlOx film. [9] Levels with different opening sizes, aspect ratios, pattern densities and cap materials need different etch margins. A double AlOx stop is useful where another protected endpoint justifies the added processing. The extra film adds formation, selective opening and cleaning steps, plus another set of interfaces to control adhesion, moisture, and stress. These blanket films are opened through the existing via pattern, so each film does not require another lithography mask. AlOₓ adds parasitic capacitance when it replaces lower-k dielectric; two thin AlOₓ films can nevertheless contain less AlOₓ than one thick film. Total thickness, placement, and theintermediate dielectric determine the electrical cost. Deposition chemistry also changes AlOx permittivity and residual hydroxyl content, which can oxidize the underlying metal. [7], [8], [10], [11] The backside stack separates into relatively fine BM0-BM2 wiring near the devices and coarser BM3-BM5 power distribution. The largest pitch increase occurs between BM2 and BM3. BM0’s pitch closely matches the logic-row height, fitting local power delivery to the cell rows. Higher levels aggregate current through larger conductors: routing density becomes less important than low resistance and current capacity as the network approaches the package. This hierarchy provides wide power wiring for the power delivery network without consuming scarce frontside signal-routing resources. [3] The SemiAnalysis STEEL teardown lab breaks down advanced datacenter and AI hardware. To learn more about our pipeline or to commission a teardown, contact sales@semianalysis.com. WE’RE HIRING: Architecture, floorplan, packaging, manufacturing, and labs experts. Opportunities from system to transistor and everywhere in between. Check out our Careers page. Frontside Interconnects Intel 18A combines Mo-lined W contacts and nano-TSVs with a separate backside Cu power network. Samsung SF2 keeps power on the frontside, using Ti-based contact interfaces and Ta-based barriers and Co liners around Cu wiring. In 18A standard-cell rows, backside power rails supply the devices through nano-TSVs within the cells, freeing frontside routing resources. Samsung’s M0 accommodates both power and signal connections. From the device toward M0, the connection runs through a Ti-based S/D interface, W contact fill, a Mo-lined W via, and the Cu M0 wire. Mo supplies a conductive nucleation and adhesion layer for W, replacing the resistive TiN liner used in conventional W integration. This increases the effective conduction volume within the feature while retaining W fill and its established polishing, cleaning and etching processes. Intel’s Mo/W patent describes this integration tradeoff. The nano-TSV uses the same Mo-lined W construction in the backside supply path. [12] The move from TiN to Mo is an incremental change. While a full Co or Mo fill can also reduce the volume lost to liners in very small features, it requires new integration schemes that increase complexity and risk. Cu remains attractive for wider wires due to its low resistance. As wires and vias shrink, the diffusion barrier consumes an increasing fraction of their cross-section. [12], [12], [14] Intel uses Co/Ru liners at M0-M1, Co at M2-M4, and Nb at M5-M9. The lower-level liners help Cu adhere and reduce void formation during trench fills. Applied Materials’ Endura has new thermal control that facilitate wetting process, so the thin film continuity is good enough that good capillary pressure will drive Cu atoms to the via bottom without voiding. Intel’s choice to use Nb is particularly interesting. Intel’s Nb patent describes a conductive diffusion barrier intended to reduce the barrier’s contribution to resistance relative to conventional Ta-based barriers, particularly at via bottoms where all current crosses the barrier. The patent pairs Nb in coarser levels with the option of lower-cost PVD processing. [15], [16] The upper metal layers support thicker barriers formed through physical vapor deposition (PVD) despite its worse coverage and uniformity. Meanwhile, the lower metal layers require thinner barriers deposited through conformal atomic layer deposition (ALD). Co/Ru adds another material interface and requires controlled deposition and Cu fill. Changing liners and barriers by metal layer allows Intel to optimize interconnect resistance, process complexity, and reliability. [15, 16] RibbonFET, Intel’s name for its gate-all-around FETs (GAAFETs), replaces the FinFET’s vertical fins with four stacked horizontal silicon nanosheets, allowing the gate to surround the channel on every side. The path to GAAFET begins with the planar transistor. A planar MOSFET places the gate above the channel between its source and drain. Pairing an NMOS with a PMOS transistor creates a CMOS inverter, in which the NMOS pulls the output low for a high input, and the PMOS pulls it high for a low input. The gate must retain electrostatic control of the channel to ensure clean switching. As gate lengths shrank, the drain began to compete with the gate for that control, increasing off-state leakage. Electrostatic control was restored through an architectural evolution that raised the channel into a vertical fin and wrapping the gate around three sides. Called “FinFET”, this new architecture packed more effective channel width into a smaller footprint. Further scaling made it harder to maintain both drive current and leakage within smaller cells, and reintroduced the same problems planar MOSFETs faced. Nanosheet GAAFETs close the fourth side by replacing the vertical fin with a stack of horizontal nanosheets, each surrounded by the gate. The tighter electrostatic control suppresses leakage at shorter gate lengths while stacking adds effective channel width within the cell footprint. In a FinFET process, channel width changes in discrete steps as designers must add or remove whole fins. Nanosheet width can instead be adjusted continuously within the process’s design rules. Wider sheets increase drive current, while narrower sheets reduce capacitance at the cost of drive current. Intel 18A uses stacks of four nanosheets each and varies their widths across logic and SRAM. At the process level, adding more sheets to each stack increases effective channel width and drive current, but complicates fabrication. RibbonFET vs MBCFET Samsung began GAAFET production in 2022 with SF3E, following with SF3 and now SF2. Its ‘MBCFET’ provides a useful structural comparison with Intel’s first RibbonFET implementation. [17] STEEL is digging deeper into SF2, used in the Exynos 2600, and TSMC’s GAAFET N2, used in Apple’s A20 Pro, in upcoming newsletter articles. We’re throwing some teasers on X. Let’s compare Samsung SF2’s MBCFET with Intel 18A’s RibbonFET. Subscribe Even to the untrained eye, Intel’s extra nanosheet is obvious. Intel stacks four ribbons to Samsung’s three. Samsung’s sheets are much wider in these fields, so both sheet count and width matter to the available channel perimeter. Sheet width also changes which silicon surfaces carry current. On conventional (001) silicon, wide nanosheets emphasize the broad top and bottom surfaces, favoring electron transport; the larger sidewall contribution in a narrow sheet favors hole transport. Thinner sheets improve gate control but increase confinement and scattering. This makes width and thickness part of the NMOS/PMOS balance, alongside strain and threshold voltage. [18], [19] GAAFET designs like 18A use different work-function-metal (WFM) stacks for NMOS and PMOS. Around each ribbon, a thin SiOx interfacial layer separates the silicon channel from the HfOx high-k dielectric, with La providing dipole tuning and the WFM wrapping the dielectric. NMOS uses a TiAl-based stack, while PMOS uses TiN WFM. W fills the remaining gate trench, providing a lower-resistivity path where the work-function layers are no longer needed. In this field, the PMOS stacks leave room for W between ribbons, while the NMOS stacks occupy more of those gaps. A silicon-based dielectric marks the P/N boundary, allowing the PMOS and NMOS gates, sharing the same gate trench, to be processed sequentially. Fast logic paths, retention circuits, and SRAM need a family of threshold options. Changing threshold without substantially changing device dimensions, capacitance or fabrication complexity is valuable. FinFET processes typically use different work-function-metal stacks. In a four-ribbon GAA stack, the narrow sheet-to-sheet gap limits how much WFM can fit around each channel. La in the gate dielectric creates interfacial dipoles at the SiOx/HfOx boundary, shifting effective work function and tuning threshold voltage. This gives Intel another control alongside its NMOS and PMOS WFM stacks. Low-threshold devices improve critical-path drive; higher thresholds reduce leakage elsewhere. Dipole tuning is especially useful in GAA because it changes threshold without consuming the narrow intersheet gap with thicker WFM. Precise control of La incorporation, diffusion and interface quality has long been a challenge, limiting viability in high volume production but is now seen from every leading-edge foundry. Intel’s patent describes depositing a dipole-forming oxide above HfOx and annealing it toward the interfacial oxide before completing the work-function and fill metals. This separates threshold tuning from the space available for metal. Newer research addresses the thermal cost: imec’s 2026 dipole-middle research inserts the shifter between two HfOx depositions, shortening the diffusion path while protecting SiOx during patterning. [20], [21]Matched-cut EDS, Intel 18A (left) vs. Samsung SF2 (right). Intel retains raised source/drain epi beneath its contacts, while Samsung recesses W deep into the epi to form a V-shaped Ti-lined interface. The deeper contact increases metal-to-semiconductor area and shortens the current path from the lower sheets, reducing contact and spreading resistance. It also removes epi volume and brings the contact etch closer to the channel ends. Retaining more epi preserves the material available for strain transfer, especially from SiGe into PMOS. These geometries balance contact access against stress engineering and etch margin. [22], [23] Samsung stacks three sheets to Intel’s four ribbons, and both processes use sheet width to tune drive strength. In our Samsung cross-sections, widths range roughly from 19 to 30 nm in the NPU rows and 37 to 50 nm in the CU cell. The Samsung nanosheets taper, with the widest sheet at the bottom and the narrowest at the top. Both processes use HfOx gate dielectric and Ti-based work-function stacks, with Al in the NMOS stack. In the Samsung devices shown here, the dielectric and WFM occupy the intersheet gaps, leaving W above the top sheet. Intel’s PMOS stack leaves more room between ribbons, and W fills those gaps while the thicker NMOS stack leaves W mainly in the upper trench. Gate-stack EDS maps. The W between Intel’s PMOS ribbons provides a conductive path close to the lower gates. Where WFM fills the entire gap, the gate still surrounds the channel, but voltage reaches it through the more resistive work-function films. Thinner WFM and dipole tuning preserve room for low-resistivity fill; Mo and Ru are alternative fill metals being developed for further scaling. [24] A masked, sequential WFM flow explains the different gate heights and inter-nanosheet fill. The proposed sequence below shows how separate NMOS and PMOS work-function steps produce that geometry. Enabled by the BSPDN process, Intel replaces the dense-logic silicon subfin with dielectric, removing the parasitic conduction path below the ribbons and reducing substrate-related capacitance. A retained silicon body as in classical, non-SOI, planar and FinFET designs needs junction and punchthrough-stop engineering to suppress leakage. Dielectric isolation makes that leakage less sensitive to the subfin doping profile but adds removal and fill steps. It also weakens the direct thermal path through silicon, making the contacts, metal stacks and package more important for heat extraction. [24], [26] Fluorine is concentrated around selected Intel device structures in the maps. WF6 is a standard precursor for W nucleation and fill, while barrier films protect adjacent dielectrics from fluorine attack. Low-fluorine W processes reduce the residual-F burden. Chloride-based precursors avoid introducing F during W deposition, but require control of chlorine attack, nucleation and fill quality. The integration target is a continuous, low-resistance W path with a thin protective liner and minimal chemical damage to the surrounding stack. [20], [27], [28] The SemiAnalysis STEEL teardown lab breaks down advanced datacenter and AI hardware. To learn more about our pipeline or to commission a teardown, contact sales@semianalysis.com. WE’RE HIRING: Architecture, floorplan, packaging, manufacturing, and labs experts. Opportunities from system to transistor and everywhere in between. Check out our Careers page. We measured cell height, gate pitch, metal geometry, and ribbon dimensions at the XTEM sites shown below. The tables group these dimensions by site and device polarity. Our “sheet cuts” cross the silicon channel and show the ribbons end-on. “gate cuts” run along the channel through successive gates. The 18A logic cell dimensions point to a five-track logic library while the N3E and Intel 3 cell dimensions evidence a seven-track logic library. The DDR-PHY uses wider M0 wires and much larger spacing than core logic. That trades routing density for lower wire resistance and weaker coupling between neighboring nets. The geometry suits the current delivery and coupling requirements of analog, clock, and I/O circuitry. PowerVia lets 18A combine a compact cell height with wider M0 geometry by moving the main power rails off the signal-routing tracks. That relaxes local wire scaling while preserving a small cell footprint. Cell height and gate pitch set the geometric density; pin access and routability determine how much of it a real block can use. [29] The biggest takeaway from our gate-pitch measurements is that Intel 18A compute logic and TSMC N3E GPU logic have similar density in the Bohr representative-cell model. The 18A example is 18.6% denser than the Intel 3 GPU example. Gate pitches are nearly identical across the three sites, so cell height drives most of the difference. The Bohr model combines a four-transistor NAND2 spanning three gate pitches and a 32-transistor scan flip-flop (SFF) spanning nineteen pitches, weighting their densities 60:40. The sensitivity column shows how independently changing cell height and gate pitch by ±1 nm changes the result. This compares representative cell geometries; whole-die density also depends on cell mix and placement. The 18A P-core gives M0 substantially more metal cross section than the N3E vector engine. Treating each profile as a trapezoid gives 2.63 times the area per line and 1.84 times the area after normalization by routing pitch. The larger section reduces the geometric contribution to line resistance and lowers current density for a given current. Taller and wider wires also add capacitance, so circuit delay depends on the balance of resistance and capacitance. The DDR-PHY has less metal area per routing width than the 18A core fields, while remaining above N3E. [30] Area = height × (top CD + bottom CD) / 2, including liners. Area/pitch normalizes by routing width. Taper is the symmetric sidewall angle from vertical, with the largest angle belonging to the DDR-PHY. Compute tile The measurements show how ribbon dimensions and gate-stack geometry vary across the compute tile and between NMOS and PMOS to balance channel drive, gate load and the space needed for the dielectric/WFM stack across logic, SRAM and the DDR-PHY. Width mainly changes available channel perimeter; thickness also changes electrostatic control and carrier confinement. Gate-stack thickness then determines the space left for low-resistivity fill P-core and LP E-core logic Both the P-core and LP E-core use multiple nanosheet widths. Widths are measured on high-magnification XTEMs while wider-field images demonstrate additional width choices within the LP E-core. Multiple widths are expected even within an LP E-core. Timing-critical paths, buffers and cells with different fanout need different drive strengths. The lower-magnification fields show this width diversity beyond the sites quantified in the table. L2 and L3 SRAM GAA gives SRAM designers another way to balance the pull-up (PU), pass-gate (PG), and pull-down (PD) transistors. FinFET bitcells set device strength through fin count while GAA adds nanosheet width as a sizing knob. In a 6T SRAM cell, a strong pull-down relative to the pass-gate limits read disturbance, while a strong pass-gate relative to the pull-up improves writability. During a write, the pass-gate and write driver pull the node storing “1” below the inverter trip point. During a read, the pull-down holds the node storing “0” low. Bias, threshold voltage, mismatch and assist circuitry set the remaining margin. FinFET high-current cells commonly use a PU:PG:PD fin-count pattern of 1:2:2, a device-sizing ratio rather than a current ratio. Ribbon width lets Intel balance SRAM strengths without adding whole fins. The L2 cell uses its narrowest ribbons for PU and widest for PD, improving writability and read stability respectively. Intel’s disclosed HCC operates without assist; its denser HDC uses negative-bitline write assist. Pulling the selected bitline briefly below ground increases pass-gate overdrive so it can overpower the pull-up at lower supply voltage. That buys density and low voltage writability at the cost of boosting circuitry, switching energy, and additional voltage stress that must be controlled. [31], [32] Four rectangular ribbons give the perimeter = 8 × (width + thickness), before corner rounding. PG/PU is 1.49 and PD/PG is 1.16. The L3 structures closely resemble L2 in layout and cell height. Fewer L3 nanosheet widths are tabulated because fewer high-magnification images were available. DDR PHY The DDR-PHY trades density for controlled analog behavior and reliable off-chip signaling. It contains drivers, receivers, delay circuits, and calibration logic that set drive strength, sampling time, and voltage margin. Repeated four-sheet devices with similar widths fit the use of regular transistor units for matching and programmable drive. Its wider local wiring provides room for current delivery and separation of sensitive signals, while consuming more area than a dense core-logic grid. The layout serves the memory channel’s electrical requirements as well as digital logic density. [33] The SemiAnalysis STEEL teardown lab breaks down advanced datacenter and AI hardware. To learn more about our pipeline or to commission a teardown, contact sales@semianalysis.com. WE’RE HIRING: Architecture, floorplan, packaging, manufacturing, and labs experts. Opportunities from system to transistor and everywhere in between. Check out our Careers page. Intel 3 GPU devices Vector engine logic Intel 3’s XVE logic uses two-fin PMOS and NMOS devices with power rails in M0. Its cell height and M0 pitch give a seven-track geometry, two tracks more than the 18A logic. One-fin groups also appear among the two-fin devices. Intel 3 L2 SRAM The Intel 3 L2 SRAM uses the familiar HCC sizing pattern: one PU fin, two PG fins, and two PD fins. N3E GPU devices Vector engine logic The N3E XVE field contains repeated two-fin devices with seven-track cell geometry. N3E remains a FinFET process, giving Panther Lake a direct FinFET-to-RibbonFET comparison. N3E L2 SRAM The N3E L2 SRAM uses the same PU:PG:PD fin-count pattern of 1:2:2. Panther Lake-U follows Lunar Lake’s floorplan quite closely. Both pair 4 P-cores with 4 LP E-cores and NPU, media and display engines in similar locations. Lunar Lake also uses Xe2, the direct predecessor to Panther Lake’s Xe3 GPU. This makes Lunar Lake the most direct basis for our comparisons. Arrow Lake differs in core count and uses the older Xe-LPG GPU architecture, so we only use it where it offers a more direct component-level comparison. Compute tile Panther Lake compute-tile floorplans remain sparse even months after launch. Intel 18A’s backside metal and dielectric stack must be removed without damaging the underlying structures before a clean transistor-level floorplan can be imaged. Most published die shots hide or heavily process the background, but we are quite proud of the die shot we achieved and are excited to show the work we have done. We measured the areas of the key components on the compute tile and compared them with their Lunar Lake predecessors on TSMC N3B. These help us to capture changes in block area and compare the two chips across process nodes and designs. Our total tile areas exclude the scribe-line area. The compute-plus-GPU subtotal below uses the PTL-U compute tile and GT1 GPU; it excludes the I/O tile and passive base. Individual block areas use the boundaries marked on the floorplans The compute-plus-GPU row is recomputed from the displayed PTL-U and GT1 areas. Component rows use their stated per-region counts and are not an additive partition of the whole tile. The P-core area remains almost unchanged between Lunar Lake and Panther Lake, despite L2 capacity increasing from 2.5 MiB to 3 MiB. Arrow Lake uses the same Lion Cove core as Lunar Lake but also has a 3 MiB L2. Cougar Cove fits 20% more L2 into the same P-core area. The larger private cache keeps more of each core’s working set close to its execution units, reducing access to shared L3 and DRAM. Extra capacity adds storage leakage and lookup energy, so designers balance it against avoided lower-level accesses. The shared P-core L3 cache also shrank by 14.8%. [2] Cougar Cove combines a similar footprint with Intel’s reported power-efficiency improvements. RibbonFET’s tighter channel control reduces leakage, while PowerVia reduces supply droop and allows tighter voltage guardbands. [1] Darkmont’s four-core LP E-core cluster is 5.0% smaller than Skymont’s on Lunar Lake, with most of the reduction in its L2 regions. The 1 MiB region shrank by 8.4% and the 1.5 MiB region by 14.9%. The tag arrays also use one fewer visible row. Tags identify which memory addresses the data array holds, so rearranging them changes the cache’s layout and wiring without requiring less data capacity. [2] The LP E-cores share one L2. This pools capacity and avoids duplicating all the cache machinery, but the four cores contend for its banks and bandwidth. Their separate cluster also keeps light work away from the performance cluster and its L3, allowing that larger domain to sleep. [1], [2] Cache area includes more than the storage cells. Tags identify each line, decoders select rows, sense amplifiers read the small bitline signal, and wires connect to the banks. Splitting an array into smaller sections shortens wordlines and bitlines, improving access speed, but duplicates peripheral circuits. Panther Lake’s smaller cache regions therefore reflect the complete memory implementation, including how much of each region is devoted to storage. [34] Unlike Meteor Lake and Arrow Lake, Panther Lake has no separate SoC tile. The NPU, LP E-cores, memory controllers, PHYs, media and display engines now share the compute tile. This removes an active die and keeps CPU memory traffic on one die. The cost is moving PHY and I/O-related circuitry onto 18A: drivers, receivers and analog circuits must still meet external voltage, loading and signal-integrity requirements, so their area does not shrink like dense digital logic. [1], [2] The biggest shrink comes from the NPU, which occupies 36.9% less area. NPU 5 consolidates the same total INT8 MAC count into half as many neural compute engines. Each of the three NCEs has a larger MAC array to make the complete NCE envelope 22.6% larger than an NPU 4 engine. Consolidation also halves the number of scratchpads and SHAVE DSPs, from 12 to 6. The MAC array handles matrix multiplication and convolution, while SHAVE executes vector and custom operations that fit the array poorly. [1], [2], [35] The paired floorplans identify each NCE envelope and its scratchpad, MAC, and SHAVE regions. Each measured MAC polygon is counted once per NCE in the area accounting below. The scratchpads store weights, activations, and intermediate results near the MAC arrays, allowing repeated use without fetching them again from DRAM. Halving their number delivers the largest measured area saving but leaves less local storage for the same total MAC count. Layers that no longer fit locally require smaller working tiles or more transfers of intermediate data. The benefit depends on keeping the enlarged arrays busy while managing that tighter storage budget. [36] NPU 5 also adds native FP8. Using half the operand width of FP16 reduces storage and transfer demand, helping workloads fit the smaller local memory budget. Lower precision and format-dependent range make scaling and model validation part of deployment. Hardware activation functions further reduce work that would otherwise occupy the programmable DSPs. [1], [2] Microsoft requires an NPU to deliver at least 40 TOPS for Copilot+ PCs. Both Lunar Lake and Panther Lake meet this threshold, but Panther Lake uses significantly less silicon. GPU tiles Panther Lake is Intel’s first product with Xe3, its latest GPU architecture. It offers two different GPU tiles: a smaller GT1 tile with 4 Xe3 cores on Intel 3 and a larger GT2 tile with 12 Xe3 cores on TSMC N3E. Panther Lake allows us to compare the same GPU architecture across both Intel 3 and TSMC N3E. Wildcat Lake adds a third Xe3 implementation on Intel 18A. A future newsletter will detail Xe3 and its implementation differences across all three process nodes. GT2 scales Xe3 to a different physical layout, with render slices arranged vertically instead of GT1’s horizontal arrangement. Slice placement sets the distances to shared cache banks and the D2D interface. Those wires consume area and add delay, so scaling the number of Xe cores also requires a new balance of cache placement, routing and timing. [1] What’s immediately obvious is that the GT2 tile on TSMC N3E has much smaller Xe cores than GT1. These block areas include logic, caches, and routing. An Xe core on the GT1 tile is ~69% larger than one on Lunar Lake, and ~55% larger than one on GT2. Intel 3 therefore uses substantially more area per Xe core. The block-area gap exceeds the measured logic and SRAM density gaps, bringing routing, timing targets, cell mix, and floorplan allocation into the comparison. The measured vector/matrix engine region is almost unchanged between Lunar Lake and Panther Lake’s GT2 tile. Xe3 retains eight 512-bit vector engines and eight 2048-bit XMX engines per core. Its gains also come from feeding those engines more effectively: more resident threads hide stalls, and variable register allocation lets shaders trade registers per thread against the number of threads kept active. [1] The shared L1/SLM capacity increased by 33% from 192 KiB to 256 KiB, while its area increased only 5%, raising effective density by 27%. L1 retains reused cache lines, while software-managed SLM lets a thread group share data locally. Both reduce traffic to more distant memory. Allocating more SLM per group can also limit how many groups reside on a core at once. [1], [37] The GT1 tile carries 4 MiB of L2 against 16 MiB on the GT2 tile. GT1 divides its L2 cache into four 1 MiB banks, while GT2 uses eight 2 MiB banks. Each bank contains 128 macros, but each N3E macro stores 16 KiB, twice the Intel 3 macro’s 8 KiB capacity. The N3E macro is only 54% larger while holding twice as many bits, giving it 30% higher density: ~23.7 Mbit/mm² versus 18.3 Mbit/mm². Including bank-level circuitry, the gap widens to ~16.9 Mbit/mm² on GT2 versus ~10.4 Mbit/mm² on GT1. GT2 gains density with its macros storing more bits per unit area, and those macros occupy more of each cache bank. Larger macros spread decoder and sense-amplifier overhead across more storage, while a more compact bank layout reduces the share spent on control and routing. The compromise is longer wordlines and bitlines that carry more capacitance. [34] I/O tile Panther Lake uses two I/O tile variants, both fabricated on TSMC N6. The smaller one provides 4 PCIe 5.0 and 8 PCIe 4.0 lanes and serves lower-tier systems as well as those without a discrete GPU, while the larger one adds 8 PCIe 5.0 lanes, bringing the total to 20 lanes, for discrete-GPU connectivity. Panther Lake SKUs with the larger 10- or 12-Xe GPUs use the smaller I/O tile. [38] The smaller I/O tile adds a PCIe 4.0 block and a Thunderbolt block to Lunar Lake’s I/O layout, providing four additional PCIe 4.0 lanes and another Thunderbolt 4 port. Its repeated N6 blocks retain nearly identical areas and layouts. Reusing these proven PHYs and controllers avoids porting and requalifying external interfaces on 18A, where faster digital logic offers less benefit to circuits constrained by the off-chip link. [38] SemiAnalysis’s teardown lab (STEEL) dives deep into the world’s advanced datacenter and AI hardware. To learn more about our pipeline or to commission a teardown, contact sales@semianalysis.com. We’re hiring technical experts from system to transistor and everywhere in between. Check out our Careers page. Panther Lake offers scalability and modularity through its disaggregated packaging that partition compute, GPU, and I/O silicon into separate tiles allowing for a suite of tile configurations. This partitioning makes the package part of Intel’s node economics as it determines how much leading-edge wafer area each product consumes, which functions can remain on other processes, and how much configuration freedom Intel can offer from a shared set of tiles. Furthermore, fabricating the compute and GPU tiles separately confines the new 18A process to the compute tile and allows graphics and I/O to use other, more established, and more cost-effective processes. For Panther Lake, the GPU and I/O tiles are assembled alongside the compute tile on a passive silicon base using Foveros-S. Intel’s current technology brief lists a nominal 36 µm pitch for Foveros-S. Through-silicon vias (TSVs) in the base connect the fine wiring above to the larger package connections below. The functional tiles sit side by side on that passive base in a 2.5D configuration. [39] Our cross-section through the compute and GPU tiles shows the package’s wiring hierarchy. Microbumps connect each active tile to the passive silicon base; its fine redistribution layer (RDL) carries the short, dense tile-to-tile links. TSVs carry connections through the base to the package substrate, which fans them out to the much coarser motherboard solder joints. The base supplies interconnect, while computation remains in the active tiles above it. [39] At the compute-tile edge, the higher-magnification inset shows a local microbump spacing of approximately 25.24 µm and a feature width of 12.33 µm. These local spacings are finer than Intel’s nominal Foveros-S value. The X-ray fields further confirm tighter neighboring bumps, consistent across every die-to-die area found on each tile. Additional X-ray analysis is offered after the paywall. Putting the memory controller beside the CPU removes the D2D transfer that CPU memory requests required in Meteor Lake and Arrow Lake. This avoids the extra transmitter, receiver, and link traversal, saving interface energy and latency. Panther Lake’s separate GPU still crosses a D2D link to reach DRAM, so its larger local caches also help contain package traffic. [1], [40] Smaller dies are less likely to contain a random fatal defect, and screening them before assembly prevents one bad tile from consuming a complete package of good silicon. Reuse also spreads design and qualification work across more products. Against those gains, Intel pays for the passive base, D2D circuits, extra bonding and test steps, and losses during assembly. Cost per working product across the portfolio captures the combined effect of wafer yield, reuse, test, and assembly. [29] Wildcat Lake packaging Intel launched Core Series 3, formerly Wildcat Lake, on 16 April 2026 for value mobile and edge systems. Wildcat Lake keeps 18A but removes the passive base and combines more functions on one die to simplify the package. The two products therefore reveal two distinct ways to commercialize the same leading-edge process. [41] Wildcat Lake’s 18A die combines up to two Cougar Cove P-cores, four Darkmont LP E-cores, two Xe3 cores and a smaller NPU. A separate platform-controller die supplies I/O, connected through UCIe, Intel’s first processor implementation of the standard. Consolidating graphics remove a tile boundary and the passive base, reducing assembly complexity for a modest-bandwidth value product. It also ties CPU and graphics scaling to the same die, giving up Panther Lake’s ability to swap in a much larger GPU. [42], [43] In July 2021, Intel CEO Pat Gelsinger set out an ambitious process roadmap aimed at regaining performance leadership by 2025, later described as five nodes in four years. Five years and one CEO later, Intel’s comeback story is not as unambiguously positive as Pat may have hoped. [44], [45] Intel once set the pace for process technology, bringing high-k metal gate technology and FinFETs into volume production years ahead of the rest of the industry. Its 22 nm FinFET process reached consumers with Ivy Bridge in 2012. [46] Intel’s integrated device manufacturing (IDM) model allowed its architects and process engineers to co-optimize products and processes. Starting with Sandy Bridge, Intel dominated x86, while AMD struggled with Bulldozer. That lead faltered at 14 nm and broke at 10 nm. Intel targeted a massive 2.7× density increase, but the node arrived years late and required several revisions before it could support Intel’s full lineup. This delay forced Intel to stretch 14 nm across six generations, while TSMC moved ahead in process technology and AMD recovered in x86. By 2019, Intel was still shipping 14 nm across most of its product stack, with its 10 nm client ramp focused on Ice Lake mobile processors. Meanwhile, TSMC was shipping N7 and N7+, and AMD’s Zen 2 compute chiplets used N7 to raise core counts and improve efficiency. Intel’s process failures were central to its decline, but unsound business decisions furthered their downward slide. Product delays compounded product mistakes, pushing client, server, and FPGA roadmaps off schedule. Several attempts to enter AI (Nervana and Gaudi) and networking (Tofino) also failed to establish lasting businesses. Intel’s recovery has focused on consumer CPUs and advanced packaging. Tiger Lake, Alder Lake, Lunar Lake and now Panther Lake have restored Intel’s consumer roadmap. On the process side, Intel 4 shipped with Meteor Lake, Intel 3 with Granite Rapids and Sierra Forest, and Intel 18A with Panther Lake. Intel has also made advanced packaging part of its foundry offering. However, Intel is still playing catch-up in servers. Several Xeon generations arrived years late and trailed contemporary AMD and Arm server CPUs in performance, efficiency, and core count. The process roadmap is back, but Intel does not hold the same process-technology leadership position it held prior to 10 nm. The introduction of gate-all-around nanosheets and backside power delivery are two of the biggest changes to transistor integration in a decade. Intel took on both changes at once: 18A paired its first RibbonFET with PowerVia in Panther Lake. Panther Lake is a substantial manufacturing milestone. Our cross-sections show how RibbonFET and PowerVia reshape local contacts and wiring, while the floorplans show where architectural consolidation and process choices save area. A sustained competitive lead depends on product performance, cost, yield, and the next implementation. The SemiAnalysis STEEL teardown lab breaks down advanced datacenter and AI hardware. To learn more about our pipeline or to commission a teardown, contact sales@semianalysis.com. WE’RE HIRING: Architecture, floorplan, packaging, manufacturing, and labs experts. Opportunities from system to transistor and everywhere in between. Check out our Careers page.. [1] Intel, “Intel Tech Tour 2025 Panther Lake recap,” Doc. 866361, 2025. Accessed: Sep. 14, 2026. [Online]. Available: https://cdrdv2-public.intel.com/866361/ITT_2025_Panther_Lake_Recap1.pdf [2] Intel, “Core Ultra processors Series 3 datasheet, volume 1,” Doc. 872188-001, Jan. 2026. Accessed: Sep. 14, 2026. [Online]. Available: https://cdrdv2-public.intel.com/872188/872188-001.pdf [3] N. Horiguchi and E. Beyne, “Backside power delivery: How to power chips from the backside,” imec Reading Room, Nov. 25, 2022. Accessed: Sep. 14, 2026. [Online]. Available: https://www.imec-int.com/en/articles/how-power-chips-backside [4] W. Hafez et al., “Intel PowerVia technology: Backside power delivery for high density and high-performance computing,” in Proc. IEEE Symp. VLSI Technol. Circuits, Kyoto, Japan, Jun. 2023, pp. 1–2, doi: 10.23919/vlsitechnologyandcir57934.2023.10185208. [5] K. Fischer et al., “Intel 18A platform technology featuring RibbonFET (GAA) and PowerVia for advanced high-performance computing,” in Proc. Symp. VLSI Technol. Circuits, Kyoto, Japan, Jun. 2025, pp. 1–3, doi: 10.23919/vlsitechnologyandcir65189.2025.11075006. [6] K.-F. Cheng, C.-L. Teng, H.-Y. Huang, and H.-C. Chen, “Metal oxide composite as etch stop layer,” U.S. Patent 12 176 247 B2, Dec. 24, 2024. Accessed: Sep. 14, 2026. [Online]. Available: https://patents.google.com/patent/US12176247B2/en [7] S. W. King, “Dielectric barrier, etch stop, and metal capping materials for state of the art and beyond metal interconnects,” ECS J. Solid State Sci. Technol., vol. 4, no. 1, pp. N3029–N3047, 2015, doi: 10.1149/2.0051501jss. [8] A. V. Mule’, D. J. Towner, D. Seghete, C. R. Ryder, and A. A. Gonzalez, “Multi-layer etch stop layers for advanced integrated circuit structure fabrication,” U.S. Patent Appl. 20220102343 A1, Mar. 31, 2022. Accessed: Sep. 14, 2026. [Online]. Available: https://patents.google.com/patent/US20220102343A1/en [9] C.-C. Wang and J. H. Wang, “Interconnect structure with dielectric cap layer and etch stop layer stack,” U.S. Patent Appl. 20230253247 A1, Aug. 10, 2023. Accessed: Sep. 14, 2026. [Online]. Available: https://patents.google.com/patent/US20230253247A1/en [10] Y. J. Wu, K.-F. Cheng, C.-C. Lee, H.-K. Chang, and H.-Y. Huang, “Etch stop layer for interconnect structures,” U.S. Patent Appl. 20240332070 A1, Oct. 3, 2024. Accessed: Sep. 14, 2026. [Online]. Available: https://patents.google.com/patent/US20240332070A1/en [11] M. G. Rainville, N. Shankar, K. S. Reddy, and D. M. Hausmann, “Deposition of aluminum oxide etch stop layers,” U.S. Patent Appl. 20180197770 A1, Jul. 12, 2018. Accessed: Sep. 14, 2026. [Online]. Available: https://patents.justia.com/patent/20180197770 [12] J. S. Leib et al., “Conductive lines having molybdenum liner and tungsten fill for advanced integrated circuit structure fabrication,” U.S. Patent Appl. 20240429126 A1, Dec. 26, 2024. Accessed: Sep. 14, 2026. [Online]. Available: https://patents.justia.com/patent/20240429126 [13] M. Naik, “Cobalt enables power and performance scaling at single-digit logic nodes,” Applied Materials, Dec. 17, 2018. Accessed: Sep. 14, 2026. [Online]. Available: https://www.appliedmaterials.com/us/en/blog/blog-posts/cobalt-enables-power-and-performance-scaling-at-single-digit-logic-nodes.html [14] Lam Research, “Breaking through AI’s invisible barrier with molybdenum,” Lam Research Newsroom, May 6, 2025. Accessed: Sep. 14, 2026. [Online]. Available: https://newsroom.lamresearch.com/molybdenum-metallization-ai-revolution [15] P. Yashar, G. Malyavanatham, and H. Vijwani, “Integrated circuit interconnect structures with niobium barrier materials,” U.S. Patent Appl. 20240112951 A1, Apr. 4, 2024. Accessed: Sep. 14, 2026. [Online]. Available: https://patents.google.com/patent/US20240112951A1/en [16] Applied Materials, “Applied Materials unveils chip wiring innovations for more energy-efficient computing,” Applied Materials Investor Relations, Jul. 8, 2024. Accessed: Sep. 14, 2026. [Online]. Available: https://ir.appliedmaterials.com/news-releases/news-release-details/applied-materials-unveils-chip-wiring-innovations-more-energy [17] Samsung, “Samsung showcases AI-era vision and latest foundry technologies at SFF 2024,” Samsung Semiconductor, Jun. 13, 2024. Accessed: Sep. 14, 2026. [Online]. Available: https://semiconductor.samsung.com/news-events/news/samsung-showcases-ai-era-vision-and-latest-foundry-technologies-at-sff-2024/ [18] C. W. Yeung et al., “Channel geometry impact and narrow sheet effect of stacked nanosheet,” in Proc. IEEE Int. Electron Devices Meeting (IEDM), Dec. 2018, pp. 28.6.1–28.6.4, doi: 10.1109/iedm.2018.8614608. [19] S. Mochizuki et al., “Evaluation of (110) versus (001) channel orientation for improved nFET/pFET device performance trade-off in gate-all-around nanosheet technology,” in Proc. Int. Electron Devices Meeting (IEDM), Dec. 2023, pp. 1–4, doi: 10.1109/iedm45741.2023.10413854. [20] D. G. Ouellette et al., “Fabrication of gate-all-around integrated circuit structures having molybdenum nitride metal gates and gate dielectrics with a dipole layer,” U.S. Patent 12 051 698 B2, Jul. 30, 2024. Accessed: Sep. 14, 2026. [Online]. Available: https://patents.google.com/patent/US12051698B2/en [21] imec, “Advancing the CFET-based device roadmap with novel integration modules and standard-cell architectures,” imec, 2026. Accessed: Sep. 14, 2026. [Online]. Available: https://www.imec-int.com/en/articles/advancing-cfet-based-device-roadmap-novel-integration-modules-and-standard-cell [22] A. Reznicek, X. Miao, C. Lee, and J. Zhang, “Wrap around contact for nanosheet source drain epitaxy,” U.S. Patent 11 302 813 B2, Apr. 12, 2022. Accessed: Sep. 14, 2026. [Online]. Available: https://patents.google.com/patent/US11302813B2/en [23] J. A. Kittl, J. G. Hong, D. R. Palle, and M. S. Rodder, “Methods for varied strain on nano-scale field effect transistor devices,” U.S. Patent 9 978 833 B2, May 22, 2018. Accessed: Sep. 14, 2026. [Online]. Available: https://patents.google.com/patent/US9978833B2/en [24] S. Gandikota et al., “Method of reducing metal gate resistance for next generation NMOS device application,” Int. Patent Appl. WO 2024/137272 A1, Jun. 27, 2024. Accessed: Sep. 14, 2026. [Online]. Available: https://patents.google.com/patent/WO2024137272A1/en [25] J. Zhang et al., “Full bottom dielectric isolation to enable stacked nanosheet transistor for low power and high performance applications,” in Proc. IEEE Int. Electron Devices Meeting (IEDM), Dec. 2019, pp. 11.6.1–11.6.4, doi: 10.1109/iedm19573.2019.8993490. [26] C. Yoo, J. Chang, Y. Seon, H. Kim, and J. Jeon, “Analysis of self-heating effects in multi-nanosheet FET considering bottom isolation and package options,” IEEE Trans. Electron Devices, vol. 69, no. 3, pp. 1524–1531, Mar. 2022, doi: 10.1109/ted.2022.3141327. [27] S. S. Pradhan, D. B. Bergstrom, J.-S. Chun, and J. Chiu, “Tungsten gates for non-planar transistors,” European Patent Appl. EP 3 506 367 A1, Jul. 3, 2019. Accessed: Sep. 14, 2026. [Online]. Available: https://patents.google.com/patent/EP3506367A1/en [28] L. Schloss and X. Ba, “Tungsten films having low fluorine content,” U.S. Patent 9 754 824 B2, Sep. 5, 2017. Accessed: Sep. 14, 2026. [Online]. Available: https://patents.google.com/patent/US9754824B2/en [29] Intel Foundry, “Accelerating AI and HPC with advanced process and packaging technologies,” Doc. 367370-001US, Jun. 2026. Accessed: Sep. 14, 2026. [Online]. Available: https://www.intel.com/content/dam/www/central-libraries/us/en/documents/2025-11/intel-foundry-hpc-ai-brief.pdf [30] M. Lofrano, X. Chang, H. Oprins, and Z. Tokei, “Mitigating the thermal bottleneck in advanced interconnects,” imec, Sep. 28, 2023. Accessed: Sep. 14, 2026. [Online]. Available: https://www.imec-int.com/en/articles/mitigating-thermal-bottleneck-advanced-interconnects [31] S. Bair, “Intel at ISSCC 2025: Navid Shahriari invited talk, eight papers, forums, panelist & product details,” Intel Community, Feb. 19, 2025. Accessed: Sep. 14, 2026. [Online]. Available: https://community.intel.com/t5/Blogs/Tech-Innovation/Edge-5G/Intel-at-ISSCC-2025-Navid-Shahriari-Invited-Talk-Eight-Papers/post/1667592 [32] D. Chandra, E. Potladhurthi, D. R. S. Reddy, and K. S. Rengarajan, “Tunable negative bitline write assist and boost attenuation circuit,” U.S. Patent Appl. 20160203857 A1, Jul. 14, 2016. Accessed: Sep. 14, 2026. [Online]. Available: https://patents.google.com/patent/US20160203857A1/en [33] Synopsys, “Advantages of firmware-based training in high-speed DDR IP,” Synopsys IP Technical Bulletin. Accessed: Sep. 14, 2026. [Online]. Available: https://www.synopsys.com/articles/firmware-based-training-ddr-ip.html [34] N. Muralimanohar, R. Balasubramonian, and N. P. Jouppi, “CACTI 6.0: A tool to understand large caches,” Hewlett-Packard Laboratories, 2009. Accessed: Sep. 14, 2026. [Online]. Available: https://users.cs.utah.edu/~rajeev/cacti6/cacti6-tr.pdf [35] Intel, “Intel Tech Tour 2024 Lunar Lake AI hardware accelerators,” Doc. 824436, 2024. Accessed: Sep. 14, 2026. [Online]. Available: https://cdrdv2-public.intel.com/824436/2024_Intel_Tech%20Tour%20TW_Lunar%20Lake%20AI%20Hardware%20Accelerators.pdf [36] Y.-H. Chen, J. Emer, and V. Sze, “Eyeriss: A spatial architecture for energy-efficient dataflow for convolutional neural networks,” in Proc. 43rd ACM/IEEE Annu. Int. Symp. Comput. Archit. (ISCA), Jun. 2016, pp. 367–379, doi: 10.1109/isca.2016.40. [37] Intel, “Introduction to the Xe-HPG architecture,” Doc. 758306, Nov. 4, 2022. Accessed: Sep. 14, 2026. [Online]. Available: https://www.intel.com/content/www/us/en/developer/articles/technical/introduction-to-the-xe-hpg-architecture.html [38] Intel, “Core Ultra processors Series 3 for the edge,” Doc. 855291. Accessed: Sep. 14, 2026. [Online]. Available: https://cdrdv2-public.intel.com/855291/Intel%C2%AE%20Core%E2%84%A2%20Ultra%20Processors%20Series%203%20for%20Edge%20Overview_2.pdf [39] Intel Foundry, “Foveros technology brief,” Doc. 366411-001US, Feb. 2026. Accessed: Sep. 14, 2026. [Online]. Available: https://www.intel.com/content/dam/www/central-libraries/us/en/documents/2025-07/foveros-25d-product-brief.pdf [40] D. Das Sharma, “UCIe: Building an open chiplet ecosystem,” UCIe Consortium, 2022. Accessed: Sep. 14, 2026. [Online]. Available: https://www.uciexpress.org/_files/ugd/0c1418_c5970a68ab214ffc97fab16d11581449.pdf [41] Intel, “Intel launches Intel Core Series 3 processors,” Intel Newsroom, Apr. 16, 2026. Accessed: Sep. 14, 2026. [Online]. Available: https://www.intel.com/content/www/us/en/newsroom/news/client-computing/intel-launches-intel-core-series-3-processors-changing-the-game-for-everyday-computing.html [42] Intel, “Core Series 3 launch press deck,” Apr. 2026. Accessed: Sep. 14, 2026. [Online]. Available: https://download.intel.com/newsroom/2026/Intel-Core-Series-3/Intel-Core-Series-3-Launch-Press-Deck.pdf [43] Intel, “Intel outlines architectures for agentic AI at Hot Chips 2026,” Intel Newsroom, Aug. 24, 2026. Accessed: Sep. 14, 2026. [Online]. Available: https://www.intel.com/content/www/us/en/newsroom/news/client-computing/intel-outlines-architectures-for-agentic-ai-at-hot-chips-2026.html [44] Intel, “Intel accelerates process and packaging innovations,” Intel Investor Relations, Jul. 26, 2021. Accessed: Sep. 14, 2026. [Online]. Available: https://www.intc.com/filings-reports/all-sec-filings/content/0001193125-21-224438/d199788dex991.htm [45] Intel, “Intel reports third-quarter 2021 financial results,” Intel Investor Relations, Oct. 21, 2021. Accessed: Sep. 14, 2026. [Online]. Available: https://www.intc.com/news-events/press-releases/detail/1505/intel-reports-third-quarter-2021-financial-results [46] Intel, “3rd generation Intel Core processors bring exciting new experiences and fun to the PC,” Intel Investor Relations, Apr. 23, 2012. Accessed: Sep. 14, 2026. [Online]. Available: https://www.intc.com/news-events/press-releases/detail/577/3rd-generation-intel-core-processors-bring-exciting
Olivia Rodrigo, Charlie Puth, Sombr, Raye and Katseye set for 2026 iHeartRadio Jingle Ball tour
iHeartRadio is loading its 2026 holiday tour with some starry acts, including Olivia Rodrigo, Charlie Puth and the Jonas Brothers.
Olivia Rodrigo, Charlie Puth, Sombr, Raye and Katseye set for 2026 iHeartRadio Jingle Ball tour
iHeartRadio is loading its 2026 holiday tour with some starry acts, including Olivia Rodrigo, Charlie Puth and the Jonas Brothers | WBBM Newsradio 780 AM & 105.9 FM
Olivia Rodrigo, Charlie Puth, Sombr, Raye and Katseye set for 2026 iHeartRadio Jingle Ball tour
iHeartRadio is loading its 2026 holiday tour with some starry acts, including Olivia Rodrigo, Charlie Puth and the Jonas Brothers | WBEN
Microsoft patent describes in-game advertising triggered by boss fights, loading screens, and cutscenes
The filing points to a possible ad-supported model for PC and console games that is less disruptive than the approaches commonly used in mobile titles.
The Big Holes in Wearable Heart Rate Variability And Readiness Scores
On September 9th, Apple announced it was revamping its Apple Watch Health Sensing System, rolling out a Readiness score (0 to 10), and increasing the frequency of heart rate variability (HRV) outputs 24-fold. This can be viewed as upping its competition with various other consumer wearable sensors. These “readiness scores” as a composite of multiple metrics, with heart rate variability (HRV) being front and center for most. The majority of Americans are now using wearable sensors, which equates to well over 100 million adults. HRV and Readiness scores are increasingly being marketed as a measurement of autonomic nervous system health, a digital marker for future disease, a clock for biological age, and a holistic metric to promote healthspan and even longevity. (Apple also introduced a new longevity tab and “Health Age”.) None of this has been proven. In this edition of Ground Truths I am going to review what we know about heart rate variability and readiness scores. Heart Rate Variability HRV is the variation in normal heart cycle timing. The variability of the heart rate, the barely perceptible millisecond changes in time between consecutive heart beats (see R-R intervals in the Figure below, left panel), is due to interplay between the sympathetic and parasympathetic (vagal nerve) inputs. Distinct from heart rate, individuals with the same heart rate can have very different HRVs. It is a rough reflection of the autonomic nervous system (ANS) activity, inadequate to say whether a person’s ANS function is abnormal. For more than three decades, heart rate variability (HRV) has been measured and several studies have found an association of low HRV and clinical outcomes, particularly a link with higher all-cause and cardiovascular mortality. There have also been less well established links of low HRV to risk of early cognitive impairment, dementia, mental illness, Type 2 diabetes and substance abuse. An important reminder is that HRV is a surrogate marker without any established cause-and-effect relationship. If you increase your HRV, that doesn’t mean it will improve health outcomes. In fact, there is no hard evidence for that. All that work linking to health outcomes was done with electrocardiogram (ECG) derived HRV. Now, in the era of consumer wearables, this is getting assessed differently, by optical pulse (yes, the lights you see) plethysmography (PPG) or what is called pulse rate variability (PRV). They are not the same, as shown below (right panel) and only concordant when the delay between the ECG and pulse is kept constant, which basically means at rest. I should mention there’s also what I will call MPV, a mattress mechanical movement sensor, a derived heart rate variability, that companies like Eight Sleep use, even further away from directly measuring HRV. HRV has not one uniform measurement but many different types of quantification, such as RMSSD, the magnitude of difference between successive R-R intervals of normal sinus beats (N-N) or SDNN, the standard deviation of NN intervals, both in milliseconds. SDNN is one of the so-called frequency domain HRVs (others are LF, HF, LF/HF). Different wearable sensors use different metics; Apple has relied on SDNN and nearly all of the others use RMSSD, which is generally considered the more accurate metric. There’s also the different length of time measured, such as for a matter of minutes, all day, or an overnight’s sleep. Short measurements are especially problematic since they don’t capture enough of respiratory modulation and other factors that influence HRV. Share Ground Truths How well does HRV correlate with PRV? There are very limited studies, especially independently done. One that is commonly cited was conducted by Air Force researchers in only 13 healthy adults assessing Oura ring 3 and 4, Whoop 4.0, and Garmin Fenix 6 and showed a correlation coefficient of 0.88 to 0.97 and a mean absolute percentage error from 6 to 10%. The correlation is not a perfect 1.0, but there’s at least a fairly high level of correlation. HRV is supposed to increase during the night due to takeover of the parasympathetic nervous system, and higher during deep sleep. A recent example of my 1 week, all day “HRV,” and one during sleep is shown below. As you can see, the N of 1 data are inconsistent for the same days from different sensors (Oura, AppleWatch, Fitbit Air, and Eight Sleep) by patterns, absolute numbers, and comparison with prior days and weeks. The largest study in over 8 million Fitbit users (the old version, not Google Fitbit Air, introduced in May 2026) gives you a sense of the effect of age, sex, and the 2 different main HRV (here PRV) metrics, with RMSSD on the left and SDRR (=SDNN) on the right below. That study, from data collected in 2018, is a major outlier, since all the more recent ones are tiny with respect to sample size. Many of the companies have not had independent evaluation of their HRV, such as Eight Sleep, but have published a low standard error on their website. There are some other published studies on the correlation between HRV and PRV, but they are all small and only in healthy adults. A scoping review emphasized the lack of study in underrepresented individuals, including the aged, people of color (which affects the PPG signal), and individuals who are underweight or obese. Add the typical adult age 60 plus with one or more chronic diseases. For example, one study in over 900 adults found poor correlation of HRV and PRV, non-uniformly underestimated across many chronic diseases (cardiovascular, endocrine, neurological, respiratory, and others), concluding PRV is “an invalid surrogate for HRV.” A recent systematic review of 43 studies comparing HRV and PRV found reasonable pooled absolute standardized error (HRV as gold standard) but only 10 of the studies provided quantitative synthesis in ideal conditions. Their main conclusion was similarly cautious: “PPG-derived HRV [PRV] should not be regarded as universally interchangeable with ECG-derived HRV across all devices, populations, and recording contexts.” Factors Affecting HRV and PRV That gets me to the long list of factors that affect HRV (and PRV) besides the device, the type of measurement (RMSSD, SDNN or others), the person’s signal, the sensor site, the duration of data capture, if weighting by sleep stage is used, how artifact is processed and corrected. And this list is not complete!: Oura puts out data from their community of users (who input data) on what affects their overnight HRV. The factors currently provided are: no alcohol (increase 8%), melatonin (increase 2%, float tank (increase 2%), wine (decrease 4%), and party (decrease 14%) in overnight HRV. Must be some big parties! Subscribe What is a PRV measurement good for? It has been falsely characterized as an index of “autonomic balance” and a specific indicator of stress. A 2018 review of the studies available for HRV and its relationship to stress, not using any of the current wearables, found that stress can lower HRV. But so can many other factors. The non-specificity of the signal, indexed to the table above, is striking. Evidence from a UK Biobank study of over 46,000 participants with actual HRV looked at genetically predicted HRV, a genetic risk score, that failed to show the expected HRV-mortality link, indicating that _HRV is likely not causa_l, but rather a reflection of person’s physiologic state. A review of consumer wearable HRV data from 5 longitudinal studies showed that nighttime PRV was not associated with perceived stress, and surprisingly higher HRV, in the largest cohort (N=717 participants), was correlated with higher stress. An Oura ring cohort of 525 first-year college students found a link between overnight PRV and perceived stress, but that was also seen with resting heart rate, sleep, and respiratory rate. Several very small studies have examined the relationship of HRV and athletic injuries or guiding training with mixed, and predominantly negative results. HRV biofeedback training with paced breathing had no significant effect on reducing stress or raising HRV, as demonstrated with sham controlled trials. When HRV for multiple days showed a decline in conjunction with body temperature, the Oura ring published data for prediction of Covid. The WHOOP company sponsored an observational study, published in 2026, of 30,000 users for 72 weeks, without a control group, that reported reduced alcohol intake (5.8 % points) by self-report. That doesn’t tell us much, and particularly about the merits of HRV for behavioral change. If you use the same device and conditions as longitudinal trends for multiple (at least 2-3) weeks that may be the one way to get something useful from the measurements. Data for overnight sleep with minimal motion and using RMSSD is the best proxy for real HRV. The reason to look at trends rather than any given night is that it more likely represents something, even though you won’t know with certainty what the “it” is. Keep in mind there are no data, no peer-reviewed evidence, to show that HRV fluctuation in-person has any correlation with health outcomes. Share Readiness Scores These are proprietary scores that integrate different metrics for each of the wearables: no algorithms have been disclosed. They are unvalidated against health outcomes. In a review of 14 composite health scores of readiness and recovery, HRV contributed 86% to the scores, followed by reading heart rate (79%), physical activity and sleep duration (both at 71%). That review noted the substantial variability in measurement protools and lack of standardization. Sleep staging is notoriously inconsistent and inaccurate by these sensors, which adds further to the HRV uncertainties for what the scores, which use sleep stage data, mean. Only resting heart rate has been shown consistently across devices to be extremely accurate. I’ve made a Table to summarize what we know about which metrics are included, the scores, any peer-reviewed studies that compared the readiness score with health outcomes, and the corresponding (if any, NA-not available) citation. You will note that some companies do not use the term “readiness," such as WHOOP for recovery, and Garmin, which has 2 different scores, one of which is Body Battery. Eight Sleep uses the term “Fitness Score.” They all include HRV; Apple includes a new metric they call “Recovery HRV” which among other components uses 7-days of sleep, but it is unclear what this means or how it is differentiated from other scores (there are clearly no data for outcomes). We have no knowledge of how the different components are weighted or whether any of these scores are better than resting heart rate, HRV alone, physical activity, or any other single metric. Since none of these are standardized, they are not interchangeable, so if you get a 90 for Oura that has no relationship to a 90 on a Google Fitbit Air. Notably, the company can update its algorithm for readiness score at any point without notification to device users. Without any useful evidence of actionability for these scores or established relationship with health outcomes, it is hard to make a case for their value. At the Apple recent announcement they showed their Readiness score (0-10) on the watch (Figure below) but there are no published data on this score, not even on their website. It’s available only on their new Watch Series 12 or Ultra 4 [of course, ;-)]. That exemplifies the problems with these scores, lack of data and evidence for being meaningful to promote health. Perhaps the best study (which isn’t saying much) is the WHOOP Recovery for golfer performance, because it did correlate with an objective outcome, even though there was no control group and the authors were all from the company. Among the 389 pro golfers, an absolute 10-per cent point increase in Recovery score was associated with about 0.5 fewer strokes per round. But that’s hardly a health outcome! WHOOP is also conducting a study in over 2,700 runners to see if their recovery score will be linked to less injuries and improved performance, but that is not yet published and has no control group or randomization. Putting This in Context For two decades I’ve been enthusiastic about the potential for digital health and particularly wearable biosensors. Over the years, we’ve seen some great progress for their ability to promote physical activity and accurately detect atrial fibrillation (the first FDA cleared deep learning AI for consumers). That work was the subject of rigorous research. But there are holes in the data and evidence for other metrics. One notable one is the “VO2 max” story that I wrote about earlier this year. At that time many subscribers asked me to cover heart rate variability, which I finally got to here. When I dived into the research and publication for HRV and readiness scores, I expected to find at least some that were of high quality and demonstrated their utility by linkage to health outcomes. To my surprise, I found none. The wearable sensor measurements for HRV (PRV) are, for the most part, accurate, but that validation work has only been done in small studies of healthy adults and does not take into account the long list of factors, from the device, software side, and the user side, that affect HRV measurements. Moreover, this metric chiefly relies on optical sensing and, as we have learned for heart rate PPG sensing, may be less accurate in people of color. Keep in mind that all of the health outcome association evidence comes from ECG-derived HRV; none are from wearable sensor data. I will repeat the key point: there's no peer-reviewed evidence to show that in-person HRV fluctuation—or efforts to raise your HRV— has any correlation with health outcomes. For those of you who look at your HRV on awakening, or even 2+ week trends of it being low, I hope this context helps to relieve any anxiety. Yes, low HRV (not PRV) has been shown to increase risk of some diseases as summarized above. But efforts to raise your HRV—a surrogate metric— has not been established for improving any health outcomes. HRV does not have any evidence of causality (the genetic evidence actually goes against this possibility). In this summary, I have not included data for other wearable sensors such as Polar, Samsung, Withings, Suunto, Amazfit, Coros, Ultrahuman, or additional mattress sensors. These are beyond my first-hand experience, and as far as I know from my in-depth review none have any peer-reviewed published data that differ from the 6 sensors I’ve reviewed here. From the points I’ve gone over above, we’re not ready for readiness scores. Besides being proprietary, they are predominantly based on metrics that have their own issues. It’s compounding the problem, like building a house without a solid foundation that has never undergone a rigorous inspection, and then selling it. Like I mentioned for PRV, you can look at trends over weeks rather than any single day, to get a handle, but even that may not be helpful. There’s simply no evidence that these scores meaningfully relate to health outcomes. I’d emphasize they might, but that requires doing prospective or randomized studies to prove it. None exist. There’s great promise for HRV/PRV utility_. For example**,**_ Prof Maiken Nedergaard, who discovered the brain glymphatics that are essential in eliminating metabolic waste products from the brain during sleep, has posited that HRV could be a non-invasive marker for neuromodulator oscillations, brain-body regulatory circuits, and brain clearance. That would be extremely useful, but like everything else on HRV and readiness scores it requires solid research and validation. The lay media isn’t helping much to get the story straight. Earlier this year The Economist published a piece entitled “The most useful indicator of your overall health” which ordained HRV as an “accumulated stress score.” That’s akin to the false assertion about VO2max: “V02 max is the singular most powerful marker for longevity.” As I’ve summarized here, that is not established. The fact is that so many things can lower HRV, including physical exercise (especially an intense workout), reduced sleep quality, stress, the list above, no less the device, signal, and software. Whatever fluctuations observed have not been correlated with any health outcome. Sadly, “datamaxxers” are widely using HRV and readiness scores that have never been validated to mean anything. We already know that for some people using the sensors for sleep metrics, it can induce “orthosomnia,” an obsession to get high sleep quality, with associated high levels of anxiety. In an experiment done by a company to promote sleep quality for its employees, “For those employees who did use the trackers, many reported feeling perfectly rested until their tracker told them they had had a terrible night. Others were told that they had slept like a baby when they had actually been lying awake worrying about the quality of their sleep. “ The same problem can result from preoccupation with HRV or readiness scores, with anxiety that would lead to further reduction in both. It you are using a wearable like >100 million American adults, it’s OK to look at these data, but contextualized with the major caveats reviewed here. If you are one to require evidence that HRV or readiness scores are linked to health outcomes, you may not even want to look. The companies make it hard to turn them off! Let me end with the companies that make and sell wearables. Apple’s doubling down on HRV (24-fold more reporting and heart rate very 5 seconds) and introduction of a Readiness score tells us that consumers have bought into these metrics and they are joining the club. However, all of this is occurring with a backdrop of tens millions of users, claims about the data that are not backed up by adequate evidence, marketing way out in front of whatever limited data exists, and not being transparent about their readiness score algorithms. The companies can well afford to do the research that is needed to connect these metrics with health outcomes show, once and for all, that increasing HRV or using readiness scores promotes our health. If they believed and invested in the products they are selling, we’d not be in this position of not knowing. That’s essentially where we are with HRV and readiness scores. Perhaps someday this will change and we’ll have good reason to embrace them. NB: I wrote this post. No AI. I have no conflicts of interest with any of its content. Loading... Ground Truths has 215,000 subscribers from every US state and 214 countries. There are over 300,000 followers of Ground Truths so more than 90,000 folks who can easily convert to be free subscribers. Your subscription to these free essays and podcasts makes my work in putting them together worthwhile. If you’re not a subscriber, please join! If you found this interesting PLEASE share it! Share Ground Truths The proceeds from all voluntary paid subscriptions go to support our summer internship program. It enabled us to accept and support a record number of 62 summer interns that joined us in 2026! These are high school, college and medical students selected from thousands of applicants. We couldn’t do this expanded program without the funds coming in through Ground Truths. Thank you!
Science Vs on the 49ers Substation: Anatomy of a Hatchet Job
In August, Pablo Torre Finds Out and Science Vs released an episode about my theory that electromagnetic fields from the electrical substation beside the 49ers facility may be contributing to the team’s run of soft-tissue injuries. As the originator of the theory, I was also a subject of the episode. I sat was interviewed by Wendy Zukerman and Rose Rimler of Sciene Vs. I sent them research, connected them with scientists, and spent hours digging through the City of Santa Clara’s and Silicon Valley Power’s records to help clarify critical details. Both Wendy and Rose came off as affable and professional, showing curiosity during the interviews, with Wendy going out of her way to question the “conspiracy theory” framing itself. Then they titled the episode We Solve the 49ers Viral Conspiracy Theory. I was familiar enough with their catalog that I never expected them to agree with my theory, but I did expect them to give it a fair hearing and to deliver their findings honestly. On both counts they came up short. They were not directly dishonest in anything they said, but rather in what they left out, in how what remained was cut, and in the picture painted by those two things in combination. They asked me to recorded the interview on my phone for better audio quality, and I held on to my copy for reference later. That turned out to be a fortuitous decision. Here is the raw audio of my first interview, unedited*. Only my side was recorded, so their questions aren’t on it, but you can generally infer them from the answers, and every claim I make below about what I said can be checked against my statements in this clip. (Note: Wendy had connection trouble at the beginning of the interview and was cutting in and out, which accounts for the false starts.) 0:00 -1:19:13 Audio playback is not supported on your browser. Please upgrade. Science Vs is a fun show. It is pitched at roughly a high-school understanding of science, for an audience that is skeptical, Snopes-oriented, and worried about misinformation. Take a claim circulating online, interview a supporter of the claim, interview consensus scientists, and then tell people which side of the line it lands on. I have listened to their episodes on raw milk, seed oils, 5G and other contested issues in the wellness space. I have problems with all of them, but each at least made an honest attempt to represent the actual range of scientific opinion by talking to credentialed scientists with differing perspectives. Dissenting researchers were heard from, answered, and often given a chance to respond. This episode did not do that. This article is not a relitigation of the science. I have written extensively about that already. This post is for the record. It is simply an accounting of what was wrong with their episode. Contents Is there a power plant next to the 49ers practice facility? — the clearest single example of getting something objectively wrong For the record — a high-level overview of all of the problems with their investigation Draw your own conclusions — why they had to make this episode at all What they left out about childhood leukemia — the exposure threshold, and playing games with statistics to minimize sick children What they left out about the mechanism — the body of literature they refused to cover, and evidence for the mechanism used against it The scientists Science Vs interviewed and did not air — four credential scientists who support myhe theory, recorded and cut, or not even called How they misrepresented me and my case — an answer severed from its question, and three telling edits How you would actually study the injuries — it doesn’t actually matter which team is the most injured, overall The theory: magnetic fields, oxidative stress, and collagen Since the episode never states my theory, here it is below. Those who are familiar with my work can skip this section. Electrical equipment produces magnetic fields, and those fields induce faint electrical currents in anything that passes through them, including people. My argument goes further than induced current. Mitochondria — the structures that generate a cell’s energy — run on continuous flows of electrons and protons, and those flows carry faint electromagnetic fields of their own. Maxwell’s principle of superposition shows that when two fields share the same space, they add together. An external field does not need to be strong to alter an internal field, it only needs to be present. The electron transport chain inside the mitochondria leaks a small fraction of its electrons by design, and those become reactive oxygen species — free radicals, which healthy cells use as signals. External magnetic fields cause the rate of leakage to change, increasing free radicals. When free radicals accumulate faster than the cell’s antioxidant systems can clear them, the result is oxidative stress. Oxidative stress is a recognized driver of chronic illness, including cardiovascular disease, neurodegeneration, diabetes and cancer, and it is in every medical school curriculum. One of its ordinary effects in connective tissue is that it switches on matrix metalloproteinases — MMPs, the enzymes that break collagen apart — while suppressing new collagen synthesis. Tendons are made of collagen. An athlete loading his soft tissue heavily six days a week is already creating oxidative stress in the tendons, so the addition of magnetic field exposure compounds the damage and inhibits repair, making him more susceptible to injury when enormous force lands on a tendon in a fraction of a second. That is the theory. We know that EMF causes oxidative stress, and we know that oxidative stress degrades collagen. The only thing missing is a study showing that direct connection. I have never claimed certainty that the substation is contributing to the elevated injury rate. I do believe it is likely, and I believe it warrants caution, and that a revenue machine like a major NFL team can easily afford to remove the exposure out of precaution. Is there a power plant next to the 49ers practice facility? Here is the clearest single instance of the problem with this episode. It requires no knowledge of electromagnetic fields at all. This story first came to me through Jon Feliciano, a former 49er, in an interview on Willard & Dibs on 95.7 The Game. Asked about why the 49ers are always injured, he said players used to joke that it was the power plant behind the practice fields. Wendy Zukerman corrects him on air. “It’s important to say that it’s not a power plant near the 49ers site.” Torre then asks her to explain the difference between a substation and a power plant. Her answer: a power plant generates electricity and transmits it to a substation, which converts the voltage into something your house can use and sends it on to your home or your business. There are two problems with that. The first is that there is a power plant on that site. The Gianera Generating Station is a natural-gas-fired peaker plant owned by Silicon Valley Power, and it has been there since 1986. I have published aerial photographs of the location from 1987 in which the gas plant is the only thing standing, before the practice facility was even there. I even mentioned the gas plant in my first interview. (Feliciano has since gone on record about this issue, and it is clear he had the substation in mind rather than the gas plant.) The second is that her definition describes a distribution substation, and the facility beside the 49ers is not that. It is the Northern Receiving Station, which takes in bulk power from outside the city’s grid and delivers it to distribution substations. Silicon Valley Power’s own map distinguishes receiving stations from substations, and Rose Rimler even reads the label out loud in the episode when she walks through that map on air. A few minutes before the map segment, Zukerman tells Torre: “like I said, we fact check everything.” Well, maybe not everything. I made a mistake On the morning of July 6th, at 10:20, I sent Rose and Wendy two studies. During the interview I had been asked which human studies show oxidative stress from electromagnetic fields, and I had not been able to name one on the spot. Rose emailed asking for the studies, I quickly searched for a couple, and I wrote that I thought these were the most relevant from that search. Unfortunately, I was busy and did not read either one closely at first. I picked them because they were topically relevant — one on substation workers, one on power plant workers. Neither were great studies, and their problems go past the ones raised in the episode. That was careless on my part. There is context that did not make the episode, though. Before that email I had been on a Zoom call with Rose talking about EMF and oxidative stress, and I pointed her to Dr. Henry Lai at the University of Washington, who has spent years aggregating the literature on this question. His tallies show that roughly 90% of studies since the late 1990s find oxidative effects from electromagnetic field exposure — hundreds of studies, systematically counted. Rose told me she was already aware of his work. So when she asked for studies, I understood her to have the aggregate view already and that she would report anything within that context. Not reading those two carefully was still a mistake, but the larger mistake was assuming the episode would be based on a thorough review of the actual body of literature, rather than on whatever I happened to send that morning. The second email At 1:49 that same afternoon — three hours and twenty-nine minutes after the first — I sent them a second email that I had been preparing for a while. It was a structured document in six numbered sections, written in response to a request they had made two weeks earlier, laying out my entire argument with citations attached to every link in the chain. The episode aired on August 13th, five weeks later, and not a single item from the second email is in it. Here is how it opens: The full email is included at the bottom of this post in Appendix B. Nearly 1,500 words in total. It contains two named mechanisms with the physics behind them. Occupational studies in humans. The literature on oxidative stress and collagen. A section addressing the strongest counter-evidence against my own theory, which I raised myself. And a closing section stating plainly what the theory cannot yet show. My entire argument, in writing, with detailed citations for every claim, provided to them five weeks before broadcast, and none of it made it into the episode. That is the single largest omission, but it is not an isolated one. The whole episode is riddled with similar problems, and the conclusion they came to is not merely a difference of opinion about the scientific evidence. Through omission and through strategic editing, Science Vs misinformed its audience — about what the evidence shows, about what I argued, and about who they spoke to. That is a notable outcome for a show whose stated purpose is correcting misinformation. For the record I am not going to tell you that Science Vs set out to discredit me, or that they had come to their conclusion before they reported the story, or that they used misleading editing to make me look like a man retreating from his own theory. I cannot know what was in anyone’s head, and I am not going to pretend otherwise, either. What I can tell you is what is on the record. They asked me for evidence and aired the weakest thing I sent. They interviewed three scientists who supported my theory and aired none of them. They discussed childhood leukemia without ever saying the exposure level at which it appears. And they took an answer I gave weeks before Ken Karipidis, the one EMF scientist who was put on air, was interviewed, and placed it directly after his critique, where it comes off as a major concession. A friend of mine who has worked as a journalist for thirty years read that last one and called it sleazy. I agree with her. Those are just the four issues that are impossible to defend as pure editorial decision, but they are not the only issues. The most glaring issues stem from what they left out: The exposure level at which the childhood leukemia association actually appears — three to four milligauss, a number the episode never states despite it being much lower than what I measured. The mechanism itself: The episode gets as far as “oxidative stress could weaken collagen,” which leaves it sounding like speculation, rather than a known pathway. The second email, and all the studies in it. Dr. Paul Héroux at McGill, whose work my proposed mechanism rests on, whose contact details I supplied, and who was never called. Dr. Beatrice Golomb at UC San Diego, who runs a research group on oxidative stress and mitochondrial function and has specific expdertice in collagen, interviewed and not aired. Dr. David Carpenter at Albany, the most prominent academic voice on the harm from magnetic field exposure, interviewed and not aired. Dr. Joel Moskowitz at Berkeley, interviewed and not even credited. The specific, testable condition I put in writing that would make me abandon the theory. My own unprompted statement of the theory’s evidentiary gap. My caveats about my own measurements. My ranking of EMF as one factor among several rather than the only one. The new equipment installed at those facilities between 2010 and 2014, and the sharp rise in system load from 2013 onward. But that is not all. What they kept, they selectively edited to serve a narrative that was not served by the raw transcripts of the interviews. A description of thousands of commercial routers in a workplace, reduced to the Wi-Fi in your house. My own acknowledgment, multiple times, that "I'm not a scientist" clipped in an episode that spends considerable energy pointing out that I am not a scientist. And a qualification I volunteered about my own measurements, cut from my answer and delivered later by their EMF expert as a correction. Together their edit produce a portrait of a wellness guy with Wi-Fi phobia and two bad studies, backing away from his theory the moment a “real” scientist looked at it. A practice facility where nothing changed because no new substation was built. Injuries written off as random because the 49ers were not quite the single most injured team in the league over the period in question. That portrait is false even though almost no individual statement made in the episode is. Draw your own conclusions At the top of the episode, Pablo Torre says his show had been flooded with requests to investigate the substation — that it was the thing they were asked to look into over and over again. Neither show went hunting for this story. It arrived from their own audience, in volume, and it could not be waved off. A thread I posted on X reached twenty-two million people. A guy with a gaussmeter and a brand-new Substack obligated a Pultizer prize winning program to spend an hour on a question about player health, that is considered fringe enough that most media operations won’t touch it. Pablo Torre Finds Out could not ignore it. So Science Vs came in and ran a “debunking” operation instead of an investigation. Everything below is what had to be left on the cutting room floor to deliver the “debunking”. They could not ignore it. So they tried debunked it instead. The demand was strong enough to force an episode, and since the episodes release the 49ers have sustained more injuries, including season ending ones, and interest has only grown. Demand from fans and playets is strong enough to force a better investigation from someone else. An omission is not a lie until you find out what was omitted. What follows is the detailed record so you can see the full picture. Draw your own conclusions. I have drawn mine. I am going to keep publishing. A note about Pablo Torre Finds Out After the episode aired I posted publicly, calling out Pablo Torre Finds Out for concerns I had about the integrity of their investigation, and about an interview directly with Pablo Torre which had been offered but never happened. I have since spoken with a producer at PTFO. The offer had rested on a misunderstanding about where things stood between us, and it fell away as the Science Vs part of the production widened past the science. He apologized for the miscommunication and I accepted. The producer also confirmed to me that they did essentially hand off the entire investigation to Science Vs for narrative purposes — so Pablo would be genuinely curious or potentially surprised by their findings — while still taking responsibility for what went out under their name. I appreciated the candor and the accountability, and I am comfortable with where we left things. So this post is not about PTFO. My disagreement is with Science Vs. What they left out about childhood leukemia Childhood leukemia appears at three milligauss. They gave no number at all. 1. They built an EMF ruler, then discussed childhood cancer without it. The sequence on air went like this: First I measured almost 9 mG, then I measured up to 50 milligauss. Torre reacts with concern — “that sounds high.” Cut to Karipidis: the public limit is 2,000 milligauss. Torre reacts again — “What happens at 2,000?” Karipidis explains that nothing happens at 2,000, that safety factors are built in. Then: at more than 50,000 milligauss you get flickers of light in the retina, and higher still, nerve stimulation. The listener now has a ruler (their ruler) to measure these numbers with. Fifty is nothing by their ruler. Then the episode turns to childhood leukemia and states, accurately, that there is a real association — but tellingly, they never give the exposure level at which that association appears. Childhood leukemia appears at three milligauss. They conveniently left that out. The pooled analyses that produced the World Health Organization’s carcinogen classification found roughly a doubling of childhood leukemia risk above three to four milligauss. That is the number that belongs in that segment, and it is more than ten times below the reading that had just been dismissed as trivial. I gave them that number myself, on tape, in the interview. Describing my own threshold, I said my comfort level for long-term chronic exposure is three to four milligauss, but preferably one. 2. Six children in thirteen million. The episode reports that these fields might explain one to four percent of childhood leukemias, then reframes it as a German estimate of perhaps six cases among thirteen million children. Torre, to his credit, is uncomfortable. What never gets explained is why that number is small. Almost no children lives above three milligauss in Europe. Fewer than two children in a hundred, in the Italian data. So even if the risk to those children doubles, the number of additional cases across an entire country stays small — not because the hazard is small, but because hardly anyone is exposed to it. Count the whole population and the few who are exposed disappear. And the reason hardly anyone is exposed is that siting rules keep housing away from transmission lines. In Europe, at least. The Dutch government has a standing precautionary policy against building new homes in any zone where the annual average magnetic field exceeds 4 milligauss — the same number the episode never mentioned. Denmark and Norway treat 4 milligauss as the level that triggers a look at mitigation. That is five hundred times below the ICNIRP figure quoted on air. The United States has no federal equivalent. That is evidence of infrastructure siting policy working. It is not evidence that there is no hazard. It is evidence that we already act as though there is one. If you listen to the raw audio you will hear that my proposed solution is EMF exposure is to build better infrastructure. What they left out about the mechanism 3. They accused me of cherry-picking, and then cherry-picked themselves. On air, Zukerman tells Torre that people who worry about electromagnetic fields dismiss the studies showing null results and collect the ones showing harm, and that this is why you have to be careful not to cherry-pick. They had two papers I had rattled off in a hurry, and a second email containing a carefully chosen set. They ran the two weakest, graded them on air, and presented the result as the state of the evidence for not only my theory, but the whole question of oxidative stress from magnetic fields. That is cherry-picking from my emails. The bigger picture is what they did with the literature as a whole. Zukerman acknowledges it exists, dismisses it, and disposes of it in a single stroke. You can go online right now, she tells Torre, and find scores of studies showing these fields increase oxidative stress. The issue, she says, is that they are not studies in people. Many of them are in people. The occupational studies in my second email are in people. That email had been in her inbox for five weeks. She waved off the literature on a factual claim that the documents in her possession contradict. (I have written separately about the evidentiary standard at work here). The effect was that they essentially baited me: ask the subject to supply the evidence, take the weakest thing he sends, and let that stand as a proxy for the whole argument. Wendy has been doing this show for ten years and clearly knows how to extract what she needs for her desired narrative. 4. They presented the collagen-improvement research as a gotcha, but I had already addressed it. The episode treats the existence of studies showing that EMF can improve collagen as a surprise that undercuts me, but I had pre-emptively addressed that five weeks earlier in the email they ignored. Those studies use measured, high-intensity pulsed dosing. The mechanism by which they work is oxidative signalling. The cell reads a controlled rise in reactive oxygen species as a warning, upregulates its antioxidant defenses, mounts a repair response, and the therapy stops before the system is overwhelmed. The key point is that it induces an oxidative response, which means those studies are actually evidence for my proposed mechanism, not against it. A field that produces a therapeutic effect cannot be biologically inert (which is the implication when they say “Nothing happens at 2,000 milligauss). Chronic ambient exposure is the same pathway with no off switch, and it is most consequential during the evening and while sleeping, when the body is in repair mode and the exposure does not stop. Section 5 of my email is titled What about the study showing improved collagen synthesis. I raised the counter-evidence myself, explained why I did not think it refuted the theory, and wrote that I was disappointed the question had not come up in the interview. That is worth holding against the cherry-picking charge in the item above: the studies they presented as inconvenient for me were studies I had gone out of my way to put in front of them. 5. Section 1 of my email stated a critical argument that was never put to the expert directly responsible for it. My email says the exposure limits exist to prevent acute effects and do not address chronic exposure at all, and that my dispute is with the entire framework. Zukerman gestures at this once. After Karipidis gives the 2,000 milligauss figure, she tells Torre that I would object because those levels do not capture “more subtle effects.” That is not my argument, and “subtle” is doing a lot of work here. My argument is that the limits are built to prevent acute effects — nerves firing, muscles twitching, etc — and were not designed to address chronic exposure at all. In fact, ICNIRP’s own documents state as much themselves. Mine is a structural objection to what the standard measures, reduced as a concern about subtlety, it becomes a matter of degree, rather than a categorical difference. They interviewed Ken Karipidis, vice chair of ICNIRP — the body that writes those limits. He was asked what the limit is, but he was never asked how it applies to chronic exposure. The scientists Science Vs interviewed and did not air 6. They interviewed four officials of the standards body. The episode credits roughly a dozen researchers. Four of them are associated with ICNIRP. Ken Karipidis is the vice chair. Rodney Croft chaired ICNIRP from 2020 to 2024 and led the development of its current guidelines. Dan Baaken is ICNIRP’s scientific secretary. Frank de Vocht sits on the Main Commission. One of them speaks in the episode. Their institution’s position is more than adequately represented — the 2,000 milligauss figure, the safety factors, the assurance that nothing happens below the line. 7. The four who were left out. Dr. Beatrice Golomb is Professor of Medicine at UC San Diego — a physics undergraduate degree, a PhD in biology, an MD. She directs a research group whose stated focus is the relationship between oxidative stress, mitochondrial function and health. That is the mechanism the episode spent its first half dismissing. Her published work includes the fluoroquinolone case series behind the FDA’s black box warning on an antibiotic class that ruptures tendons, where the accepted mechanism is oxidative stress and mitochondrial toxicity — the back half of my chain, already on a drug label. She is in the credits. She is not in the episode. She has told me she finds my theory biologically plausible, and that she has a recording of her interview. Dr. David Carpenter directed the Institute for Health and the Environment at the University at Albany for decades and is the most prominent academic voice arguing these fields cause harm. Childhood leukemia is the reason these fields carry a WHO classification at all, and Carpenter is the person you call if you want that case made by someone who has spent a career on it. He is in the credits. He is not in the episode. He has told me that when he was interviewed they could not have been less interested in what he had to say. Dr. Joel Moskowitz directs the Center for Family and Community Health at UC Berkeley’s School of Public Health. He told me directly that they interviewed him and that he thought it went well. He does not appear in the episode, and unlike Golomb and Carpenter, he does not even appear in the credits. Dr. Paul Héroux at McGill is the researcher whose work provides the foundation for the oxidative stress mechanism my theory rests on. I gave Rose his email address and phone number. He was never interviewed. They never even called him. 8. “He’s not a scientist.” The episode leans hard on the fact that I am not a credentialed scientist. Zukerman says it. Torre repeats it. “At all.” Having built a segment on that, they then left out the physicist-biologist-physician, the professor of public health, the director of a university environmental health institute, and the McGill researcher whose work the primary mechanism comes from. They spent the episode telling their audience I am not a scientist, and then kept four actual scientists who support the theory off because their argument does not survive otherwise. How they misrepresented me and my case 9. They played my words as a response to an interview I had never heard. Near the end of the hour, Zukerman says she went to me and told me the evidence was lacking. Then she plays me saying I am not wedded to it being an EMF, that I have staked my reputation on it needing more study, and that I am not certain of anything. That segment was placed immediately after Karipidis had finished taking the cherry-picked papers apart. It plays as a man folding under expert scrutiny. Commenters have used it against me since, but it is incorrect. Here is the sequence as it happened. I was interviewed. Karipidis was interviewed weeks later. I was never told what he said, or even who he was. I was never asked to respond to it. I did not know his name until the episode aired. That is not a simple editing decision. That is a quote from one interview, moved to follow a critique from a different interview it was not a response to, so that it would land as a concession. A friend of mine who has spent thirty years as a journalist read that sequence and called it sleazy. She is right. The full passage is reproduced in Appendix A, and it is worth reading. What airs is the preamble. Four paragraphs later I say “to directly answer your question” — and that answer, about what would change my mind if the team relocated, is not in the episode. Neither is the sentence two paragraphs on, where I say the EMF might turn out to be ten percent of the problem. I will say the underlying point again, though, because I meant it and I still do. I am not wedded to my theory that magnetic fields from the substation are contributing to the 49ers injuries. I could be wrong. What I am wedded to is the process of scientific inquiry, and the fact that we should be cautious around these exposures and we need more independently funded studies. 10. Three telling edits. They cut “I’m not a scientist” mid-sentence. What I said: “...and I am not certain of anything. I’m not a scientist. To me, I find the mechanisms compelling. I will go back to the fact that I got into this not because of the 49ers, but because of clients I was working with who suffer from EDS...” What aired: “...and that I am not certain of anything. I’m not.” The clip ends on the first two words of the sentence in which I say I am not a scientist, in an episode that spends considerable energy establishing that I am not a scientist. The full sequence is in Appendix A. They cut my own description of my work. Asked what I do, I said I run a wellness business, that I am focused on writing software to help people navigate environmental stressors, that environmental health is my focus, and that I also do independent journalism and research. What aired is the wellness business and the environmental health, with the software and the journalism removed from my own answer. The software engineering does come up later — narrated by them, as a hacker who grew up to write firmware — but my current work, in my own words, arrives as “wellness guy.” Thousands of commercial routers became home Wi-Fi. What I described: working at a wireless networking company surrounded by thousands of Wi-Fi routers, and a stress-induced health collapse afterward that working in that environment contributed to. What aired: Wi-Fi, the kind you have at home — reinforced by a clip of Christian McCaffrey talking about turning his router off at night. Three cuts, and every one of them makes me seem less credible. That is not a coincidence. 11. I gave them my own caveats and the test that would disprove me. In the first email I wrote that a well-designed chronic, ambient-level exposure study in tendon showing no collagen degradation, run over many months, would undercut my confidence in the theory. That is a specific, testable condition, stated in advance and in writing. There was a second interview, weeks later, which was not used in the episode, but in that I asked again for the falsification criterion to be included. What the episode presents instead is a man saying “we need more study,” which reads as a hedge, an open-ended request for the benefit of the doubt. It is the opposite of a hedge. Section 6 of my second email says, unprompted, that as far as I am aware there is not a single study showing EMF-driven oxidative stress degrading collagen in a human in vivo or in human tendon in a lab. I wrote that because it is true and because it is the first thing anyone should check. Under their standard it becomes the hinge of the entire first half, framed as something they had to surface without my cooperation. I had handed it to them five weeks earlier. And in the interview I said that my 25 to 50 milligauss readings were taken at knee level, and that these fields attenuate quickly, so by the time it reaches your head it is not that high anymore. That did not air either. The qualification arrives later, from Karipidis, as a correction. How you would actually study the injuries 12. They measured the wrong thing. The episode spends its final act on whether the 49ers are the most injured team in the NFL, finds no clean ranking, and lands on “injuries are random”. That is the wrong test. You would define the exposure — measured field levels at the facility — and the outcome narrowly: soft-tissue injuries, not a single MSK injury count that a broken finger and a ruptured Achilles both land in. Then you would compare the 49ers to themselves before and after the exposure changed, and compare that change against every other team over the same window. That controls for league-wide shifts in rules, turf and training load. It is a difference-in-differences design, and it is standard. I sent them that argument on July 28th, in writing, along with the fact that in eleven of the past twelve seasons the 49ers had finished top ten in Adjusted Games Lost. And by the episode’s own reporting, the 49ers are second in games lost to injury over the past decade, with FTN’s Aaron Schatz confirming that once you adjust for starters they come out second worst in the league — after having been one of the healthiest teams for years before that. A finding like that is where an investigation starts, not where it stops. 13. I did a lot of substation research for the episode that was never aired. I had reported that the substation next to the facility was expanded in 2014. That was wrong, and Rose Rimler and I sorted it out together. They gave me credit on air, which I appreciated, but framed it as a concession on my part to something they had found. In fact I had been investigating it myself, was uncertain before they asked, but was waiting for clarification from Silicon Valley Power. The uncertainty centered around a substation named “Esparanca” that was proposed right behind the facilities as part of the stadium construction, but now appears to never have been built. I spent days in Silicon Valley Power’s records, and did eventually reach Silicon Valley Power for confirmation and indentified the mistake myself. I also gave them highly detailed documents on what we do know about those facilities: new equipment installed between 2010 and 2014, and load growth across the system with an inflection around 2013–2014, with the Northern Receiving Station carrying the highest capacity in the system and sitting closest to the data center corridor. None of that is in the episode. I have a separate piece about it coming out soon. *I cut 2 seconds from the raw audio to remove personal information that had no connection to anything in the episode. Appendix A: the passage the clip was taken from This is the sequence from my first interview, in full. The episode uses one fragment of it. It’s tricky because my understanding is that there’s something like a 15-year half-life on collagen remodeling issues. They’re going to get a new crop of players. What I really need, though, is for the injuries to be broken down so we can see the soft tissue injuries. Right now, they’re categorized as MSK injuries, but let’s say we can actually identify them more specifically. Is that right? He was able to do that? Oh, cool. I wasn’t able to find that information. Interesting — I didn’t know that data existed. I tried so hard to get it, and I didn’t know it was available. Maybe you could connect me with that person. That would be amazing. I’m not wedded to it being an EMF, honestly. I’m not. I would say that, to a certain extent, I’ve staked my reputation on the fact that it needs more study. It’s important to look into it, and I am not certain of anything. I’m not a scientist. To me, I find the mechanisms compelling. I will go back to the fact that I got into this not because of the 49ers, but because of clients I was working with who suffer from EDS, Ehlers-Danlos Syndrome, a tendon hypermobility issue. This condition is documented to be partially caused by oxidative stress. I’ve worked with them to mitigate their EMF exposure and have seen people improve. I think it’s a bigger story. In a way, the 49ers represent an occupational cluster with specific exposure risks. If someone would conduct a study on this, I believe we could learn a lot. Perhaps we would discover that it’s like, okay, maybe it’s 10% of the problem. I don’t know the exact number, but I think either way, conducting the study would provide valuable insights. To directly answer your question, it’s challenging for me because it’s part of a bigger story. If the 49ers move and stop having injuries, I might conclude that it was related to the substation. However, if they move and still experience injuries, I would want to analyze the nature of those injuries. Are they primarily soft tissue injuries? I don’t think it’s necessarily a case closed, but the story might lose traction at that point regardless. Three things are worth noting. The aired fragment stops immediately before the sentence in which I say I am not a scientist — the cut described in item 10. It also stops four paragraphs before the words “to directly answer your question.” What the episode plays is the throat-clearing. The actual answer, about what would change my mind, never airs. And two paragraphs in, I say the EMF might turn out to be ten percent of the problem. That is the ranking the episode says I never gave. Appendix B: the full email, July 6th, 1:39 PM Hi Rose, You asked me to send you some studies a couple of weeks ago. I apologize for the delay. I have been incredibly busy. I also wanted to give you a chance to dig in yourself, and get your own perspective before I influenced you too much with mine. That said, I do want to make sure that the actual argument is abundantly clear, so here it is! One note before I begin, there are really two separate questions here: Does the substation cause the soft-tissue injuries? That was my specific claim, and what the podcast is about. Is the theory biologically sound? Can 50/60 Hz fields at these levels do anything biological below the safety limits? Question two is why people have been calling me a conspiracy theorist. The regulatory and advisory bodies (ICNIRP, etc) maintain that anything below their exposure limits is biologically inert, and that position is broadly accepted within the mainstream, but the actual science paints a different picture. To defend my theory I’ve had to defend a much larger claim: that the advisory bodies have the biology wrong. I can’t honestly answer question 1 without first addressing that. Here’s is a simplified outline of my theory, with links to studies: 1. Advisory body thresholds for biological effects The ELF magnetic-field limits (ICNIRP’s ~2,000 mG public / 10,000 mG occupational) exist to prevent acute effects from exposure. If a magnetic field is strong enough, it makes nerves fire and muscles contract. Muscles twitch. You can see faint flashes of light in your vision. People can even feel tingling or even a shock. That happens due to Faraday’s law, which states that a time-varying magnetic field induces an electric field inside the body via the creation of Eddy currents.The standards assume that those currents do nothing below the twitch threshold. They do not consider chronic effects at all. My argument is that they are biologically active and do have chronic effects. 2. There are multipe ELF-EMF mechanisms for inducing oxidative stress. This is basic biophysics. Below are the two primary mechanisms. Simplest first. Dr. Paul Héroux of McGill has the most intuitive explanation, which rests on Maxwell’s principle of superposition: because Maxwell’s equations are linear, two sets of electromagnetic waves sharing the same space simply add together. In the interview I pictured a pebble dropped into a pond, sending out clean ripples. Then a boat motors past at distance, sending larger waves, and where they meet the ripples, the two add together and the original pattern gets distorted. In our cells, the mitochondria run on continuous currents of electrons and protons that carry their own faint electromagnetic fields. An external field from a substation, by the principle of superposition, adds to those internal currents and shifts the pattern. The electron transport chain in the mitochondria leaks a fraction of its electrons by design as reactive oxygen species (ROS), which act as signalling molecules in a healthy cell. Distorting that field increases that electron leakage, creating oxidative stress. A tiny electromagnetic interference results in a large downstream effect. https://www.sciencedirect.com/science/article/pii/S2405844025006474 The more complex pathway is Dimitris Panagopoulos’s Ion Forced-Oscillation / Voltage-Gated Ion Channel (IFO-VGIC) mechanism. Weak, polarized, low-frequency fields force the ions (calcium, potassium, sodium) next to an ion channel’s voltage sensor to oscillate; because that force scales with the inverse cube of the distance involved, fields far below the stimulation threshold can still force the channel open, disrupting the cell’s ion balance and driving ROS overproduction (Nitric oxide and superoxide interact and produce peroxynitrite, a highly unstable cytotoxin that breaks down into free radicals). https://www.researchgate.net/publication/366088934 https://www.spandidos-publications.com/ijo/59/5/92 https://www.mdpi.com/1422-0067/22/18/10041 https://www.frontiersin.org/journals/public-health/articles/10.3389/fpubh.2025.1585441/full https://pmc.ncbi.nlm.nih.gov/articles/PMC3772193/ 3. Evidence of oxidative stress, in vivo, in humans from ELF-EMF There are many studies, and the strongest are occupational, which applies to the 49ers situation. 115 power-plant workers vs. 145 office controls. Serum malondialdehyde (MDA, the standard lipid-peroxidation marker), superoxide dismutase, and catalase were all significantly higher in the exposed group, and the markers rose with increasing field exposure. https://journals.sagepub.com/doi/abs/10.3233/WOR-203244 A follow-up where they reversed the oxidative stress in power plant workers. A double-blind randomized controlled trial: 91 power-plant workers given vitamin E, vitamin C, both, or placebo for 90 days. The antioxidant groups showed reduced lipid peroxidation and raised antioxidant capacity versus controls. https://pubmed.ncbi.nlm.nih.gov/32191586/ A spot-welder study: 46 welders occupationally exposed at 50 Hz, showed significantly altered red-blood-cell antioxidant enzyme activity and concluded ELF-MF “might act as an oxidative stressor… even at the recommended levels of exposure.” https://pubmed.ncbi.nlm.nih.gov/18504600/ A substation-specific study found oxidative stress and DNA damage https://pubmed.ncbi.nlm.nih.gov/24460415/ 4. Oxidative stress degrades collagen This half isn’t controversial at all. Oxidative stress activates the matrix metalloproteinases (MMPs) that break collagen down and suppresses its synthesis. Here are three example studies: Siwik et al. 2001: Oxidative stress raises MMPs and lowers collagen synthesis in cardiac fibroblasts. https://pubmed.ncbi.nlm.nih.gov/11121376/ Fisher et al. 2009: Collagen fragmentation drives oxidative stress and elevates MMP-1 in aged human skin. https://pmc.ncbi.nlm.nih.gov/articles/PMC2631323/ Varani et al. 2006: Decreased collagen production in oxidatively aged skin. https://pubmed.ncbi.nlm.nih.gov/16723701/ 5. What about the study showing improved collagen synthesis I’m glad you asked this question, and I’m bummed we didn’t get to it in the interview because it is a great question, and very important to understand. Superficially it may appear to be a refutation of my theory, but I would argue that it actually supports the underlying argument. That study used a short, high-intensity pulsed dose: therapeutic-PEMF, the same type of field used clinically to stimulate bone and tissue healing. The reason this works is because it triggers the very same biological pathways that can lead to oxidative stress. The exposure still stimulates ROS production, but done at a controlled dosage, for a limited duration it has a different effect. The cell reads the ROS as a warning signal and it upregulates its antioxidant defenses, and mounts a repair/proliferation response. The therapy is then stopped before the cell gets overwhelmed. That is essentially a hormetic response: a small dose of a stressor producing a beneficial adaptation. Two things follow: At the very least, this means the field is biologically active, which by itself undercuts the regulatory claim that these fields are inert below their thresholds. There is a wealth of literature on PEMF (pulsed EMF) bone regeneration and calcium channel signalling (which is the second mechanims in section one). Here is one. https://www.nature.com/articles/srep13856 At best, it leads us straight to my actual concern, which is that under chronic, chaotic and unregulated exposure, the cell’s antioxidant capacity gets overwhelmed, resulting in oxidative stress. There are plentiful examples of things that are beneficial at controlled doses, but harmful at uncontrolled doses. Excecise is one obvious, uncontroversial example. Botox is perhaps a more interesting example. 6. What about a study that shows EMF causing collagen degradation in humans? As far as I am aware there is not a single study showing EMF-driven oxidative stress degrading collagen in a human, in vivo, or in human tendon in a lab. There are studies involving isolated fibroblasts though, which secrete collagen. The most relevant one showed that human fetal scleral fibroblasts exposed to 50 Hz ELF-EMF showed a significant decrease in collagen synthesis and that MMP-2 expression was upregulated, which is the exact mechanism I described in section 2. https://pmc.ncbi.nlm.nih.gov/articles/PMC3626379/ That’s it! Thanks again for the interview. Best, Peter Anthony Cowan Environmental Health · Wellness Tech · Researcher & Journalist Publisher & Editor, Living Energy Bio: petercowan.comSocial: X · Instagram · TikTokApp: sunlightis.lifePhone: [redacted]
Engrams Embedding Entendre: Codesign for Efficient DRAM/SSD Offloading
Engram extends standard token embeddings with learned multi-token lookups. Recurring local patterns retrieve vectors directly, reducing the need to reconstruct them through attention and feed-forward layers. With Engram model architecture optimization, it allows for lower HBM capacity to be needed for models at the same quality. This does not mean there won’t be an insane demand for HBM but it just means that model architecture will continue to innovate around constraints. This model architecture design is naturally codesigned for parameter offloading: each token accesses a few embedding rows whose addresses depend on token IDs, not hidden states. The runtime can prefetch those rows from host DRAM while earlier layers compute, keeping the table outside HBM without transferring entire weight matrices. Our Memory model contains our latest estimates of quarter by quarter HBM, DRAM, & NAND supply and demand. Offloading frees HBM for model weights and KV cache, potentially supporting larger batches or more concurrent sessions. When DRAM becomes the next constraint, NVMe offers another tier. Recommendation systems already cache frequently or recently accessed embedding rows in faster memory while backing colder rows with SSDs. After NVIDIA roadmap had to change due to massively despec’ing Rubin Ultra from 1024GB to now ~200GB of HBM per chip, model architecture optimizations like emgram maybe helpful. Our DeepSeek-V4.1-Flash configuration uses roughly 189 GiB of memory for Engram. We replace it with a memory-mapped (mmap) file and measure serving performance with offload to SSD. Later on in our report, we will show our Engram offloading experiments along with the official InferenceX agentic inference serving results on Engram models like DeepSeekv4.1 Flash across all 6 NVIDIA GPU SKUs along with MI355X. Unsurprisingly, the CUDA Moat is still mogging MI355X on the ultra popular DeepSeekV4.1 Flash model. We also show how even on high capacity HBM SKUs, offloading emgrams to DRAM could result in even better performance for most of the pareto than keeping the emgram in HBM. Our benchmark has been widely reproduced, validated and/or supported by almost every major buyer of compute from Google Cloud to Microsoft Azure to Oracle, to Meta and many more. Furthermore, it has the support of the ML community including from vLLM, LMCache, SGLang, PyTorch, Huggingface and the support of major labs like OpenAI, MiniMax, ZAI, Qwen, Moonshot Kimi, etc. Star the InferenceX GitHub repository if you find the open-source benchmark and data useful!. InferenceX is the only inference benchmark in the world to have TPUv7, Jalapeño, Nvidia Rubin NVL72, AMD, and soon, SambaNova and Trainium. Due to how realistic AgentX scenario is to real world agentic inference workloads, AMD has committed to collaborating on MI455X UALoE72 too. DeepSeek did not release the original paper’s two trained Engram models. We replicated its setup on fineweb-edu using the released code and training hyperparameters, at an estimated 6E18 FLOPs per run. We observed the same U-shape scaling: Engram improved performance over pure MoE baselines. We also reproduced DeepSeek’s results, where earlier-layer representations with Engram resembled those of later layers. Like the original Engram paper, it is possible to probe Engram’s gate scores to see what n-grams DeepSeek-V4.1-Flash makes the most use of. Our gate scan finds names, code fragments, relational phrasing, and boilerplate. These examples prioritize interesting-ness over gate strength. One unexpected result was Wright : Ace Attorney. These examples suggest learned memory optimizes the training objective, not a judgment of which facts deserve storage. Licenses, bibliography fragments, API scaffolding, and website furniture can provide prediction shortcuts, so the value of additional Engram capacity may depend on what survives data preparation. This does not show that table capacity is “wasted”: the evaluation-corpus scan establishes neither training exposure nor the capacity occupied by each category. For offloading, strong gates do not identify cache-hot rows. Low gates do not automatically save reads either: computing the gate requires the retrieved key and undoes the performance gain of a fused kernel. Skipping reads would require a separate usefulness predictor before retrieval. In the original paper’s inference-time ablation, factual-knowledge benchmarks retained just 29–44% of their original performance, while reading comprehension retained 81–93%. This is due to the training–inference mismatch. The resulting degradation therefore measures this trained model’s dependence on Engram, not the performance difference between models trained with and without it. In our ablations, suppressing Engram worsens token likelihood across all evaluated domains, especially encyclopedia text and several code corpora. Surprisingly, GSM8K accuracy stays within measured run-to-run variation and removing Engram has no effect. Engram is not a detachable dictionary beside an unchanged MoE. Removing it changes downstream features and expert selection. We tested whether rerouting hurts or compensates by holding tokens fixed in a teacher-forced experiment on CRUXEval, a code-reasoning benchmark of small Python functions where the model predicts a function’s output from its code and an input, and scoring the reference answer. Removing Engram raised answer loss from 0.2848 to 0.3093 bits/token. Forcing the ablated model to use the original Engram-on expert choices made it worse still, at 0.3375 bits/token. Rerouting partially compensates for the missing memory. Memory features and expert selection work together, rather than following a clean “memory stores facts; experts reason” division. On the same CRUXeval, removing Engram during either phase reduced accuracy and increased generated tokens; removing it throughout produced the largest changes. Keeping Engram for prefill leads to more correct answers than keeping for decode likely due to semantically richer KV cache transferred to decode workers, allowing it to mitigate some of the performance loss. Engram’s table is large, but each lookup is small. DeepSeek-V4.1-Flash requests 24 rows at each of two Engram layers, about 12.4 KiB per processed token position across the model, or 3.1 KiB per GPU when split across four GPUs. Currently as of Day 7 since Model Release, MI355X is still 2-4x worse performance per dollar compared to B200 even when normalized by Mi355X’s lower TCO. Our full total cost of ownership breakdown comes from our AI Cloud TCO Model along with monthly market surveys of over 100+ gpu clouds & gpu cloud customers. On the Day 0 release of DeepSeekv4.1 Flash, NVIDIA vLLM works out of the box with zero issues across all 6 SKUs: H100, H200, B200, B300, GB200, GB300! This was thanks to the amazing work by the NVIDIA & Interact teams! In comparison, AMD vLLM did not work on day 0 for DeepSeekv4.1 Flash.Source: SemiAnalysis InferenceX AMD’s vLLM documentation points to using vllm/vllm-openai-rocm:deepseekv41-flash-0909, but from hour 0 of the model release to hour 23, AMD has not publicly released the image. AMD claims, “SPEED IS THE MOAT,” yet it has still not released it by the 23rd hour. We wish that, going forward, the AMD team has a better process for hour 0 model releases. Eventually when they did publicly release for “day 0” image support, performance-wise, it is currently up to 14.8x worse perf per dollar than H200 and up to 42x worse perf per dollar than B200/B300. The power of the CUDA MOAT is NVIDIA’s collaboration with its massive 6 million-developer community ecosystem including most of the vLLM & SGLang & Tokenspeed maintainers which means that CUDA is optimized on day 0. Overall, AMD did make significant improvements but the performance per dollar is still currently 2-4x worse than B200. AgentX Engram DRAM Offloading Improving Performance The HBM and DRAM offload use the same GPU kernel to select and dequantize rows. With HBM, it reads GPU memory; with Unified Virtual Addressing (UVA), it reads pinned host memory directly. Both support full decode graphs. Moving the table into HBM accelerates only the sparse lookup, leaving decoder computation and communication unchanged, resulting in little overall benefit while consuming memory otherwise available to KV cache. Another benefit of Engram offloaded to DRAM which means you can reduce the communication overhead by using less HBM GPUs per replica. For example, when enabling Engram offloading on B300, we are able to switch from TP4 to now TP2 which improves the pareto curve by up to 1.6x. When iso-model quality, less HBM is required as the engrams could be offloaded to host DRAM. Thus HBM bandwidth matters way more than HBM capacity. For inference workloads where memory bandwidth matters the most, 4-hi HBM provides the best $/bandwidth and therefore lowest cost per token. If China continues to make more and more revolutionary model architecture innovations, soon it could potentially 0Hi HBM stacks. [ ](https://newsletter.semianalysis.com/p/long-live-the-short-king-why-4-hi) [ Long Live the Short King: Why 4-hi HBM Wins ](https://newsletter.semianalysis.com/p/long-live-the-short-king-why-4-hi) Myron Xie, Bryan Shan, and 3 others · Sep 13 [ Read full story ](https://newsletter.semianalysis.com/p/long-live-the-short-king-why-4-hi) Moreover, on B300 and week-0 stack, moving the Engram table back to HBM did not improve results and stay within run-to-run variance. This is a result of the work optimizing DRAM offload, such as async, and overlap. SSD Offloading On B200, we created a unoptimized vLLM fork and stored the Engram tables in memory-mapped files on local SSDs. File backing lets the OS reclaim table pages when other applications need RAM. Pages already cached in memory can be served without reading the SSD again. Furthermore, note that we were unable to turn on GDS The unoptimized SSD implementation changes how rows reach the GPU. It copies row IDs to the CPU, deduplicates them, gathers the requested rows into pinned buffers, copies those rows back to the GPU and dequantizes them. This work runs between segments of the GPU execution graph. Native UVA performs row selection and dequantization directly on the GPU, avoiding the CPU round trip. A warm filesystem cache removes physical SSD reads, but leaves the coordination, row gathering and transfers. This is why a file already cached in RAM can still perform worse than a pinned DRAM table. The comparison measures the whole serving path; it does not separate the time spent on each of these operations. B200 DRAM dominates both measured SSD serving curves in total tokens per dollar and P90 interactivity. Near 125 tokens/s/user, DRAM delivers 121 million total tokens per dollar versus 52 million for SSD. For production serving, SSD offloading is likely not worth the tradeoff. On the B200 configurations we measured, SSD offloading loses on both measures: every observed SSD point has a DRAM alternative that delivers higher P90 interactivity and more total tokens per dollar. The only points where Cheaper storage does not automatically produce a cheaper inference service. Moving Engram to SSD leaves the same four expensive GPUs and the rest of the server in place. Reclaiming RAM only creates an economic benefit if it enables a cheaper server configuration or additional useful capacity. The current unoptimized path provides neither benefit, and the filesystem cache still consumes RAM when table pages are resident. Next we will look at the mechanisms and specific implementation of ngrams from DeepSeekv4.1 Flash, LongCat, Qwen3.8 Flash Next.
Can researchers trust that AI models won’t leak their ideas?
Research funders are asking how they can stop confidential scientific ideas in grant proposals from leaking out through large language models (LLMs), for example if reviewers use AI to help them judge applications. Last week, a row erupted in mathematics after OpenAI announced it had solved a notoriously difficult problem, only for another mathematician to demand answers over whether the company had used ideas he had fed into its models. The spat, and the broader issue of idea leakage through AI models, was discussed by funders at an R&D policy conference held in Paris on September 14. At worst, it threatens to undermine the correct attribution of breakthroughs, one of the foundations of modern science. The mathematics dispute around OpenAI is “certainly something we're looking at” Jason Gallo, head of the newly established Office of Metascience at the US National Science Foundation (NSF), told the Blue Sky Forum, hosted by the Organisation for Economic Co-operation and Development. The controversy started on September 8, when OpenAI announced it had cracked a mathematical problem that had gone unsolved for 90 years. Using a swarm of AI agents, millions of dollars’ worth of computing power and a new internal model, the company said it had finally proved that a long-used formula for modelling fluid dynamics breaks down under certain conditions. However, Tristan Buckmaster, a New York University mathematician who had been independently working with collaborators on the same problem using OpenAI’s Codex tool, hit back with a statement questioning whether the company had snooped on his Codex sessions to gain an advantage. OpenAI has always denied it looked up the “specific user data” of Buckmaster and colleagues. However, the company initially admitted that “while unlikely, we cannot rule out that de-identified data derived from their usage of our products helped improve our models,” although its proofs “differ significantly” from Buckmaster’s. The firm later said that, following an investigation, it concluded that Buckmaster’s Codex prompts “could not have influenced the [OpenAI] system in any way, including through training.”Funder concernYet OpenAI’s initial admission that its model might have ingested Buckmaster’s ideas, then spat them out to its own mathematicians, has scientistsworriedthat their work could leak out if shared with AI. Some AI tools give users an opt-out option to stop their sessions from training the model. If such guarantees can be trusted, this gives scientists some confidence they can use an LLM to develop an idea without the risk of it leaking out.But there’s a bigger potential vulnerability. To win money to investigate ideas, scientists have to share them with funding bodies. These grant proposals are normally kept strictly confidential, to prevent other scientists from scooping them, or to protect commercially valuable intellectual property. To preserve this confidentiality, when the NSF uses AI models, “we have to be sure that it's not making server calls out to an external party, leaking that data,” said Gallo to delegates in Paris. Leaks through reviewersHowever, funders normally use external academic reviewers to assess proposals. And there are growing worries that these reviewers are feeding grant ideas into AI models to help them make a judgement. Research funders have noticed that peer review reports and funding recommendations have “suddenly become much longer, more detailed and complex,” said Diogo Machado, head of the data science unit at Technopolis, a science and technology-focused consultancy. “The sense is that reviewers are increasingly using generative AI in their review work, even where guidelines sometimes forbid it,” he told Science|Business. “This is very difficult for funders to control or prove.” Happening at scaleThis sharing of confidential proposals with AI models is “happening at scale already,” said James Wilson, executive director of the Research on Research Institute, based at University College London, during the debate in Paris. In response, Gallo said the NSF had “strict guidance” around the uploading of proposals to AI models. But he added the NSF was mulling new types of assessment to make sure applicants were truly the originators of their ideas, such as following a short-form proposal with an oral presentation, which would allow the foundation to “really interrogate the investigator on their intellectual contribution.” Related articlesHow can Europe cope with a flood of AI-assisted research grant proposals?The risks and benefits of AI in screening research grant proposals Viewpoint: AI could make scientific overload worse“For us as funders, a lot of the ways to ensure that the application remains the applicant's own intellectual contribution, those ways are quite regressive,” Peter Clifford, head of strategy at Research Ireland, told delegates. “They're maybe ten or 15 or 20 years old, and they don't scale particularly well.” Disputes over the origin of ideas are as old as modern science. Isaac Newton and Gottfried Wilhelm Leibniz spent more than a decade arguing over who invented calculus, for example. And long before LLMs emerged, academics would agonise over how much of their early work to reveal at conferences, in case their colleagues scooped them, Jack Leahy, a senior metascience advisor to the UK government, told delegates. But AI was “accelerating” the “problems of the system,” he said.
Carrick accepts blame for Manchester United's League Cup exit
Michael Carrick took responsibility for Manchester United's embarrassing League Cup collapse to Brighton and demanded a reaction after his side were booed off the pitch at Old Trafford. An evening that began with teenager Shea Lacey marking his first senior start with a dream goal and Mason Mount adding a second in the 10th minute ended in furious fans unloading at full-time.
Oil falls as US crude inventories rise despite Saudi supply concerns
Oil prices fell on Wednesday after an unexpected build in U.S. crude inventories, while investors assessed supply risks after Saudi Arabia suspended oil loadings at its Yanbu port following an attack on its East-West pipeline to the Red Sea.
LG responds to TV spying allegations
The company claims reports about its capturing and uploading audio are misleading.
New Port of Baltimore facility to streamline grain exports for farmers
A new grain transloading facility at the Port of Baltimore is expected to simplify exports for Maryland farmers by allowing soybeans, corn, wheat and other grains to be loaded directly
Hackers breach F5 BIG-IP APM devices to deploy Linux rootkit
A Linux rootkit targeting devices in F5 BIG-IP APM environments can intercept PHP file loading and inject a fileless web shell directly into memory, avoiding the need to write malicious code to disk.
Apple Releases iOS 26.6.2
Apple today released iOS 26.6.2 with a fix for "an issue that prevents downloading a software update over a cellular connection". The minor software update is rolling out now for the iPhone 11 series and newer. To update your iPhone, open the Settings app and tap on General → Software Update. It might take a few minutes until the update is visible to everyone. iOS 26.6.2 arrives approximately three weeks after iOS 26.6.1, which delivered security fixes.
User Demonstrates Offloading DLSS 5 Neural Rendering To The Second RTX 5060 Ti Using ReShade Add-On, Restoring Original Performance
DLSS 5 Neural Rendering can now run on a second GPU with MGPU Bridge, boosting performance in games like The Blood of Dawnwalker and Cyberpunk 2077.
SportsZone Top 5 Plays of the Week 9/6/2026
SIOUX CITY, IA (KCAU) – Here is a look at this week’s top 5 plays! 5. Number five takes us to Winnebago. The pass loading for Zzikiya Redhorn, and he doesn’t take his eyes off the ball as he catches the rebound and takes it all the way to the house for the Winnebago score. […]
Football Frenzy (9/4/26)
NORTHEAST KANSAS (KSNT) – High school football in Northeast Kansas is back. This article will be updated as scores are made available. 6A 5A 4A 3A 2A 1A 8 man For more sports news, click here. Keep up with the latest breaking news in northeast Kansas by downloading our mobile app and by signing up for our news email […]
The Fierce Battle For Kids’ Eyeballs...Plus Two “Devils” Face Off
(Welcome to my weekly streaming ratings report, the single best guide to what’s popular in streaming TV and what isn’t. I’m the Entertainment Strategy Guy, a former streaming executive who now analyzes business strategy in the entertainment industry. If you were forwarded this email, please subscribe to get these insights each week.) During the week of this report, Spider-Man: Brand New Day hit theaters and went on to become one of the biggest films of all time. So, you know what that means! It’s time for a poll! Recently, Project Hail Mary came to Prime Video and became a top five film for them, but didn’t smash streaming records, a feat I cautiously thought it could accomplish. Can Spider-Man 4 do better? To set the stakes, some of the biggest films of all time do just okay on streaming. Wicked, Top Gun: Maverick, and Barbie come to mind... But unlike those films, Spider-Man: Brand New Day is a Sony film which means it’s headed to Netflix, the world’s largest streamer. Now, per friend of the newsletter and my go-to Netflix expert, Kasey Moore of What’s on Netflix, Sony films usually appear on Netflix 120 days after they premiere, which means Spidey 4 would land on Netflix on, checks notes, 28-November, one of the two biggest streaming weekends of the year! But Sony occasionally delays its biggest hits for a bit longer, to give them more time to make money on PVOD, so Spider-Man: Brand New Day might not swing onto Netflix until late December or even February 2027. Loading... Okay, on to this week’s issue. Since The Super Mario Galaxy Movie made its way to Peacock, I’m looking at the true workhorse of streaming viewership: kids films. And the surprising film that bested most Netflix animated movies. All that, plus Hulu has a new show that might be a hit (for them…?), two “Devils” square off on streaming, Zendaya tries to find some streaming magic, Netflix’s I Will Find You continues to do well, House of the Dragon tries to break into the season three top ten, some big NBC sports and summer programs, what friendly neighborhood Spider-Man character actually generates the most viewership on streaming, all the flops, bombs and misses, and a whole lot more. Let’s dive right in! (Reminder: The streaming ratings report focuses on the U.S. market and compiles data from Nielsen’s weekly top ten viewership ranks, Luminate’s Top Ten Data, JustWatch and Reelgood interest data, Samba TV household viewership, company datecdotes, Netflix hours viewed data, Google Trends, and IMDb to determine the most popular content. While most data points are current, Nielsen’s data covers the weeks of July 27th to August 2nd, 2026. You can find a link to my terminology here.) Subscribe Film - As The Super Mario Galaxy Movie Hits Peacock, How Have Kids Films Fared in 2026? Like Mario hopping on a Koopa, The Super Mario Galaxy Movie landed on Peacock on 30-July. (Yeah, I kinda forced that Mario reference.) How did it do? Great, just like the first film. It got 13.0 million hours on its opening, and that’s good enough to take second place among Peacock films on the Nielsen charts all time: As I mention all the time, Peacock just isn’t the world’s biggest streamer—in fact, nowhere close—so its films struggle to put up numbers like Netflix (or even Prime Video and Disney+). So this is a solid debut for that streamer. And as we saw last time, The Super Mario Bros. Movie had a HUGE second life on Netflix when it moved streamers. This chart doesn’t even show how long the first Mario lasted on Netflix: a whopping, KPop-esque 31 weeks. For NBCUniversal, though, the big win was the second Mario movie grossing $429 million in America and over $1 billion globally. But yeah, to add a very strong streaming performance only makes its success even greater. Illumination pretty clearly has another hit franchise on their hands. This is a good time, though, to check in on kids films so far in 2026. We’ve had a few big, big titles, a rare Netflix success, and some disappointments. Let’s start with the top animated films since November of 2025, over their first four weeks on the charts, along with a few of the “best all time” animated films: (As a reminder, if a film doesn’t make the Nielsen top ten lists, I don’t have data for it, so it gets a “zero” in terms of viewership. Now, it had some viewership, but I don’t know what it is, so it gets a zero.) I added a few titles to get an idea of what the best of the best looks like in terms of animation. And you can see, through four weeks, Zootopia 2 can now claim “one of the biggest streaming films of all time” titles. GOAT too did surprisingly well, besting Netflix’s biggest films (Swapped and In Your Dreams) through four weeks. When I’ve talked about how films can have varying performances on streaming compared to their box office, this is what I mean. Goat just managed to gross even $103 million at the box office—so it’s popular, but just barely—but did very well on streaming. Of course, what set Encanto and KPop Demon Hunters apart from the rest was their absolutely insane longevity. Here are the kids film charts through eight weeks to show that: And really KPop Demon Hunters just kept going, and no 2025 or 2026 kids films have had that sort of longevity. I can’t wait to see the sequel in theaters. Film - Two Devils and a Drama As I wrote last week, in the past, theatrical-first film flops weren’t eligible for “Miss of the Week” or “Miss of the Year” honors/dishonors, but since so many more films go to theaters first these days, I’ve made them eligible now! But it’s complicated. You see, this is a Streaming Ratings Report, emphasis on streaming, which makes analyzing this week’s two biggest streaming misses—The Drama on HBO Max (A24’s dramedy starring Zendaya and Robert Pattinson) and Prime Video’s The Devil Mouth (a killer shark movie starring young attractive people)—somewhat tricky. The Drama isn’t eligible for Luminate, but The Devil’s Mouth only had 1.5 million hours in its first week: The Devil’s Mouth notched 5.2 million hours according to Nielsen. Meanwhile, The Drama missed the charts entirely. Not great! But Prime Video is a much bigger streamer than HBO Max. That said, HBO Max titles are on Prime Video, but I think Prime Video prioritizes their original/exclusive titles, which helps boost their viewership. Complicating this further, The Drama made $48 million in the US—while A24 sells foreign distribution; the $57 million of overseas grosses helps A24 sell overseas rights for money in the future—off of a $28 million budget (which seems low for Zendaya and Pattinson, so they might have backend). The Devil’s Mouth, though, had zero dollars at the box office. The other stats for The Devil’s Mouth are bleak, and by “bleak”, I mean horrendous: a 4.7 on 15K reviews on IMDb (yikes) and a 46 on Metacritic (double yikes). The Drama, though buzzy online, only has a solid (but far from elite) 7.1 IMDb score on 121K reviews and a 59 on Metacritic (which is far, far lower than I’d have guessed). So which film is the “Miss of the Week”? I’m giving it to The Devil’s Mouth for one simple reason: it’s a horror film. And almost all horror films, even poorly reviewed horror films, should get a shot in theaters to at least try to recoup some of their budgets. The Drama made money at the box office and The Devil’s Mouth didn’t. It’s just like the situation last issue with Obsession and Masters of the Universe. Sure, Obsession didn’t pop on the streaming charts, but at least it already made it money at the box office. And as we’ve been seeing for years now and we’ll see all fall, theatrical films just do better on streaming. After all, the big “devil” winner this week was actually The Devil Wears Prada 2 with 14.8 million hours according to Nielsen. That’s right, The Devil Wears Prada 2—one of the box office success stories of May—came to streaming and parlayed its $220 million box office into 14.8 million Nielsen hours. Disney also put out a datecdote that it had 15.2 million “views” globally in five days, but again we don’t have a lot of great context for what that means. Here’s how it stacks up to other Hulu films, since I’m categorizing this mainly as a Hulu title (though it was also on Disney+): Share Quick Notes on Film We’re just getting started with this issue, but the rest is for paid subscribers of the Entertainment Strategy Guy, so if you’d like to find out… Whether Hulu has a new hit show on its hands... The top first and third seasons of all time and which shows are climbing those charts… What version of Spidey is the most popular... What sport and TV show are performing better on NBC than the NBA... All the flops, bombs and misses... ...please subscribe! We can only keep doing this great work with your support. Coming Soon! Next week, wow, things stay so busy. One of streaming’s biggest shows of all time, Ted Lasso, returns along with new YA-ish shows from Hulu (The Shards, from Ryan Murphy) and Prime Video (Sterling Point). The Challenge moves to Paramount+, plus My Life with the Walter Boys returns to Netflix and Futurama returns to Hulu. And Netflix has a new horror film, The Last House. Lots to talk about! All month, things stay busy. We’re talking Reacher on Prime Video! Outer Banks and Tires on Netflix! Lanterns on HBO Max! UFC 330 on Paramount+ and the MLB Field of Dreams game on Netflix. And a ton of smaller films come to streaming. Long term, the VMAs are back on 27-Sep! I feel like I haven’t heard much about this show in past years, but I’m seeing more news coverage and buzz for this show than in past years. I’m curious to see how it does.
gamescom 2026’s Steam wishlist winners
gamescom wrapped up last weekend, and it was bloody fantastic. It really felt like the industry had a spring in its step again. The show itself was BUZZING, with 368K attendees (+3% over last year), maxed-out exhibit space, and Opening Night Live pulling over 78M online views (+7.5%). Let’s have a look at the games at the event that brought in the most new Steam wishlists between August 23 and 30. One quick thing before we start, and stop me if you’ve heard this one before, but wishlists are NOT a barometer for launch sales (they’re bookmarks, not baskets is what we always say over here at Alinea). With that aside, wishlists are a nice gauge of hype, so let’s dive in. Witche(r)s and Warlocks take the top two spots The biggest game of the event (by new wishlists, anyways) wasn’t a game at all. It was an expansion. The Witcher 3: Wild Hunt’s upcoming expansion, Songs of the Past, nabbed 451K Steam wishlists by the end of the weekend. It’s a big example of something we bang on about constantly: proven IP with built-in fandom is the most powerful force in this market. But said IP can also ripple into other external success stories (more on that later!). Anyway, Songs of the Past debuted at Opening Night Live, and it’s the first Witcher 3 expansion in over a decade, landing next year after a full Witcher 3 remaster (free for existing owners!) arrives next month. The Witcher 3 has sold around 66M copies, and I’ll be one of those owners jumping back in to refresh myself on the core game (like there isn’t enough to play!), then play the expansion next year. This is all part of CDPR’s plan. The studio is clearly whetting everyone’s appetite for The Witcher 4 in 2028 via a classic bridge strategy. Want more free estimates and insights in your inbox twice a week? Subscribe below! Subscribe Releasing a free 4K remaster next month – followed by a proper expansion in 2027 – keeps IP top-of-mind across a huge gap between content (games-wise, anyway). It kind of echoes Capcom’s evergreen catalogue strategy. And outsourcing co-development to Fool’s Theory (a studio led by trusted Witcher 3 veterans) keeps CDPR’s internal headcount focused on The Witcher 4. #2 by wishlists between Aug 23 and 30 went to another giant IP: Warlock: Dungeons & Dragons, published by Wizards of the Coast and developed by Invoke Studios (a team of ex-Far Cry, Watch Dogs, and Splinter Cell devs). It’s due to launch in 2027, and it looks brilliant so far. It’s a dark-fantasy open world with third-person spell-slinging. It’s a very Witcher-ish aesthetic. As you might expect, 48% of Warlock’s Steam wishlisters have played Baldur’s Gate 3, and 46% have played The Witcher 3. One to watch. For sure! Mash-ups: Anime GTA (!?) and a city-builder for kids (and adults) Ananta took #3 with 302K new wishlists. It’s an open-world game that looks like GTA and Persona had a baby. It was published by NetEase, developed by Naked Rain, and got a trailer at Opening Night Live. I got hands-on at the show, and it wears its influences openly; there’s literal web-slinging (very Spider-Man) and a LOAD of mini-games (Like a Dragon). It’s free-to-play and out in January. For Ananta, Naked Rain is taking the free-to-play, cosmetics-only route. That’s certainly a choice for a single-player open-world game. Without gacha mechanics driving high ARPU, Ananta is betting purely on top-of-funnel reach and long-term retention. This puts pressure on skin appeal when there’s no multiplayer lobby to flex in (although, co-op is reportedly coming later). Want more free estimates and insights in your inbox twice a week? Subscribe below! Subscribe Then there’s Lego Skylines (273K wishlists between Aug 23 and 30), which does exactly what it says on the tin. It’s a cosy Lego twist on the city-builder genre, as shallow or as deep as you fancy (five min quick sesh or ten-hour no-lifer session; latter for me). There’s some family-friendly stuff in there like couch co-op, which is only really viable on console. Iceflake Studios, which Paradox acquired back in 2020 is developing this one. So Paradox is probably looking for a family-friendly onramp to its rad Cities: Skylines games. The overlap there is plain, with 33% of Lego Skylines wishlisters having played Cities: Skylines and 16% Skylines II. There’s no doubt plenty of parents here (and uncles, ahem) planning to indoctrinate the next gen of city-builders. I’ll definitely be playing with my youngest nephew. This one could be big and educational. I’m very curious to see more. Want more free estimates and insights in your inbox twice a week? Subscribe below! Subscribe Mature IP, new and old, rounds out the list As for the rest of the list of gamescom games by new wishlists between Aug 23 and 30: Heroes of Might & Magic III Remake got 189K new wishlists. It remakes the iconic 1999 turn-based-strategy classic in full 3D and 4K. It follows this year’s Heroes of Might & Magic: Olden Era (~1.8M copies sold, as per our estimates), and around a quarter of the III Remake wishlisters have played Olden Era, including 7% of them this month. Wardogs – a $40 PvP shooter – added another 146K to an already massive pile (nearly 1.6M total). It’s published by Team17 and made by Bulkhead. It plays somewhere between the tactical squad stuff of … Squad and the all-out-chaos battlefield of … Battlefield. It’s up to 100 players, and it’s out in about 10 days. It’s already pre-sold over 500K copies on Steam ($16.6M), boosted by a pre-order-only beta. It sold another 70K copies during gamescom, too. Afterworld managed 121K new wishlists. It’s a Paradox-made post-apocalyptic grand-strategy game, and it was announced at the show. It’s a departure from Paradox’s usual historical stuff, and the studio’s first new setting in over a decade. The Paradox purists are into it, with over half its wishlisters having played Hearts of Iron IV (19% this month) and 41% Crusader Kings III (17% this month). Metro 2039 got 110K during the show, tunnelling its total past 1.3M. Nearly 40% of wishlisters have played 2020’s Metro Exodus, the previous mainline title in the series. The Metro novels are Russian, dev 4A Games is Ukrainian, and it’s set in Moscow, so it’s unsurprising that – after the US (17.3% of wishlisters) – the next two countries are Russia (12.7%) and Ukraine (10.0%). The Blood of the Dawnwalker drew in (the blood of) 108K new wishlisters, now totalling an impressive 1.7M (at a $70 price point, though). It’s an open-world dark-fantasy action-RPG set in 14th-century Europe (with vampires!), made by some former Witcher devs at Rebel Wolves. Again, it shows. Over half of Blood of the Dawnwalker’s wishlisters have played The Witcher 3, and its #2 market is Poland, home of Rebel Wolves. That’s that hometown-hero effect we keep banging on about again! And Blood of the Dawnwalker plays very well, by the way. While it looks like a Witcher clone on the surface, it actually brings a lot to the table mechanics-wise. I love the whole time-as-a-currency spin on RPG gameplay. It gives the game a different, refreshing pace. Essentially, you start Blood of the Dawnwalker with a pool of 480 action points spread across 30 days and 30 nights. As for how the game performs, we’ll be posting some numbers next week, so hit subscribe below to get that in your inbox when it’s ready. Subscribe The Witcher’s legacy ripples Make no mistake: on a high level, this was The Witcher‘s gamescom. The #1 game by wishlisters was a Witcher 3 DLC. Warlock is clearly Witcher-inspired – as is Blood of the Dawnwalker, which was made by ex-Witcher devs. There was a lot of other cool stuff on the floor that we’ll no doubt be covering in subsequent newsletters, though! As always, wishlists went through the roof for sequels, remakes, DLC, and licensed IP. As always, though, these wishlists will likely convert to about 5% to 10% of launch-day buyers, with the probable exception of The Witcher DLC – the exception that proves the rule. Either way, we’ve got a nice, diverse list here – one that is very Witcher-y, Paradox-y, and – if you’re a fan of great games on Steam – a bloody good list. We even got a bunch of new IP to get excited for. Great gamescom! Even if we are still processing that unfathomable number of mini Kölsch beers! One more thing: GTA 6 broke the internet (again) Move aside, Spider-Man and The Odyssey (I love you both, really). But the decade’s defining entertainment zeitgeist is round the corner. Unless you’ve been living under a rock, you’ll know that Rockstar dropped an extended look at GTA 6 last week. It was INSANE, the most impressed I’ve ever been by a showcase like this. You can see the budget seeping through every second of the extended look, and this level of game is unobtainable for any other publisher. Everyone’s going nuts, and rightly so. GTA 6 is out in holiday 2026, and it’ll be the biggest entertainment release ever, full stop. For those without a PS5 or Xbox Series console, a $500-$900 console is a big ask in this economy, but GTA will pull a lot of last-generation-console fence sitters over the current-gen fence either way: The autumn timing puts GTA 6 on (digital) shelves for the holidays, meaning a PS5 or Xbox (let’s be honest, mostly the former) will be on many a Christmas and Hanukkah list. Demand will outweigh the struggling supply, though, and PS5 Pros are already sold out and being sold at a markup via scalpers. Every console that isn’t on a shelf is a sale deferred or lost, and worse, it hands scalpers the margin. GTA has cultural clout that reaches everyone. It’s been a mainstream fixture since 2001’s GTA III and its clout has only grown exponentially since then. There’s a touchpoint for every generation alive. The FOMO-word-of-mouth double punch does the rest. Loading... It’ll reactivate the lockdown cohort, the millions who picked up gaming in 2020-21 and churned. GTA 6 will bring them back. Remember when College Football 25 brought a bunch of US gamers to current-gen consoles? GTA will do that globally. Once they’re back (or upgraded), and all GTA’d out, the fresh current-gen-console folks will check out the rest of the current-gen catalogue they missed. GTA Online, if not free to play, will also lead to a boost for PS+ and Xbox Game Pass, another boost for the ecosystem players. Want more free estimates and insights in your inbox twice a week? Subscribe below! Subscribe Even some of the smug PCMR 60FPS crew won’t be able to resist their FOMO and will bite the bullet on a console to play this one early (then double dip). The consoles cost the same as a decent stick of RAM these days anyway lmao… Don’t get me wrong, though. GTA 6 is no silver bullet for the mature console market, but it does prime the current gen for a couple of strong swan-song years. And it’ll also give the consoles more users from which to maximise ARPU. ‘’We are in the latter half of the lifetime cycle of [the PS5],’’ said Sony President and CEO Hiroki Totoki in a Wall Street Journal interview. ‘’That means we do not need to aggressively sell the console today; rather than that, we are focusing on the current revenue from the current user base.’’ The biggest, best premium games always arrive late in a console cycle, so getting more players aboard now, ahead of that content storm (the new current-gen-only Call of Duty included), is a clear win for the platforms. On the Netflix thing, though: let’s all chill out a bit, like. Half my LinkedIn feed was going bonkers treating the reveal debuting on Netflix like it’s the future of entertainment. It wasn’t. It was a six-hour head start, and Netflix left that detail out of the “first-of-its-kind partnership” press release. The extended look hit Netflix at 3pm ET on 27 August, then went up on YouTube and Rockstar’s site for everyone six hours later. And it was all over social immediately, anyways. This reveal was going to reach an untold number of people whether it debuted on Netflix, YouTube, a Times Square screen, or a group chat. The last trailer did nearly half a billion views in a day with no gatekeeper at all. Netflix just stood next to that audience for six hours, borrowing it. It’s was a tidy deal for Netflix, sure. But every outlet on earth was breaking the thing down frame by frame by the evening anyway. Anyway. I’m hype. And we’ll be covering some GTA numbers on the newsletter soon. So hit subscribe below to get that in your inbox when it’s ready! Subscribe The last word Reply to this email – or reach out here – if you have any feedback for the newsletter – or want to request a game for us to cover. [Alinea Analytics boasts the most accurate PC and console estimates in the business. Game makers use our platform to understand their audience, keep an eye on the competition, monitor sales trends, and spot new opportunities. We equip game studios and financial institutions with accurate data and the confidence to make smarter, data-driven decisions. Want to talk about all things games market data? We’d love to chat!]
Patriots reloading by using strength in numbers
YATESVILLE, LUZERNE COUNTY (WBRE/WYOU)— The energy was high, and the pads were popping ahead of Pittston Area’s season opener against East Stroudsburg South. The team will have a new look, though, as they look to replace eight starters on both sides of the ball. “We put ourselves in a position a year ago where we […]
Vertical 6-Bike Rack Review: Thule Revert vs. Velocirax Max vs. Yakima HangTight
Three big-name vertical hitch racks go head-to-head on loading ease, trail stability, weight capacity, and overall value.
Review: In a deluge of free entertainment, Alternating Currents 2026 flowed into a loading dock, record store and the QC's coolest venues
The Quad Cities’ annual Alternating Currents Festival landed on a rainy weekend this year (Aug. 13-16), but the weather didn’t dampen the spirits gathering for over 200 events across Davenport, Bettendorf and Rock Island. The smiles outlasted the downpours as free music, art, comedy and film events carried on, albeit with few a cancellations, a […]
Google ‘gradually’ rolling out Android’s ‘advanced’ sideloading ahead of developer verification
Android is now "gradually" rolling out the new "advanced" sideloading flow used with the developer verification system.
Mozilla Firefox 155 Enters Public Beta Testing, Here's What to Expect
Firefox 155 open-source web browser is now available for public beta testing with support for Happy Eyeballs v3 for faster page loading.
Google begins rolling out Android’s controversial new sideloading flow
The new advanced flow for sideloading adds a 24-hour delay and extra warnings for apps from unverified developers.
Hinton volleyball ready to roll ahead of 2026 season
HINTON, Iowa (KCAU) – The Hinton volleyball program has its sights set on reloading ahead of a year full of opportunity. Hinton finished last season as the Class 2A runner-up with a 30-3 overall record, entering this year with a new look. The Blackhawks graduated three depth pieces and a trio of starters, including IGHSAU […]
Mountain biking is an investment in children, community
On Father’s Day, my son and I were unloading our mountain bikes at Slaughter Pen, looking forward to spending the day together on the trails. As we were getting ready...
Washburn Athletics announces 2026 hall of fame class
TOPEKA (KSNT) – A new set of Ichabods will be inducted into the Washburn Athletics Hall of Fame on Oct. 9. Entering the hall of fame this year is one former head coach and three former athletes. For more sports news, click here. Keep up with the latest breaking news in northeast Kansas by downloading our mobile app and by […]
Washington Capitals Offseason Recap
In this series, Heavy Sports is looking at what every NHL team has done over the offseason. Next up is the Washington Capitals. Washington had a disappointing season this past year as they dropped out of the postseason, posting a 43-30-9 record. The team has been busy loading up over the offseason in the hopes […] The post Washington Capitals Offseason Recap appeared first on HEAVY.
SNAP changes push more North Carolina families to food pantries, food banks
In some places, locally supported nonprofits are stepping up, but leaders say their work can never fill the hole left by cuts to the federal food stamps program.
Are AI scribes useful tools in medical education, or a crutch that imperils learning?
Educators want future physicians to be fluent in AI scribes, but they worry what off-loading clinical notes to AI could do to learning.
2026 Tobacco Heritage Day at L.W. Paul Living History Farm in Conway
For many in attendance of Tobacco Heritage Day on Saturday, Aug. 1, tobacco farming wasn't just history — it was part of their childhood. Generations of Horry County families made a living in the tobacco fields, where children often spent summers priming leaves, loading wagons or stringing tobacco by hand before it was cured in barns. Once the county's dominant cash crop, tobacco shaped the local economy and way of life for decades, with thousands of acres planted each year and warehouses bustling during auction season.
Splatoon Raiders has a loading screen Easter egg that can also crash the game
What a load of...
Business coalition opposes NYC Council last-mile delivery legislation
Opponents to Intro. 518 say the City Council bill would restructure the city’s last-mile delivery network.
Xenoblade Chronicles 2 version 3.0.0 is now live, adding support for the Switch 2 Edition
The update brings in support for the Switch 2 Edition with pre-loading now live.
EA Is Salivating At The Prospect Of Loading Games Up With Ads
The last un-monetized frontier of gaming: virtual billboards
GTA 6 Preloading: Start Date And What To Know About Digital And Physical Copies
GTA 6 approaches.
Video shows totes of clothes burning outside Upstate business
Firefighters in Duncan, South Carolina, responded to a fire at a business early Saturday morning and found totes of clothes burning on a loading dock.
There's one group of Android users Google's new sideloading rules won't affect
LineageOS says Google's new developer verification system for sideloading apps on Android will not affect devices running its custom ROM.
3 Reasons Why I'm Loading Up on Penn Entertainment in the Second Half of 2026
Penn Entertainment stock is soaring and it has the potential to build on those gains in the second half of 2026.
3 Reasons Why I'm Loading Up on Penn Entertainment in the Second Half of 2026
Penn Entertainment stock is soaring and it has the potential to build on those gains in the second half of 2026.
Exec blames malware threat for Amazon blocking sideloading on new Fire Sticks
New Fire Stick OS helps Amazon block third-party homepage launchers, ad blockers.
Eastern Conference Contender Showing Interest in Kings G LaVine
Detroit eyes Zach LaVine’s scoring punch as Sacramento weighs offloading his massive $49 million contract. The Pistons could leverage young assets to bolster Cade Cunningham’s perimeter support.
Assassin’s Creed Black Flag Resynced: Here’s What the Remake Looks (and Plays) Like Compared to the Original
After playing through a few missions of Assassin's Creed Black Flag Resynced recently, I now have a better understanding of what to expect from this ambitious remake, which included narrative additions, new options for carrying out my assassinations, and other bits and baubles along the way that have increased my interest in replaying through Kenway’s epic tale of pillaging and unloading flintlocks into the chests of unsuspecting lawmen.
Steam Workshop abused to spread malware via Wallpaper Engine app
Threat actors are abusing Steam Workshop, Valve's community hub for downloading game-related content, to push various malware hidden in wallpaper packages.
Pokémon Champions Now Available For Mobile Devices!
Good day, Pokémon trainers! It’s finally time, after months of waiting, Pokémon Champions is now available on Android and iOS devices! If you’ve been eager to participate in thrilling battles with your favorite Pokémon on your smartdevice, now is your chance. You can get Pokémon Champions by downloading it frome either the Google Play or […]
Is SMIC N+3’s Metal Pitch Smaller than Intel 18A’s?
Almost four years ago, we published that SMIC had started shipping 7 nm (N+1) chips. Now, SMIC is shipping its third-generation 7 nm (N+3) process in Huawei’s Kirin 9030, with a minimum metal pitch of 32.5 nm, about 10% tighter than the 36 nm minimum metal pitch shipping in Intel’s latest Panther Lake CPUs on 18A. The headline is true, but incomplete cherry picked metric. N+3 reaches the density of TSMC N6 through aggressive DUV multi-patterning and design-technology co-optimization (DTCO), but it pays for that in complexity, efficiency and process control. We found this and more in our reverse engineering and teardown where we cover SMIC’s N3 process technology, Huawei’s packaging, memory, architecture, and more. SemiAnalysis has been building a state-of-the-art teardown lab in Oregon capable of analyzing the world’s most advanced and important chips over the last year and half. We have already generated revenue on advanced datacenter chip teardowns including our recent reverse engineering of a major TSMC customer’s COUPE CPO optical engine + EIC 3D stack. This is the first public report from the SemiAnalysis Teardown Engineering & Evaluation Lab, or STEEL for short. The lab is aggressively scaling up and out and we’re excited to announce it publicly. This is a bit of inconvenient timing for TechInsights as they are private equity owned and currently being sold while having enjoyed virtually no credible competition for decades. This has led to TechInsights underinvesting in CAPEX. SemiAnalysis exceeds TechInsights in revenue despite no venture or private equity ownership and being founded only 6 years ago. Because we have no external investors and are founder led, we move faster, build faster, and we can release client chip teardowns for free regularly, while focusing on datacenter for our major clients. Here’s the first public image from our lab, the HiSilicon Kirin 9030 Pro SoC: This report will detail our teardown of the Kirin 9030 and our findings on SMIC’s N+3 process, the most advanced in China. For comparison, we’ll show our teardown of the MediaTek Helio G99, made on TSMC N6. Through this comparison, we can look at the effect of export controls – SMIC N+3 and TSMC N6 are comparable nodes, but one is heavily export-controlled, the other free to use the West’s most advanced equipment. Here we see both China’s progress and constraints. SMIC N+3 reaches TSMC N6-class logic density, but it requires far more aggressive DUV multi-patterning, so it does not match N6 on process maturity or cost. The Kirin 9030 Pro performs similarly to three-year-old Android flagships, and trails far behind the current flagship SoCs from Apple, Qualcomm, MediaTek, and Samsung. The efficiency gap is even wider. Export controls have not stopped Huawei and SMIC from shipping advanced silicon, but they have forced a different path. Without EUV, SMIC is leaning harder on DUV multi-patterning, DTCO, and increasingly complex integration. The roadmap continues forward through tighter design rules and backside power, but each step adds cost and process risk. Huawei’s τ scaling and LogicFolding show another path: stacking active logic and recovering density through advanced packaging and system-technology co-optimization (STCO). To understand the Kirin 9030, we must first understand Huawei’s SoC history. HiSilicon is Huawei’s chip design arm, responsible for the Kirin smartphone SoCs, Kunpeng server CPUs, Ascend AI accelerators, and switch/router networking silicon. Before export controls, Huawei was TSMC’s largest customer – the only customer on TSMC’s first EUV node, N7+, and among the first on N5, alongside Apple. That ended in late 2020. Huawei switched to Qualcomm SoCs in its flagship smartphones, though export controls limited them to 4G-only variants. In late 2023, Huawei returned to in-house silicon with the Kirin 9000s, a successor to the Kirin 9000, fabricated on SMIC N+2 instead of TSMC N5. In the following years, they released the Kirin 9010 and 9020 on the same N+2 process. These chips used Huawei’s in-house TaiShan CPU cores and Maleoon GPU. We have not torn down a Kirin 9020 ourselves, so the predecessor die shot is from Kurnal. The die shots show how Huawei spent its silicon budget: which functional blocks are where, and how their areas compare to the predecessor. First, a quick guide to the major blocks on the die. The total die area is nearly identical, but the 9030 uses that area more aggressively. A denser process lets Huawei fit an extra middle CPU core, more GPU and NPU cores, and larger caches into the same footprint. In contrast, the Helio G99 is a much smaller, low-cost SoC, built for budget smartphones rather than a flagship device. While the Kirin 9030 is ~140 mm², the G99 is only ~29 mm², roughly one-fifth the area. The underlying TSMC process technology, however, is directly comparable as a baseline for analyzing SMIC’s. The Kirin 9030 is an evolutionary refresh, not a clean-sheet design. Its CPU, GPU and NPU cores carry over the 9020’s families, and the gains come from three levers: the SMIC N+2-to-N+3 process step, DTCO and floorplan work, and incremental microarchitecture. Area is where the first two show up, and the 9030 scales well here. Performance and efficiency are the harder test. Huawei’s design holds up better than its node would suggest, but the chip still trails, both because N+3 sits behind the leading-edge nodes and because its cores, while competent, remain a few generations behind the newest designs. The new prime core is an incremental update. The main changes are a 10% frequency increase from 2.5 GHz to 2.75 GHz and a doubling in the L2 cache from 1 MiB to 2 MiB. Despite the increased cache, the core size decreased by 7.6%. Excluding the private L2 cache, the core size decreased by 21%. This is a large reduction for an incremental node. Compared with the TaiShan New V120 core in the Kirin 9020, the Kirin 9030’s middle core is almost unchanged architecturally, yet each core shrinks by ~22%. Most of that comes from the move from N+2 to N+3, with layout likely accounting for the rest. Visually, the most noticeable change is the increase from 3 middle cores to 4. There is also a 20% increase in the shared L3 cache of the big cluster. This helps to improve multi-core performance without sacrificing much in terms of area. Even with each core shrinking, the big CPU cluster’s total area is essentially unchanged. The per-core savings went back into an additional middle core and larger caches. The tiny cores shrank less than the prime core (excluding its L2 cache) and less than the middle cores. This is likely because fixed overhead is a larger share on a small core. We cannot resolve any architectural changes from the die shot alone, but the per-clock and efficiency gains shown below point to more than pure process and layout scaling. The area reduction was offset by a doubling of the shared L2 cache from 2 MiB to 4 MiB, leaving the total tiny CPU cluster area slightly larger. Area is the easiest improvement to see from a die shot, but it is only one part of PPA (power, performance, area). For modern logic, power and performance matter just as much, and often more. Since Dennard scaling broke down in the mid-2000s, voltage and frequency have not scaled in step with transistor dimensions, so each node has had to fight harder for gains in performance and efficiency. The starkest comparison is not Kirin 9020 versus Kirin 9030 Pro. Apple’s efficiency cores run circles around Huawei’s prime core. Apple’s low-power core delivers 20% higher integer performance while drawing only 1 W, compared with 4.5 W for Huawei’s prime core. N+3 matches TSMC N6, but N6 is several generations old. Apple and Qualcomm build on N4 and N3P, which are denser and sit on a better voltage-frequency curve, giving them a larger transistor budget and more performance per watt. The 9030’s own cores did improve. The middle and tiny cores gained 17% and 14% in per-clock integer performance over the 9020, with floating-point flat on the middle core and up 11% on the tiny. The tiny core improves cleanly, with performance rising while power falls and efficiency increases by 45% in integer and 24% in floating point. The middle core is mixed: integer performance rises but power rises faster, cutting integer efficiency by 7%, while lower power lifts floating-point efficiency 16%. Per-clock gains at the same or lower frequency are microarchitectural, so the cores are tuned, not just shrunk. Both also failed to hold their rated maximum frequencies, pointing to thermal, power, or stability limits. Per clock, the middle core sits around Arm Cortex-A720 and the tiny core near the Cortex-A520; absolute performance trails because Huawei clocks them much lower. The prime core is roughly Cortex-X2 class per clock, a 2021 design. Apple’s 2020 M1 Firestorm core is still 35% higher per clock and 57% faster in absolute integer performance at a similar 4.5 W. The current leading edge is further ahead again: the Apple M5 P-core is 60% higher per clock and 2.7× faster, the Arm C1 Ultra 45% higher and 2× faster. Matching older high-end cores per clock is a genuine design achievement. What Huawei cannot match is the voltage-frequency curve and transistor budget of leading-edge nodes, which let Apple, Qualcomm and others spend more transistors in the same area on wider cores, larger caches and deeper buffers while running at lower voltage. Huawei’s LogicFolding roadmap is one answer, stacking active logic to recover density and shorten signal paths. We return to it later. The GPU compute units (CUs) changed more visibly than the CPU cores, moving to a more rectangular layout for both the arithmetic logic unit (ALU) clusters and the CU overall. Even with ray-tracing support added, a CU shrank ~28%. However, that shrink is offset by the increase from 4 to 6 CUs and the area outside the CUs grew 33%. Overall, the GPU cluster is larger by ~10%. The GPU is where Huawei makes its biggest gains. The Maleoon 935 is not competitive with current flagships, but it is a large step up from the 920 and reaches older-flagship territory. In 3DMark it is 70% faster in Wild Life Extreme (WLE) and 79% faster in Steel Nomad Light (SNL) than the 920; with 11% higher clocks and 50% more CUs, the ~67% theoretical uplift roughly matches WLE and is beaten by SNL. It edges ahead of the Snapdragon 8+ Gen 1 in WLE and SNL, and the Dimensity 9200 and Apple A16 in WLE, but stays far behind newer parts: the Snapdragon 8 Elite Gen 5 and Dimensity 9500 are ~2.4–2.6× faster in WLE and ~3.2× faster in SNL. The Maleoon 935 is Huawei’s first GPU with hardware-accelerated ray tracing; there it lands slightly ahead of the Exynos 2200, and on par with the Apple A16, with current flagships up to 3.7× faster. The Neural Processing Unit (NPU) saw the largest structural changes of any block, moving from a Lite and a Tiny core in the Kirin 9020 to a Lite and two Tiny cores in the Kirin 9030. Both core types also show significant layout changes. This is a reversal in Huawei’s NPU design. The Kirin 9000 5G, its last flagship chip on TSMC N5, used two Lite and one Tiny core. The series of SoCs on SMIC N+2 moved to one Lite and one Tiny core, likely for area savings. With the Kirin 9030, Huawei has shifted back toward a larger multi-core NPU cluster, but with the additional area going to a Tiny core rather than a Lite core. We’re diving deep into the most advanced datacenter and AI hardware hitting the market. To learn more about what’s in the pipeline or to commission a custom teardown, contact sales@semianalysis.com. Interested in joining us on this ride and think you can be a difference maker? Check out our Careers page. Before diving into the process stack, the package and memory are worth separating from the SoC itself. The Pro variant of the Kirin 9030 carries 12 GB of Samsung DRAM, with two stacks of four dies each. The dies were identified as the K4L2E165YD, a 12 Gb LPDDR5X-9600 device fabricated on Samsung’s 1a node, the fourth generation of its 10 nm-class DRAM after 1x, 1y and 1z. 1a has shipped in volume since 2022, so this is current memory rather than older-node inventory. The 16 GB Pro Max variants we obtained were found with both CXMT and Samsung packages. The CXMT package is marked CXDD7JEDM, with two stacks of four dies, packaged in week 45 of 2025. The inferred die dimensions from X-ray computed tomography (CT) are consistent with a known density of ~0.3 Gib/mm² for the CXMT G4 process, roughly equivalent to other manufacturers’ 1z processes. The Kirin 9030 uses a typical integrated package-on-package (iPoP) stack: multiple DRAM dies in a memory package sit above an organic redistribution layer (RDL) interposer, which sits above the SoC and package substrate. The full package is then mounted to the printed circuit board (PCB) through ball-grid array (BGA) solder bumps. The memory package substrate is a thin bismaleimide-triazine (BT) laminate carrying the LPDDR5X stack. The organic RDL interposer over the SoC routes the PoP signals around the die and carries possible dummy thermal copper pillars. The package substrate, a thicker Ajinomoto Build-up Film (ABF) build-up over a BT core, fans the flip-chip bumps out to BGA pitch and embeds the power planes. The whole stack is organic. The only silicon is the SoC and the LPDDR5X dies; there is no silicon interposer. Keeping it all-organic brings the package’s coefficient of thermal expansion (CTE) close to the PCB’s, reducing board-level warpage, and avoids the cost of a silicon interposer the SoC’s bandwidth does not need. In an iPoP stack, the memory package connects to the organic RDL interposer through an array of solder bumps. Underfill fills the gap around those bumps, adding stiffness and protecting the joints from mechanical stress. The Pro and Pro Max variants differ here, which we cover behind the paywall. The die shot and architecture tell us how Huawei allocated its silicon budget. The process tells us what SMIC can manufacture. We use the Helio G99 as the process reference for TSMC N6. Both SMIC N+3 and TSMC N6 are evolutions of previous 7 nm-class nodes. We used targeted TEM cross-sections through logic and memory regions, imaged in both fin-cut and gate-cut directions. Each cross-section caption gives its horizontal field width (HFW), the real width of the imaged area. We start at the transistor fins, then move up through standard cells, local interconnect, and SRAM. SMIC has not overtaken Intel or TSMC. It uses aggressive DUV scaling and DTCO to reach N6-class density, but that density doesn’t translate into comparable performance and efficiency, for two reasons: the node gap to leading-edge nodes, and Huawei’s core designs. Fin Profile One of the most important knobs in a FinFET process is the fin profile: the shape of an individual fin and the channel where current passes from source to drain. The ideal fin is tall, narrow, and nearly vertical. A taller fin increases effective channel width, while a narrower fin improves electrostatic control by thinning the body the gate must control. Push either too far, and the process pays for it: weaker drive current, fragile fins, taper, footing and line-edge variation that hit yield and device variability. The Intel 22 nm, 14 nm, and 10 nm fin cross-sections show how FinFET nodes have improved over time. 22 nm fins were a first-generation structure, relatively short, wide and strongly tapered. The shape limits current density and reduces gate control uniformity across the height of the fin. At 14 nm and 10 nm, Intel pushed the fins taller and narrower while also making the sidewalls more vertical. Rather than shrinking the device, these changes increase the effective channel width per fin and improve electrostatic control. The trade-off is that taller fins at tighter pitches are much more difficult to manufacture. Now, let’s compare the Helio G99 on TSMC N6 with the Kirin 9030 on SMIC N+3. Both processes are in the same class, with fin pitch of 30-32 nm on N+3 and 34 nm in our N6 cross-section. The pitch for N6 is especially interesting as N7’s HD library is generally listed with a 33 nm fin pitch, and N6 did not shrink pitches directly. Its density gains came from DTCO instead of tighter pitches. The 34 nm pitch was stable across our sampled region and serves more as a comparison against the SMIC N+3 we have not investigated further. Pinning down N+3’s fin patterning scheme takes more than one core unit. The CPU cores show a dense ~32 nm pitch, with the pitch between N-P fin pairs alternating between 78 and 88 nm. Logic alone may be consistent with dual-pitch mandrels of 120 and 110 nm, but this is a complex and unusual approach. Combining the pitch from the the 8T SRAM, which has more complex repeat unit, with the CPU core sequence allows us to reverse engineer the patterning steps with more confidence. As both the logic and SRAM should share the same base grid, a single CD mandrel lithography pattern with 128 nm pitch undergoing SAQP produces a die-wide ~32 nm grid (128 nm/4), which supports the pitch sequencing seen in both logic and SRAM cells. In the sampled cross-sections, N+3 shows a taller, narrower, higher-aspect-ratio fin than N6. The measured fin aspect ratio is ~9.5:1 on N+3 versus 7.8:1 on N6. N+3 also shows less top rounding, with an estimated radius of ~2 nm, compared with 2.8 nm on N6. Even though the fin widths differ, the ratio of top rounding to fin width tells the same story, with N+3 at 0.37 and N6 at 0.44. In a geometric sense, lower is better; a perfectly rectangular fin would have no top-rounding penalty. These are single-digit-nanometer features measured from a handful of cuts, so treat the absolute numbers as approximate. The important result is the relative gap: N+3’s fins are consistently taller, narrower and less rounded than N6’s. We’re diving deep into the most advanced datacenter and AI hardware hitting the market. To learn more about what’s in the pipeline or to commission a custom teardown, contact sales@semianalysis.com. Interested in joining us on this ride and think you can be a difference maker? Check out our Careers page. Standard Cell A standard cell is the basic building block of chip layout: a fixed-height row pairing one NMOS and one PMOS transistor that share a gate, tiled in a grid to build logic blocks. The key dimensions are contacted gate pitch (CGP), cell height (CH), fin count, and the lower-metal routing grid. To measure density, we use the Bohr metric: a weighted average of NAND2 gate area (60%) and scan flip-flop area (40%). This represents a realistic mix of combinational and sequential logic. This metric has its limitations, especially for complex cell layouts like TSMC’s FinFLEX, which alternates cells with different fin counts. Even so, it is the best metric for a pure process-level comparison. Another important measurement is the fin pitch; it refers to the distance between two fins of the same transistor. In a FinFET process, multiple fins are used in each transistor to increase the drive current and thus performance. TSMC N6 ships both a high-density (HD) library with 2 PMOS and 2 NMOS fins per cell, and a high-performance (HP) library with 3 of each. More fins under the shared gate mean more effective channel width. HP cells switch harder at the cost of area. Designers mix the two on a die, primarily spending HP cells on timing-critical paths, and matching their PPA targets. In the Cortex-A55 core of the Helio G99, we found a cell height of 240 nm for the HD cell. MediaTek has used HD cells in the G99 to minimize die size and thus cost. As an SoC for budget smartphones at ~$100, this is essential. By contrast, we found only one library in the Kirin 9030, with 2 NMOS and 2 PMOS fins. This suggests a narrower library strategy than TSMC N6, where both HD and HP libraries are widely used. This likely reflects the smaller customer base and the more constrained domestic design and electronic design automation (EDA) ecosystem. In all three CPU cores of the Kirin 9030, we found cell heights of 228 nm, 5% smaller than on N6. This is also a reduction of 9.5% over SMIC N+2’s cell height of 252 nm. SMIC N+3 and TSMC N6’s HD library both feature a CGP of 57 nm. For SMIC, this is a 9.5% shrink over N+2. In the past, CGP and cell height alone may have been enough to compare transistor density. Now, however, we must consider scaling boosters and DTCO as well. SMIC’s density gain does not come from EUV. It comes from using every available DTCO booster aggressively. First is fin depopulation: reducing the number of NMOS and PMOS fins in each cell. The first FinFET nodes started with 3 or 4 fins for each transistor. SMIC N+3 and TSMC N6 HD both use only 2 fins per transistor, trading drive strength for density. Next is contact over active gate (COAG). By landing the gate contact directly over the active gate, instead of out over the isolation region, the cell height drops. N+3 integrates COAG while N6 does not. Our N+3 gate-cut cross-sections indicate COAG, with the gate contact sitting over the active region, while N6 shows an off-gate contact. Last is single diffusion break (SDB). Diffusion breaks are inserted between cells in the same row to provide electrical isolation, but they also introduce local layout effects (LLE), layout-dependent shifts in electrical characteristics. In the past, a double diffusion break was used, consuming the space of two CGPs. SMIC N+3 and TSMC N6 instead use SDB, saving area but increasing LLE sensitivity. This must be controlled at the process level and accurately modeled in the process design kit (PDK) so that EDA tools can account for it. Overall, SMIC N+3 has a transistor density of 113.4 MTr/mm², slightly above TSMC N6 at 107.7 MTr/mm². Even without EUV, SMIC has achieved density beyond TSMC’s mature N6 node which utilizes EUV. Metal Stack The smallest critical dimension in the teardown is M0; SMIC N+3 uses a 32.5 nm local metal pitch. That is smaller than the 36 nm M0 pitch on Intel 18A in Panther Lake. However, this does not mean that SMIC has a better process than Intel 18A or TSMC N3P. M0 is a local intra-cell routing layer. Its usefulness depends on the full interconnect stack: M1 and M2 pitch, track count, via and line resistance, design rules, mask count, overlay control, and routing flexibility. The 32.5 nm M0 is consistent with self-aligned quadruple patterning (SAQP), whose four-population line-width loading we read coarsely as alternating widths of 21.5 to 24 nm; M1 and M2, at 38 and 40 nm, are consistent with self-aligned double patterning (SADP), a single A/B split. On TSMC N6, M0, M2, and M3 sit at a relaxed ~40 nm and are consistent with SADP-class double patterning, with no need for quadruple patterning. That said, we measure M2 for example at ~43 nm, likely inflated by sparse routing. We do not assign any specific layer to EUV from our cross-sections; the distinction we can draw is double versus quadruple patterning, not lithography wavelength. Transistor-level density in the front-end-of-line (FEOL) sets an upper bound, but the design is ultimately limited by what the interconnect stack can route. The lowest metals are the most important for standard-cell density, but the semi-global and global layers determine how usable that density is at the block and chip level. Two axes are commonly used for chip cross-sections: the fin-cut and the gate-cut. The micrograph above is a fin-cut and shows metals 0 through 3. This axis lets us see the even-numbered metals, with M0 right above the fins. There are two kinds of M0 lines. The first are the power rails; these are wide wires for the VDD and VSS running horizontally at the top and bottom edges of each standard cell. The wide wires measure 55 nm across, more than double the other M0 lines. Their width minimizes resistance and reduces IR drop. The second kind are intra-cell wires, short segments within the cell that connect terminals to M1. These have alternating widths between 21.5 and 24 nm. The M0 pitch is 32.5 nm, a 19% reduction versus N+2 and N6. At this pitch, DUV patterning requires more aggressive multi-patterning, increasing mask count, overlay sensitivity, process complexity and cost. M0 is below what a single DUV-defined spacer (SADP) can resolve, so SMIC cascades a second spacer step (SAQP). The cross-section reflects the cost: the M0 trenches are visibly more re-entrant (narrower at the bottom than the top) than M1 or M2 on the same chip and carry a bright barrier-rich foot where the trench meets the etch-stop layer. That shape is partly the intended damascene profile, as a slightly narrow bottom helps void-free copper fill, but its magnitude at M0 is driven by the tight pitch and the higher trench aspect ratio. Intel 18A supports an M0 pitch of 32 nm, although Panther Lake has only shipped with a looser 36 nm pitch. This is due to Intel’s heavy usage of HP libraries. Among leading-edge nodes, 18A has the loosest M0 pitches due to PowerVia. With power routing moving to the backside, congestion is reduced, and the entire front-side metal stack can be used for signal routing. M2 is the first true inter-cell routing layer. It runs horizontally like M0 but spans across multiple cells to carry block-level signals. The M2 pitch sets the cell’s track height – the number of M2 tracks that fit between the VDD and VSS rails, defining what the library calls a 6-track or 7.5-track cell. This layer is the most important, limiting the routing of entire blocks. SMIC N+3 features a 5.7-track cell. The M2 pitch is 40 nm, a 5% decrease over N+2 and the same as N6. This shrink keeps the pitch at the edge of what is possible with double patterning. Future nodes will need to increase the number of masks for M2 as reducing the number of tracks is much harder due to the limitations in routing. The micrograph above is in the perpendicular direction, the gate-cut, and shows metals 0 through 4. This allows us to see and measure the odd-numbered vertical metal layers. The M1 pitch is 38 nm, 9.5% less than N+2 and 33% less than N6. The M1-to-gate ratio matters because it sets local routing flexibility. N+2 and N+3 use a 3:2 ratio, while N6 uses a 1:1 ratio, explaining the huge differences in M1 pitch. The more M1 lines there are compared with the gates, the more flexibility there is for power and signal crossing within the cell. Routing flexibility enables more complex and better cells. Clean fractional ratios are also preferred as a grid is periodic and improves layouts. The 3:2 ratio gives SMIC more local routing flexibility than a strict 1:1 grid, but it also complicates layout and patterning. This is a DTCO choice, with SMIC increasing process complexity to recover density and routability without EUV. This 3:2 ratio is not very popular in the leading-edge nodes. TSMC has only used it on N7+, the N5 family, and the short-lived N3(B). They have switched back to a 1:1 ratio for N3E. Intel has only used it on the 10 nm/Intel 7 family, with Intel 4, 3 and 18A all using a 1:1 ratio. Samsung is the only one still using a 3:2 ratio at the leading edge, using it in the SF4 and SF3 families. It remains to be seen if SMIC will remain at a 3:2 ratio or move to a 1:1 ratio with its future nodes. The industry is still actively exploring these local-routing ratios. At VLSI 2026, imec will be presenting work on even higher ratios, including a 2:1 scheme that can reduce area by up to 14%. We will be covering the conference in a future newsletter article. Subscribe We’re diving deep into the most advanced datacenter and AI hardware hitting the market. To learn more about what’s in the pipeline or to commission a custom teardown, contact sales@semianalysis.com. Interested in joining us on this ride and think you can be a difference maker? Check out our Careers page. The final local interconnect layer for N+3 is M3, with a pitch of 44 nm. The M3 pitch is the same as on N+2 and 10% larger than on N6. The semi-global layers carry the majority of block-level signal routing. They have a coarser pitch than the lower local layers. On leading-edge nodes, they are designed to sit at the limit of DUV single patterning. M4 through M11 pitches were found divided between 80–82 nm (M4–M6), 128 nm (M7–M10), and 148 nm (M11). Given limited sampling, it is possible these are divided further in dense routing areas. At the top are two giant metal layers, M12 and M13. These have kept the same pitches as N+2 at 1920 nm and 4600 nm respectively. While the lower layers’ pitches are generally fixed by the process and library, the upper layers vary much more in pitch and count, depending on the design. Even two smartphone SoCs on the same process can have wildly different metal stacks. The Helio G99 carries fewer routing layers, reaching coarse metal pitches of 850 nm by M9, while the larger and higher-performance Kirin 9030 keeps fine pitches until M11. SRAM At the leading edge, SRAM is much more difficult to scale than logic. TSMC’s latest nodes have seen little to no bitcell scaling, while logic still has more DTCO levers to pull. While looking for other logic libraries in the GPU compute units, we stumbled upon the SRAM. The most common type of SRAM has 6 transistors (6T), but this cell had 8 transistors (8T) instead. 8T SRAM adds two transistors to form a dedicated read port. Unlike a 6T cell, where reading disturbs the storage, the decoupled read port removes read-disturb, improving read stability and letting the cell be pushed harder for performance. At first glance, the cut looked like an unusual logic library, with each cell row having 3 fins of one polarity and 5 fins of another. The rows also alternated in orientation. Energy-dispersive X-ray spectroscopy (EDS) resolved our confusion. The cut had not landed on the GPU logic, but on the SRAM macro beside it. The unusual fin pattern was due to the SRAM library. We return to EDS in the process flow analysis behind the paywall. SRAM libraries are not like traditional logic libraries. Due to the unequal number of PMOS and NMOS transistors, they require specialized rules and layout libraries. They do not need the flexibility of logic libraries, so they are hyper-optimized for one purpose: dense, reliable memory. The SRAM cell we found is a 1:2:2-2:2 cell. This means there is 1 fin per pull-up (PU) PMOS transistor, and 2 fins per pull-down (PD) and pass-gate (PG) NMOS transistor. These 2 PU, 2 PD and 2 PG transistors would usually form a single 6T high-current cell (HCC). An 8T HCC adds a read-pull-down (RPD) and a read-pass-gate (RPG) NMOS transistor, each with two fins. We measured a cell height of 406 nm, which brings the bitcell size to 0.0463 µm². That is a theoretical peak density of 21.6 Mib/mm². We estimate that a 6T HCC would have a cell height of 292 nm and a size of 0.0337 µm². This is ~12% larger than a 6T HCC on Intel 3 and 4. We also estimate the 6T high-density cell (HDC) to have a cell height of 228 nm and a size of 0.0260 µm². This is coincidentally the same as the logic standard-cell height measured earlier. The estimate puts the cell near Samsung 7LPP/5LPP and slightly below TSMC N7/N6. That is a theoretical peak density of 38.5 Mib/mm². 6T HDC is arguably the most important cell as it is used for the largest caches in a chip, the L3 caches and system-level cache (SLC). Both the Kirin 9020 and 9030 have split the SLC into 4 banks to raise total SLC bandwidth. In the Kirin 9030, the SLC increased from 2 MiB to 3 MiB per bank. Correspondingly, the number of arrays within the bank also increased by 50%, from 16 to 24. Each array can store 128 KiB and forms an orderly pattern on the die shot. From the Kirin 9020 to the Kirin 9030, the area of a 128 KiB SLC array decreased from 0.0477 mm² to 0.0392 mm², an 18% shrink. The achieved density is 25.5 Mib/mm², 66% of the theoretical maximum. While the SLC was quite similar across both chips, the L3 has seen some major changes, particularly in terms of its layout. The total capacity also went up from 10 MiB to 12 MiB. Much like the SLC, the L3 is also split into 4 banks. In the Kirin 9020, an L3 bank consisted of 16× 128 KiB arrays and 16× 32 KiB arrays. However, an L3 bank in the Kirin 9030 instead consists of 48× 64 KiB arrays. In the Kirin 9020 L3, a 128 KiB array was 0.0513 mm² and a 32 KiB array was 0.0154 mm². The size of the 128 KiB array is different on the L3 and SLC as the assist circuitry for the two arrays differs depending on their purpose. In the Kirin 9030 L3, a 64 KiB array is 0.0210 mm². Although not a like-for-like comparison, normalized for capacity, it is 18% smaller than the 9020’s 128 KiB L3 array and 31% smaller than its 32 KiB L3 array. The achieved density is slightly lower than the SLC, at 23.8 Mib/mm², 62% of the theoretical maximum. Unlike the L3 and SLC, the prime cores’ private L2 cache uses a 2-bank design. As the prime cores’ L2 is latency-critical, it likely uses 6T HCC instead of 6T HDC. The 9020 has 16 arrays in each bank while the 9030 has 32. Each array has a capacity of 32 KiB. A 32 KiB array in the L2 shrank from 0.0171 mm² to 0.0142 mm², ~17% smaller. The density is 17.6 Mib/mm², ~59% of the theoretical maximum for 6T HCC. SRAM scaled well from N+2 to N+3, shrinking by ~19%, close to the theoretical logic shrink. The caveat is that N+2’s bitcells were unusually large, bigger than comparable 7 nm-class nodes, so part of the gain is catch-up rather than true scaling. With the insights from STEEL’s teardowns, we will be doing a deep dive into SRAM in a future newsletter article. Subscribe Everything above came out of a single STEEL teardown: die annotation, block-level area analysis and TEM cross-sections through logic and SRAM. We’re diving deep into the most advanced datacenter and AI hardware hitting the market. To learn more about what’s in the pipeline or to commission a custom teardown, contact sales@semianalysis.com. Future Roadmap The same cross-sections that pin down N+3 also show where SMIC can go next. Although N+3 is already close to the practical limits of DUV multi-patterning in several layers, SMIC still has a few scaling levers left. A theoretical N+4 would likely start with cell height. N+3 uses 5 M0 tracks between its power rails. Moving to 4 M0 tracks, as on SMIC N+2 and TSMC N6, could reduce cell height by roughly 15%. The routing grid is only one side of the shrink; the front end also must fit into the smaller cell. One possible FEOL lever is reducing the p-to-n isolation spacing from two diffusion grid units to one. Intel used this scaling booster on Intel 4, and TSMC did so on its N3 family. This path trades layout flexibility for density. Fewer M0 tracks reduce local routing resources, while tighter p-to-n spacing raises integration and design-rule difficulty. M2 is also constrained by the cell height shrink. For SMIC to maintain a ~5.7-track cell, M2 would need to move toward ~35 nm. That would move another layer into SAQP territory. SMIC could also reduce the CGP from 57 nm to 54 nm. Intel reached a similar CGP on Intel 10 nm/Intel 7 without EUV. The local interconnect is also tougher. If SMIC keeps the 3:2 M1-to-gate ratio, M1 would need to shrink to 36 nm and would likely require SAQP as well. If SMIC moves to a 1:1 ratio, M1 could relax to 54 nm, but it would give up routing flexibility. Under this theoretical path, we estimate that SMIC N+4 could reach a cell height of 198 nm and a CGP of 54 nm, implying a Bohr density of 137.8 MTr/mm², on par with TSMC N5 or Samsung SF4. However, the difficulty is cumulative. Each step is individually plausible, but together they make N+4 harder than the transition from N+2 to N+3. It will likely take longer, cost more, and carry less process margin. A theoretical N+5 would require a larger integration shift. One possible path is backside contacts (BSCon), moving power routing and source/drain contacts to the backside, which would reduce front-side routing pressure and enable another reduction in cell height. The front-side metal pitches could relax to reduce the process complexity. M0 would likely relax slightly to ~34 nm, and M2 and M4 pitches could relax further. CGP is unlikely to shrink much further. Even with EUV, 48 nm has been the practical limit for yield and process control. This approach would allow N+5’s cell height to fall to 170 nm and its CGP to 53 nm. This implies a Bohr density of 163.6 MTr/mm², on par with Intel 18A’s HP library. However, this would not make N+5 cost-competitive with the leading edge. It would reach a similar density through a much more expensive route. The integration difficulty rises sharply, with new process flows for backside alignment, wafer thinning, contact reveal, and backside metallization. Past this point, standard density and interconnect scaling become increasingly unattractive. That is where Huawei’s roadmap stops looking like a normal foundry roadmap and starts looking like a packaging roadmap. At ISCAS 2026, Huawei unveiled its tau (τ) scaling law, reframing process scaling in the time domain. τ is the time cost of data movement and processing: switching delays in transistors, RC signal propagation delays in circuits, compute, memory, and networking latency. Outside Huawei’s terminology, this is called system-technology co-optimization. This is Huawei’s answer to its lack of EUV lithography. Without EUV, planar density cannot keep pace with TSMC, Intel, or Samsung. If transistor density cannot shrink further, Huawei’s alternative is to shorten wires, reduce buffering, and stack logic vertically. “LogicFolding”, Huawei’s implementation of this new scaling idea, is, in practice, an aggressive 3D stacking approach. AMD V-Cache places SRAM above or below a CPU die. AMD’s MI350X places active interposer dies (AIDs) underneath accelerator and compute dies (XCDs), with the AIDs handling cache, IO interfaces, the network-on-chip (NoC) and embedded metal-insulator-metal (MIM) capacitors. With LogicFolding, parts of the same logic block are split across multiple active dies bonded face-to-face at ultra-fine pitch. This allows Huawei to shorten some critical paths and reduce buffer overhead, not merely add cache capacity or offload the IO and interconnect. Shortening wires is where the higher clocks come from. A large share of a modern core’s delay and energy budget goes into driving long interconnects and the repeater buffers along them. LogicFolding distributes a block’s critical-path gates across multiple stacked tiers bonded at a very fine pitch, so the bond interface behaves like an additional metal layer and the longest paths get shorter. That is how Huawei expects to recover frequency and efficiency it cannot get from the process alone. Huawei’s roadmap shows its intent. Prime core frequency is targeted to rise from 2.75 GHz in the Kirin 9030 to roughly 5 GHz by 2031, far beyond what planar scaling alone could deliver. Prime cores with 3.1 and 3.39 GHz clocks are being tested in its labs, although their power consumption is unknown. Beyond that, chips are in the design, simulation or pathfinding phase, meaning the frequencies are targets. However, the direction matters more: LogicFolding also helps with performance, not just density. The catch is that Huawei’s density claim is not directly comparable to foundry densities. A stacked design can report more transistors per package footprint by adding active layers, even if each patterned die remains well behind TSMC or Intel in front-end density. This is how Huawei can claim to reach foundry 14A-equivalent density by 2031. This is not a like-for-like foundry comparison, with Huawei using stacked logic and measuring density per package footprint. On a normalized Bohr-density basis, SMIC N+3 is ~114 MTr/mm², 38% less than Intel 18A’s HD library. Huawei’s 3D roadmap closes the gap by stacking active logic, reaching 215 MTr/mm² by 2030. In 2031, the roadmap density jumps to 295 MTr/mm², implying either a third active layer, partial EUV insertion or aggressive planar DUV scaling. Huawei’s methodology makes other foundries look much denser as well. Applying it to AMD’s MI450X with an N2 top die on an N3P base die yields a theoretical density of 460.2 MTr/mm² in 2026, compared with Huawei’s 295 MTr/mm² in 2031. This Kirin 9030 does not use LogicFolding, remaining in a conventional mobile SoC package. Instead, it forms the baseline for how far Huawei and SMIC can push planar scaling. Future teardowns of Kirin and Ascend chips will show both planar logic density and Huawei’s hybrid bonding solutions. Export controls changed China’s optimization problem rather than ending it. EUV restrictions raised the cost and complexity of leading-edge manufacturing without freezing it. SMIC reaches N6-class logic density through DUV immersion, SAQP and DTCO, while Huawei shifts more of the burden onto architecture, packaging and system-level integration. Future nodes will be tougher. N+3 still had room to tighten local metals and reduce cell height and CGP. Further scaling without EUV leaves fewer levers. More aggressive multi-patterning adds masks and overlay error. SMIC can keep pushing DUV, but each step will get more expensive and less forgiving. The design side is just as critical. Huawei had domestic EDA tools and flows before the Kirin 9030, with the Kirin 9000s, 9010 and 9020 making that clear. Huawei was able to ship multiple consumer SoCs on SMIC N+2 and N+3 while cut off from the Western EDA stack. US export controls restricted EDA tools for advanced chips in 2022 but did not target tools for more mature chips. In 2025, the US government briefly placed much broader restrictions on EDA software from Synopsys, Cadence, and others, before lifting them less than two months later as part of a trade deal tied to rare earths. Huawei has been unable to access those tools because it remains on a US trade blacklist. That forced Huawei, SMIC and Chinese academic institutions to build their own tools and flows. Researchers at Peking University recently announced a prototype EDA tool for Huawei’s LogicFolding architecture, which requires a new flow to handle the multilayer layout and floorplan. This is not the same as replacing the full Synopsys or Cadence stack, but it shows where domestic EDA is headed: toward tighter co-optimization between architecture, process and packaging. These advances are also diffusing into the Chinese ecosystem. SMIC is licensing its N+2 and N+3 processes to HLMC/Hua Hong at the government’s direction rather than by choice. If the same process learning feeds into Ascend accelerators for AI training and inference, the choke point shifts from one named fab to an ecosystem. Alibaba’s T-Head silicon arm and Cambricon, a Chinese AI chip designer that is expected to supply ByteDance, could also be major beneficiaries. Sanctions aimed at SMIC alone become less effective once the manufacturing knowledge has spread to other fabs and design houses. China is not closing the gap with Intel, Samsung and TSMC. The teardown shows the opposite in several places: no EUV, no backside power, higher process complexity, and visible trade-offs. But China is still advancing. If domestic chips become good enough for phones, inference, networking and security-sensitive workloads, they can matter strategically without matching TSMC at the leading edge. Behind the paywall, we show what else STEEL can do, with material and process flow analysis of SMIC N+3, and analysis of the Kirin 9030 package. We’re diving deep into the most advanced datacenter and AI hardware hitting the market. To learn more about what’s in the pipeline, access the full Kirin 9030 and SMIC N+3 analysis or to commission a custom teardown, contact sales@semianalysis.com.
How documenting everything changes your brain
Psychologists call it cognitive offloading, and it has upsides and downsides for your memory.
Why You Need to Prepare Your iPhone Before Downloading iOS 27
Prepare for the iOS 27 beta release at WWDC 2026. Learn about new AI features, Siri 2.0, and which older iPhone models will lose support.
Filtr is a new privacy tool that blocks ads in almost every iPhone and Mac app
This popular ad blocker app for iPhones, iPads, and Macs can now block ads from loading inside apps, including web browsers, thanks to a new feature in the latest Apple software.
Myles Garrett traded from Browns to Rams as Los Angeles goes all in for Super Bowl run
Los Angeles is loading up for a Super Bowl run by acquiring the reigning Defensive Player of the Year in a historic blockbuster trade.
Ebola, chemical plants and health, hantavirus, common colds, heat and more
As they say, bugs have ears. Public health has had a relentless May. Chemical plant (near) explosions, Ebola, hantavirus, and that's before you even get to the usual suspects: common colds, heat, and ticks. Here’s an attempt to keep you up to speed and, more importantly, what it means for you and your community. But first, an announcement: World Cup The World Cup starts in 15 days (not like anyone is counting down), and we are hanging TVs in the Health Security Operations Center. It’s going to be a gorgeous, chaotic celebration of basically every culture on earth. And it’s going to take a team effort—from communities to individuals, from public health to health care—to keep people healthy. That’s where you come in. If you’re going to the games or live within 30 miles of a stadium, we would love to hear from you. Sign up for a weekly survey here. Thanks to the more than 1,280 people who have already signed up…you will be hearing from us soon!! Global disease “weather report” Hantavirus: Two more cases The global count of cruise ship passengers with hantavirus has now increased to 13 cases (11 confirmed), with three deaths. The two new cases are overseas and among people who were already quarantining: A crew member in the Netherlands. A Spanish national passenger who was quarantining at home tested positive through daily monitoring. They are now in a biocontainment unit. In the U.S., everyone remains negative. There may be more cases, given that the incubation period of hantavirus (the time from exposure to infection) is 45 days. But this week, we will reach a major milestone: the median incubation window is 18 days, and that will pass on May 29. What this means for you: Your risk from this cruise ship outbreak remains essentially nil at this point. Ebola rages on in Central Africa Big thanks to Dr. Craig Spencer, a humanitarian physician who has treated and survived Ebola, for jumping in to provide the YLE community with an update. Craig, take it from here… The combined confirmed and suspected Ebola cases in DRC are now more than 1,000. All signs are pointing to a very long and catastrophic outbreak in Central Africa: This is a vast undercount. We know this because the test positivity rate is hovering around 50%, only 20% of contacts are being traced (and in some areas, no contacts at all), and more cases keep popping up with no known connection. This all points to widespread and undetected community transmission. This is in only a week of detection. Compared to previous outbreaks, the growth is very fast, as the huge West Africa outbreak in 2016 was first detected at 49 cases and rose to 208 cases a month later. It took four months for that outbreak to reach the size of the current one in the DR Congo. The cases are spread out across 16 health zones. There are now multiple epicenters, making containment very difficult. Next door in Uganda, the case count is seven. While this number is low compared to DRC, a concerning development is that two health care workers recently tested positive with uncertain exposure histories. If they weren’t treating known Ebola cases, this means it’s spreading undetected in Uganda as well. On the ground, backlash, including the burning of health centers, has emerged, a pattern seen in nearly every outbreak and rooted in deep community distrust. It often stems from outside actors working in communities without fully understanding or addressing local priorities. Affected populations may recognize the severity of Ebola while still holding other concerns as more pressing, such as where their loved ones are buried. Community trust is essential to an effective response, but difficult to build during an active emergency. It is best established long before a crisis begins. U.S. priorities are made clear, and may backfire. The U.S. Administration has shown that its first priority is keeping Ebola out, with helping end the outbreak in the DRC a secondary concern. That's meant travel restrictions broader than anything we've imposed before — covering travelers from across the region, and reportedly some green card holders and permanent residents as well. For example, just these past two days, news broke that high-risk American travelers will be subject to a mandatory quarantine in Kenya before they are allowed to return. If an American is infected, the U.S. government plans to send them to a hospital it is standing up from scratch in Kenya. (Currently, Americans who were in the area are allowed to return but are diverted to three airports—Houston, Atlanta, and DC—for further screening.) Past administrations have used travel notices and stepped-up screening; this goes much further. At first blush, these strict precautions may sound prudent. But this approach could backfire in three ways: Restrictions this blunt give people every reason to hide where they’ve been and whom they’ve been near — making the people we most need to find harder to track, not easier. They also breed a false sense that this is someone else’s problem. But diseases are humbling. They find the small cracks in even the most impenetrable-seeming defenses. Lives lost. There is no treatment for this Ebola strain, which means survival depends heavily on the quality of the health system. We have that system in the States, but we are choosing not to use it for infected Americans. This is unbelievable and infuriating. (See a deeper dive from me here.) The only real way to lower the risk to Americans — and everyone else — is to end the outbreak in the DRC and across the region. What this means for you: If you have travel plans to this region, it’s time to cancel them. This is a high-risk situation in Central Africa, and CDC released a Level 4 Travel Advisory. There is also great uncertainty if and when the Administration will let you back in. To the general public in the U.S., your risk remains very, very low right now. U.S. disease “weather report” Ticks: past peak season? Good news, especially for people in the Northeast and Midwest: tick numbers continue to decline. Though we are at the peak of the season, this unusually early year is trending favorably. Heat-related illnesses in the North Heat risk is certainly not blanketing the entire country yet, but this week it will be moderate in the Midwest and the Southeast. What this means for you: Be sure to check the CDC HeatRisk tool. A red day isn’t the best day for that soccer game for a kid with asthma, and a stretch of orange days is a great time to check on your elderly neighbor. Common colds surging As far as respiratory viruses go, the only thing really going around right now is the common cold. It is higher this year than last year, but should peak very soon before returning this fall. Spotlight: Chemical plants Over the weekend, YLE California covered the serious situation that unfolded around a chemical plant in Orange County, California. Thankfully, the worst-case scenario (a chemical explosion) was mitigated, but only through luck. The health implications in a densely populated area could have been catastrophic. As the situation was winding down, another chemical explosion struck Washington State. This one was fatal. Health officials say the risk to surrounding residents from chemicals in the air and ground remains low, though the community deserves far greater clarity. Clean air and clean water consistently rank as Americans’ top public health concerns — yet when chemical disasters strike, communities are too often met with reassurances rather than transparency and accountability. Residents near the 2023 East Palestine, Ohio train derailment, for example, are still waiting to understand the long-term impact on their health, with few clear answers. What this means for you: Many are wondering if facilities processing potentially dangerous materials are in their neighborhoods. This interactive map allows you to enter your ZIP code and see all facilities monitored by the EPA, including any violations. Good news Ebola vaccines and treatment for the Bundibugyo species are in development. Two vaccine platforms are being explored, at least one of which may be deployable in as little as 2 months. One vaccine is being developed by Oxford, and the other uses the same biotechnology as the FDA-approved Zaire species of Ebola. (It may provide cross-protection, but it’s unclear at this point.) On the treatment side, two promising monoclonal antibody cocktails are likely to be deployed, though doses are limited and logistical hurdles remain significant challenges. Anni over at The Biotech Tea had a great explainer about the Ebola vaccine gap, if you want to read more. [ The Biotech Tea The Ebola vaccine gap, reading vessel damage, & the Makary exit 🗝️ Paid subscribers get the Biotech Term of the Week at $30/year (~$2.50/month). I try to keep it intentionally accessible, but if cost is ever a barrier, send me a message. A few people asked about expensing your subscription, so I put together a short email template… Read more 6 days ago · 10 likes · Annicka Evans, PhD ](https://www.thebiotechtea.com/p/the-ebola-vaccine-gap-reading-vessel?utm_source=substack&utm_campaign=post_embed&utm_medium=web) Polycystic ovary syndrome (PCOS) has been officially renamed polyendocrine metabolic ovarian syndrome (PMOS), following more than a decade of debate and input from roughly 22,000 clinicians, researchers, and patients worldwide. This is good because the old name was medically misleading, leading to patients without visible cysts being dismissed or overlooked entirely. The WHO estimates that 70% of people with the condition remain undiagnosed. A more accurate name should improve recognition, reduce stigma, and ultimately help the millions of people living with this condition get diagnosed and treated sooner. Poll Loading... Bottom line Public health situations are everywhere right now. Staying healthy takes public health professionals and systems working tirelessly behind the scenes and each of us showing up for our neighbors and communities. Love, YLE Big thanks to Ed Nirenberg for staying on top of the Ebola vaccines, Hannah Totte for all the figures, Dr. Craig Spencer for the Ebola insight, and Dr. Matt Willis for covering the chemical situation in CA. Your Local Epidemiologist (YLE) comprises a team of experts, ranging from physicians to immunologists to epidemiologists to nutritionists, working together with one goal: to “Translate” ever-evolving public health science so that people are well-equipped to make evidence-based decisions. YLE suite of newsletters reaches over 475,000 people across more than 132 countries. This newsletter is free to everyone, thanks to the generous support of fellow YLE community members. To support the effort, subscribe or upgrade below: Subscribe
Kash Patel’s personal merch site hacked to trick users into installing malware
FBI Director Kash Patel’s personal merchandise site went offline after a hack apparently tricked visitors into downloading malware.
Live and Die by the Fastball – Marlins 3, Rays 6
On a national telecast through Peacock, the Tampa Bay Rays got a well deserved spotlight this Sunday, starting with the star of the show – Junior Caminero – clearing the fence: After giving up the lead, the Rays would battle back the second time through the order, and after loading the bases, Taylor Walls – […]
How to Prevent Chrome Browser from Downloading 4GB of Local AI Model Files
If you’re a Mac user of the Chrome web browser, as many are, you might be interested to know that the latest versions of Chrome default to downloading a large local Gemini AI model that can t…
Microsoft Edge to stop loading cleartext passwords in memory on startup
Microsoft is updating the Edge web browser to ensure it no longer loads saved passwords into process memory in clear text at startup.
Windows 11 KB5089549 can now be downloaded for USB installs using official Microsoft tool
Microsoft has made an update related to the Media Creation Tool, its official utility for downloading and installing Windows via ISOs.
Stop Chrome Browser From Downloading a Hidden 4GB AI File
If your Mac's storage has been mysteriously shrinking recently and you use Google Chrome, you may have already identified the culprit. The browser has been downloading a 4GB AI model file onto computers without explicit user consent. Here's how to reclaim the space.
BMO to Sell Truck, Trailer Financing Businesses to Stonepeak
Bank of Montreal agreed to sell its business that finances trucks, trailers and equipment to Stonepeak, offloading a portfolio that totals about C$14.5 billion ($10.6 billion).
Google Chrome Is Downloading a 4GB AI Model Onto Your Device Without Consent, Researcher Warns
Maybe Google doesn't ask permission because it knows what the answer would be.
Google Chrome Is Silently Downloading a 4GB Gemini Nano AI Model to User Devices Without Consent
Google Chrome has been automatically downloading a 4GB folder of Gemini Nano AI weights to user devices without notification, and redownloads it if deleted.
Missouri tests medical drones to speed up rural care
A medical drone program in Missouri aims to speed up delivery of blood and lab samples, as rural patients face growing challenges accessing timely care.
Hackers exploit file upload bug in Breeze Cache WordPress plugin
Hackers are actively exploiting a critical vulnerability in the Breeze Cache plugin for WordPress that allows uploading arbitrary files on the server without authentication.
Amazon’s new Fire TV Stick HD signals the end of sideloading
Amazon is weirdly warning only some customers that they will not be able to sideload apps on the new Fire TV Stick HD.
Amazon won’t release Fire Sticks that support sideloading anymore
The two newest Fire Sticks block apps from outside of Amazon's store.
USC women’s basketball officially signs five-star Sara Okeke
The Women of Troy are loading up on the recruiting trail, and have now officially signed three five-stars in the 2026 cycle
Dangerous Microsoft Windows Update Confirmed—Do Not Download
As security researchers warn about a dangerous Microsoft Windows update that isn’t, users must pay close attention to what they are actually downloading.
Dangerous Microsoft Windows Update Warning—Do Not Download
As security researchers warn about a dangerous Microsoft Windows update that isn’t, users must pay close attention to what they are actually downloading.
New Google Photos trick lets you save time while sharing images
Google Photos now supports a new feature that lets you share images more quickly, without downloading the originial first.
Android Developer Verification Rollout Begins Ahead of September Enforcement
Google expands Android developer verification globally after September rollout, adding authentication and delays to sideloading to deter malware.
Android's new developer verification rollout begins, sideloading changes are next
Google says the new advanced sideloading flow will launch in August, followed by a wider global rollout.
New ‘Android Developer Verifier’ app coming to phones as Google shares verification timeline
After detailing the “advanced flow” for side loading earlier this month, Google today is opening Android developer verification to developers...
Is it true that … you can never eat too much fibre?
Fibremaxxing is everywhere, and most of us should eat more roughage, but your gut won’t thank you for overloading
‘Crimson Desert’ Patch Notes: New Mounts, Glide Stamina, Faster Loading And More
Pearl Abyss just released a new patch for Crimson Desert, continuing its tradition of rapid-fire, quite substantive patches that keep hitting the game. Here's what's new
Your Thymus and Your Healthspan
Our thymus gland plays a central role in the development or our immune system, specifically for supporting T cell development (how these cells got their name, maturation and differentiation in the thymus) and discriminating between self and foreign, non-self antigen proteins with production of dendritic cells. As we age, our thymus gland shrinks—the process known as involution—with progressive change from spongy to fatty tissue, with loss of functionality. But this process, with respect to timeline and extent, markedly varies from one person to the next. And to make things even more complicated, our thymus gland anatomy and precise location in the chest also is highly variable. So the famous Frank Netter diagrams from the 1960s (such as at left below) don’t capture the remarkable heterogeneity that a random set of 12 chest CT scans (below at right). What I consider as 2 landmark papers in Nature this week (here and here) used AI to quantify health of the thymus in 2 large cohorts, and then correlate the metric to a broad array of health outcomes. In this edition of Ground Truths, I’m going to cover three questions: (1) How was thymus health determined by AI?; (2) How is thymus health linked to key clinical outcomes? and (3) Are there ways we can promote a healthy (aka rejuvenate) thymus gland in our later years? Subscribe 1. How was thymus health determined by AI? This was accomplished by extensive work developing and validating an AI pipeline for 2 stages, the first for localization and segmentation of the thymus bed, and the second to quantify a digital marker, termed the thymic health score. There’s a lot to this algorithmic development, so I won’t go into all the details, but just provide a rudimentary outline of what was done. The 1st stage used supervised learning with 2 radiologists reviewing 2,461 chest CT scans. That work trained a 3D U-Net model to automatically identify the thymic bed’s 3D cropped region using center of mass coordinates, which reached 99.8% accuracy. The 2nd stage used a foundation model that was pre-trained with SwAV (which stands for Swapping Assignment between multiple Views), a self-supervised model, that generated a high-dimensional thymus representation of 4,096 features. That compares to the radiology reductionist and subjective scoring of 0 to 3 where 0 means the thymus is fully degenerated and fatty, and 3 is considered dense, intact glandular soft tissue. Notably, the self-supervised learning (SSL) performance was superior to supervised (AUC ~0.75 vs 0.55, respectively) reflecting the shallowness of the 0-3 classifier vs SSL’s holistic, self-taught AI. The thymic health score ranged from 0 to 100, with the highest number indexed to fully preserved thymic health. The algorithmic work extended to explainability with both Shapely value distribution, meaning the model was holistic not relying on any “magic” pixel, and occlusion sensitivity, demonstrating the model’s performance was not affected by ribs, sternum, lung tissue, but rather focused on the thymic bed directly. Below are a few saliency maps from the occlusion sensitivity that indicate the thymic bed focus, not affected by neighboring structures, using the jet color scale. Now the AI was ready for processing ~25,000 CT scans, each over 30 MB files, from the 2 large participant cohorts to provide a thymic health score, it did so in less than 14 hours, which is less than 2 seconds per scan! Share 2. How is thymus health linked to key clinical outcomes? The two cohorts were the National Lung Screening Trial, NLST (N=25,031) and the Framingham Heart Study, FHS (N=2,581) each with baseline demographics and long term follow up of >12 years for health outcomes. From the NSLT, you can see below that increased age and body-mass index were correlated to reduced thymic health. And, overall, men had reduced thymic health scores compared with women. There are a lot of graphs in the paper for outcomes, for both the NLST and FHS cohorts. To simplify the major outcomes, below is a composite Figure with striking reduction of all-cause mortality and for the different causes of death. The results were consistent between the 2 cohorts. For example, the cardiovascular mortality hazard ratio average was 0.57 in NLST and 0.38 in FHS (weighted to be 0.53 for the aggregate). The reduction of mortality also extended to digestive diseases and pulmonary disease (data not shown below). Lung cancer incidence was reduced by 36% for high vs low thymic health (Figure below, with a panel -incidence and b panel-mortality, adjusted for age, sex, BMI, smoking status). Notably, there was a smoker’s paradox: lower incidence and improved survival for lung cancer in smokers who had a high thymus health score. Smoking was associated with lower thymic scores whereas alcohol was not. Higher thymic health was linked to higher HDL cholesterol, lower triglycerides, lower fasting blood glucose, and lower systolic and diastolic blood pressure. Inflammation biomarkers, such as C-reactive protein and interleukin-6, were elevated in the people with lower thymic health scores. In their second paper, the same team looked at the relationship between thymic score and outcomes for 3,476 patients receiving cancer immunotherapy. The non-small cell lung cancer progression-free survival was reduced in patients with low thymic health score. A parallel relationship for survival was seen for other types of cancer including melanoma, breast, and kidney with a 44% lower risk among high thymic scores (Figure). A similar pattern was seen with different immune checkpoint inhibitors. The thymus health score outperformed programmed death ligand-1 (PD-L1)and tumor mutation burden (TMB) assay of the tumor and was an independent prognostic marker of progression-free survival. In this dataset, a direct connect with thymus score and adaptive immune function was noted with a correlation before cancer treatment for both T-cell diversity and thymus T-cell production (a metric known as T cell receptor excision circles). Thymus Removal Study There is a highly relevant citation about thymus function before moving onto ways to rejuvenate it. In 2023, a very important study of thymus removal during cardiothoracic surgery was published, with 1146 patients undergoing thymus removal vs 1146 matched controls (mean age 55 years). The adverse outcomes for thymus removal were striking, with a 2.9 higher risk of all-cause mortality and doubling of cancer incidence, 1.5 fold increase in cancer mortality, and 1.5-fold increase in autoimmune diseases. The immune function showed marked compromise after thymus removal, as reflected by CD4+ and CD8+ T cells (Figure below) Share Ground Truths 3. Are there ways we can promote a healthy (aka rejuvenate) thymus gland in our later years? In recent months we’ve learned a lot about the process of thymus involution and the pathways by which this may be modulated. Thymus involution is primarily due to loss to thymus epithelial cells (TECs) and there are 2 subtypes with different functions: the cortical, responsible for positive selection of T cells, and the medullary, for negative selection. When adipose tissue infiltrates the thymus, thymic adipocytes are pro-inflammatory, knocking out T-cell output. Likewise, systemic inflammation, or “inflammaging,” from smoking, obesity, and chronic stress promote thymic adipocytes. Now, from preclinical studies, we know about key pathways that account for these processes. Historically, back in 2014, FOXN1 was recognized as a single master transcription factor essential for TEC function, with studies in mice with forced FOXN1 upregulation leading to thymus regeneration. That laid the foundation for potential regrowth of quiescent thymus cells. It took awhile to find pathways to accomplish this goal and to even refute its role as the master regulator. The Liver-Thymus Axis Feng Zhang and his team recently showed that injection of an mRNA vaccine to the liver encoding DLL1, IL-7 and FLT3-L in combination, factors known to support thymopoiesis (Figure), successfully enhanced immune function in aged mice. However, the effect was transient. The FGF21 Story Two reports in 2025 addressed the thymus gland’s production of FGF21, a growth factor (fibroblast growth factor 21) and its control of thymus involution. This growth factor is also produced by the liver, but it doesn’t have impact on the thymus in aged rodent models. In contrast, The liver’s hepatocyte growth factor (HGF) has been shown to reverse senescent TEC structural changes. Growth hormone stimulates production of insulin growth factor-1 (IGF-1) in the liver and the thymus. Two small studies in human participants (TRIIM, for Thymus Regeneration, Immunorestoration, and Insulin Mitigation) of 6 and 50 people, respectively, treated with combined human growth hormone, metformin and DHEA, suggested the potential of slowing epigenetic aging and improving thymus mass and function. But the combination of drugs, small sample, and lack of controls make conclusions murky. Increased thymus FGF21 led to increased CD8 T cells in old mice and extended healthspan and improved physical performance. Building on the previous FOXN1 data, ablation of β-klotho, the obligatory co-receptor for FOXN1, accelerated thymic aging. The benefit to FGF21 was confirmed in a companion paper using an FGF21 knock-in mouse model showing thymus enlargement and increased TECs throughout the lifespan. (schematic Figure below). The accompanying editorial for these 2 papers speculated: “The proposed link between age-associated thymus involution and organismal aging is intriguing, and suggests we might seek to achieve systemic rejuvenation by counteracting thymic involution.” Other Factors Lessons from the Axolotl (Mexican Salamander) In an elegant set of experiments of the axolotl (Figure), known for its complete thymus regeneration after total surgical removal, mediators were identified. Surprisingly FOXN1 was dispensable. But midkine, a growth factor, appeared to be the driver, the initiator of thymus cell growth, survival, and repair. Other components that contributed were Postn+ (periostin-expressing mesenchymal niches) and Ccl19, a chemokine protein from the bloodstream. RANK and RANKL RANK (receptor activator of nuclear factor κB its ligand (RANKL) have been considered key regulators of medullary TECs. In a study of both aged mice and human thymus cell cultures, their key role was confirmed, with multiple benefits of resorting TEC function, recruitment of progenitor cells, and T cell development. Subscribe Summing Up This body of work strongly supports the health of the thymus gland as a critical regulator of human healthspan, not just a correlate or link. The new landmark studies are reinforced by the regenerative biology experimental models that have shown, via an array of mediators, that healthspan of aged mice is at least in part dependent on thymus gland function and its critical role for maintaining adaptive immunity. It sure seems that we’d be better off having an immune reservoir of naive T cells instead of a profile in older adults of exhausted, low quality, memory T cells, with a senescent phenotype. It is striking that we ignored the importance of the thymus gland for many decades. A wake-up call was the results of the thymectomy matched control study reviewed here, with the opposite of high thymic health score improved health outcomes. The convergence of AI and healthspan here is notable. The exceptional and laborious work for developing and validating (no less explaining) the AI to quantify thymus health set the foundation for probing the connect the major health outcomes in two cohorts with extensive follow-up. We’ve never had a way to meaningfully quantify the thymus gland before, so this represents important leverage of supervised, self-supervised, and transformer models. The study wouldn’t have been possible without AI. And it begs the question as to whether there are ways we can maintain our thymus at a high health score. But it’s not so simple to rejuvenate our thymus. There are multiple risks including the induction of autoimmunity, increasing the risk of cancer, and inducing a pro-inflammatory state. For example, factors that increase TEC proliferation could compromise the medulllary negative selection process and leak out auto-reactive (self-attacking) T cells. Thymic cells could be induced to proliferate, but we know, for example, that growth hormone and IGF-1 are associated with an increase risk of cancer. If thymus rejuvenation doesn’t clear out the accumulated senescent cells in older adults, then inflammation within the gland could block the production of TECs. Keeping these risks in mind, we now have identified many ways to keep our thymus healthy as we age that undoubtedly will be tested in clinical trials going forward. Ultimately, the benefit-to-risk tradeoffs will be defined. In the meantime, we have desperately needed an immunome. As I’ve stressed in multiple prior Ground Truths, we have no test in the clinic to assess a patient’s immune status! It was very encouraging for me to get an email from Hugo Aerts, the senior author of the landmark papers in the days after they were published: “I also wanted to say that your Substack on “Why We Need an Immunome” was inspiring for this work. We found it a very compelling perspective and greatly enjoyed reading it.” Just having an assessment of a person’s immune status via a chest CT scan, with millions of these performed per year, might help in guiding the right immunotherapy for cancer (e.g. more intensive, combinations for those with low thymic health scores) and help define people who are increased risk for age-related diseases of cancer, cardiovascular, and neurodegenerative. The main theme of Super Agers is that we now have many layers of data (genetics, proteins, biomarkers, organ clocks that can be integrated via multimodal AI) for us to define a person’s specific risk decades before one of these age-related diseases leads to symptoms, enabling prevention. From the new studies this week, we can add thymus health score as yet another way to understand a person’s risk, a window into their adaptive immune system and healthspan. And perhaps someday we will be able to safely rejuvenate the thymus, or maintain its health through one’s life, and use the thymus health score to monitor. NB: I wrote this post (No A.I.) A Quick Poll Loading... ********************************************************************* Thanks to Ground Truths subscribers (now > 200,000) from every US state and 212 countries. Your subscription to these free essays and podcasts makes my work in putting them together worthwhile. Please join! If you found this interesting PLEASE share it! Share Ground Truths Paid subscriptions are voluntary and all proceeds from them go to support Scripps Research. They do allow for posting comments and questions, which I do my best to respond to. Please don’t hesitate to post comments and give me feedback. Let me know topics that you would like to see covered. Leave a comment Many thanks to those who have contributed—they have greatly helped fund our summer internship programs for the past two years. It enabled us to accept and support 47 summer interns in 2025! We aim to accept even more of the several thousand who will apply for summer 2026.
This is Android’s new ‘advanced flow’ for sideloading apps without verification, includes one-day waiting period [Gallery]
Later this year, Google is introducing new developer verification. For power users, Android will offer an option to disable verification...
Google introduces a new way for users to sideload Android apps that still protects against scams
Google is making Android sideloading easier with a new advanced setting that lets users disable app verification after a multi-step security process.
Valve has thrown a curveball at Counter Strike 2 with a massive reloading change, forcing players to shake off old habits
Valve has added a bold change to reloading in Counter Strike 2, which is shaking up old habits for players.
Intel Precompiled Shaders Accelerate Game Load Times By Up To 3x On Arc GPUs, Now Available & Supported In 13 Games
Intel Arc GPUs can finally enable Precompiled Shaders for faster loading times that are supported in 13 games at launch.
Intel enables precompiled shader selivery on Arc B-series and Core Ultra iGPUs in new driver, up to 3x faster loading times
Intel Arc gets Graphics Shader Distribution Service, 13 games supported at launch Today, Intel has rolled out its Graphics Shader Distribution Service in
Texans Reloading for 2026 After Fourth Offensive Line Addition
The No. 1 Nutrient to Prioritize for Better Heart Health, According to a Cardiologist
In TODAY.com's Expert Tip of the Day, a cardiologist shares why loading up on this nutrient can help prevent heart disease and more.
Google calls Android’s new sideloading flow ‘high friction’
Google announced changes to Android app sideloading last year that would require developer verification as a safety measure, but the...
Google confirms 'high-friction' sideloading flow is coming to Android
Google says the added friction is meant to educate users about the risks of sideloading.
Video: Animal Crossing: New Horizons Nintendo Switch 2 Edition vs. Switch comparison
A new video comparison has come in for the new Nintendo Switch 2 Edition of Animal Crossing: New Horizons and its Switch original. Probably the biggest here is load time improvements. On Nintendo Switch 2, loading is reduced by about half. Elsewhere, things are mostly the same. Nintendo hasn’t touched up the graphics in any way, and unfortunately, it’s still...
Hytale's early access launch isn't quite ready to step out of Minecraft's shadow, but as a block-loving survival fan I'm still having a blast
I did not anticipate it'd take me an hour to start my first game of Hytale after loading it up, but you can thank its s…
Mac users hit with Logi Options+ loading bug, workarounds inside
Logi Options+ not working on Mac. Users report endless loading screens and broken mouse settings. Two temporary fixes available.
Why is Great Canadian Entertainment suddenly offloading most of its B.C. casino assets?
Great Canadian Entertainment, one of Canada’s largest gaming and entertainment companies, will be reduced to a presence of just two properties in B.C. after a recently announced agreement to sell its casino operation in Coquitlam to the Snuneymuxw First Nation.Founded in B.C. more than 40 year
Why is Great Canadian Entertainment suddenly offloading most of its B.C. casino assets?
Indebted from a private-equity takeover, it's selling off most of its B.C. casinos, with a First Nation taking over the core of its assets.
Boost Your Windows SSD Performance by up to 85% With This Server Driver Trick
It's almost as good as downloading more RAM.
Hacker hijacks Amazon accounts via Kindle ebook
An ethical hacker has warned Kindle users against downloading ebooks from third-party sites after discovering flaws that can expose credit cards
