SCROLL NEWS / DISCOVERY
Search headlines
Find the stories shaping the conversation.
Results for " agentic " 196 found 🔀 AI-powered Shuffle
How American Express Is Helping Business Owners Prepare for Agentic Commerce
How American Express Is Helping Business Owners Prepare for Agentic Commerce
Replit Expands Government Business Worldwide, Taps Former Amazon Executive to Lead Corporate Affairs
Replit, the agentic software creation platform used by builders in 85% of the Fortune 500 companies, today announced the expansion of its government business and the appointment of former U.S. Treasury diplomat and Amazon Web Services (AWS) executive Michael Greenwald as President of Corporate Affairs.
Sundar Pichai: Google’s Frontier Imperative, Compute Scarcity Dilemmas, Future AI Products and Businesses
Alphabet CEO Sundar Pichai has little patience for doubts about Google’s willingness or capacity to compete at the frontier of AI. His company’s latest Gemini 4 Argon model — long-promised and recently delivered — beats OpenAI and Anthropic’s latest public models on coding, math, science, and knowledge-work benchmarks and has placed Google’s AI back among the world’s best. Amid chatter that Google was fine with good-enough lightweight ‘Flash’ models, Pichai pushed back hard against any suggestion that company is limiting its research ambitions. “We won’t do it any other way,” he told me in a Big Technology Podcast interview recorded this week at Google’s Mountain View headquarters. Pursuing cutting-edge research is costly for Google, requiring compute and chips that otherwise could be sold at a high markup by its cloud business. But Pichai said allocating resources to frontier AI development is “ground zero” for Google. “Our cloud business is at any given moment in time forgoing short-term revenues because we are making compute trade-offs to support frontier AI development,” Pichai told me. In our conversation, we discussed Argon (which Pichai says processed three times as many tokens as Gemini 3 in the 72 hours after launch), AI’s role in developing better models, the AI interface of the future, competition with OpenAI, Google’s partnerships with Anthropic and Apple, data centers in space, and the nature of his optimism. You can read the full conversation, lightly edited for length and clarity, below, or listen to it on Apple Podcasts, Spotify, YouTube, or your podcast app of choice. Alex Kantrowitz: Google is back at the frontier with Gemini 4 Argon, which beats OpenAI’s and Anthropic’s public models on benchmarks spanning coding, math, science, and knowledge work. The conventional wisdom was that Google had given up on the frontier to focus on fast, good-enough models for its products. That was wrong? Sundar Pichai: Very wrong. Look, we are so committed to being at the frontier. It’s in our DNA. We won’t do it any other way. We’ve been thinking about the AI frontier for well over a decade now. The frontier is very dynamic. It’s very competitive. It’s very exciting. I couldn’t be more pleased with how Argon has come together. We took our time with it. We made sure we tested it safely and deployed it carefully. We took a bit longer than our other models, given its capabilities. Across the board, be it for agentic work, knowledge work, coding, creative tasks, Argon is one of the really good frontier models. The first 72 hours since we launched, it’s almost up to 25 trillion tokens used, which is about three times what Gemini 3.0 had in the first 72 hours. And we are capacity constrained as we are serving it. Excited, but there are always areas….. You know, the frontier is very dynamic, and you have to keep evolving the models. There are areas which we can see we need to improve as well. We’ll be at work on all of that as well. Which areas? On any given model, Argon would be across the board better at everything if you just look even two months back. But the space is so evolving. Even on coding, for example, there are pockets where we know we need to do better, and there’s clear feedback. We’ll make it better. The argument for standard models is that they’re already good enough to build great products, with the frontier valuable only for certain tasks. Meta built Muse with good models rather than the best, and Google could put standard models into Maps, Gmail, and Search. Why isn’t that sufficient? It’s definitely not how I have thought about it at all. We are deeply committed to driving the state of the art and actually pushing where this technology can go, all the way from transformers. That’s a journey we’ve been on. You will see us continue to do that. When you’re at the frontier, you get an earlier look into the emerging capabilities, earlier than others, which helps you better prepare for the future. We are fortunate to have a wide set of opportunities across consumer and enterprise. Some of our cloud customers, for sure, want the model at the frontier. The frontier models will be better at developing the next set of models in an era of recursive self-improvement. I would argue it matters there. Internally, we have definitely seen a steep point of inflection by deploying Argon to our developers. It definitely tackles problems which our previous models could not do. Being at the frontier puts you at the innovation frontier and helps you drive that as well. There is room for models of all sizes and capabilities. That’s why we did Flash, Flash-Lite, etc. I agree with you that you can develop very good product experiences with not necessarily the most frontier model, because they tend to be slower and more expensive too. But we care about the Pareto frontier, and we’re committed to doing all of that. You said developing frontier models gives you a look at what’s coming. What do you see on the horizon? What makes this moment so exciting is you build something which you think is a culmination of months of work, and you put it out. You already see clear headroom ahead. We’ve all been working on these agentic, long-horizon capabilities, but clearly the frontier is going to have multiple of these agents working together to do things. That’s an example of an exciting paradigm underway. It’s also clear to me that users, both in consumer and enterprise, will have access to some multi-agent framework for them, working all the time doing things. It feels like a moment of infinite possibilities. Sergey Brin said one of Argon’s core innovations was relying more on the model to help develop the model. Yet incidents involving OpenAI’s bots breaking containment and Google’s bots accessing a site they shouldn’t have suggest these systems still aren’t perfectly aligned. Isn’t there reasonable cause for concern about accelerating toward recursive self-improvement? There are three parts to your question. Let’s first talk about RSI itself. You’re right, Sergey mentioned RSI being a critical contributor as we are developing Argon. But to be very clear, today, in a continuum, we are pretty far from the end state of RSI or something like that. The right way to describe it is you have RSI loops along with AI researchers in various aspects of model development. You almost have agents helping AI researchers in some automated way work together to build these models. It’s good to at least understand where we are today, and it’s super exciting. I expect that continuum to keep making progress as we go into next year. The second question is around safety. Security comes from defense in depth. You’re talking about alignment, obviously. We all are doing a lot of work to make sure the models are as aligned as possible. But it’s an active area of research, and there are clear gaps. That’s not the only thing we rely on. You have the work you do on alignment, but beyond that, we put a lot of effort into real-time monitoring as these models are working. Making sure you have good telemetry into what’s happening so that you can intervene in real time if you see unexpected behaviors. How you do that, the architecture with which you do that, how you make sure it’s all sandboxed and isolated in a way that it can continue doing its job during real-time monitoring — these are all areas of investment. Obviously, when we deploy these models, giving customers, enterprises, control, both with real-time monitoring, but their ability to decide where they deploy, what type of data which models ever have access to, etc. All that work is very, very important, and we are doing a lot of that. As an industry, we need to constantly up our game and take the time to do all this stuff, including dedicating the compute required to make sure you can safely deploy these models. The third part is: Is it all enough? This is why there are a lot of conversations in the industry, and I think it’s reasonable to be worried. The right way to approach something you worry about is early on thinking about what more we can do. As an industry, you’re definitely seeing increased levels of work on alignment, testing for safety, etc., as we deploy these models. The next step would be developing better standards so there are cross-industry standards of how we do these things, including the ability to evaluate using independent evaluators. If you look at areas like financial controls, we have accounting standards, we have independent auditors. Those are all the next natural steps for us as an industry to work toward, and regulation will end up playing a role as we go through all of that as well. Google pursues both language models and world models, while some other AI companies like OpenAI have focused more heavily on LLMs as the path to AGI. Why is world-model research important to you? As humans, you obviously perceive the world around you, and you’re taking all those inputs. Making sure these models are able to understand the laws of physics, simulate the world in high fidelity. All that, I think, are key enabling blocks for intelligence. We are working across all possible applications of this. Take an area like robotics. It’s not an accident: If you look at embodied reasoning benchmarks, Argon is state of the art. That all comes from our experience building world models. Inherently, I would expect there will be architectures which will converge all of this in the future. But for us, we view this as, today, this jagged frontier that is intelligence. We want to be at the frontier of all the possible approaches, and that will help us both drive better research, better insights, and develop a much better AGI model, is how we think about it. Some people argue Google should focus on selling its computing resources rather than using them for AI research. Cloud wants to sell compute at a high markup, while researchers need a very big computer to build the best models. How do you balance those demands? Ground zero for us is making sure we are investing in the compute that’s needed to be at the AI frontier. In 2015, I set the company to be AI-first. AI is that key horizontal ingredient across everything we do, all the way from Search to Cloud to Waymo. That is ground zero for us, making sure both for research and developing the models that we have enough compute needed. Beyond that, we have these independent businesses which each have their own long-range plans. They’re using those long-range plans to project forward, and we plan for compute for all of them. The cloud team, they have a business and they have a projection and they have demand signals. They have committed contracts. With that help, we can plan on a long-term compute trajectory for them. We’re fortunate to have this wide set of opportunities. There’s no company in the world which isn’t making trade-offs at any given moment in time. But the first principles we operate on is making sure we have enough compute to invest in the frontier, and that’s non-negotiable, and everything else works on top. Remember, we are building our own compute, our own data centers. There are long lead times. The compute we have today is a function of decisions I made in 2024. Even if you think back to a year from today, last year the agentic coding takeoff hadn’t happened. Today, obviously, when you sit and look at anything, everyone is short—not just us, other companies—and everyone wishes you had two times what you had planned for. That comes with the territory. At any given moment in time, we are planning for what we would need two to three years out in the future. You’re going to find some decisions you make may look suboptimal two years out, depending on where the state of the art is. I don’t spend too much time thinking about it. The way I think about compute allocation is making sure Google DeepMind, both for research and frontier model development, is cutting edge in terms of compute access. After that, we plan for all our businesses. It’s great we can provide our compute to other people too, like we have done. It allows us to scale that up, get real tests for it in the widest application set possible, test it in the real world, and make sure we are at the cutting edge of technology at every layer of our AI stack as well. Ben Horowitz relayed on a podcast that Andy Jassy had said compute was sold out until 2028. Are you seeing the same thing? If today a customer is coming out and asking us for compute in ’27, that is incredibly difficult to do. Even last year we went out and leased some short-term compute to accommodate some customers and help them out. But yes, we are all dealing with similar timelines. Every major technology shift seems to involve overbuilding, from railroads to fiber. Do you think that’s happening here? It depends on the time frames you’re looking at. This level of build, if I take a long enough time frame, it makes complete sense to me. We are at early stages of the most consequential technology ever, and this is not like fiber. It’s not dark. We are using it. We are capacity constrained on a daily basis. The applications, the underlying technology is improving rapidly. But running a company, the way I work through the moment is you look at your demand signals. You look at your committed revenue opportunities that you’re getting in the form of contracts. You balance that with the underlying technical capability, or internally you balance how we are using it internally, the model trajectories. Based on all that, you take a disciplined approach—a confident, aggressive, but disciplined approach—to plan out compute. But when you talk about the broader economy as a whole, you’re having people operate with various levels of risk. I focus on what I can control as a company and plan through it. If your question is, can there be an overbuild? Yes, because you can have some constrained components. Remember, what matters is a set of everything you need to build out that compute. Memory, electricity, electricians…. Land. Like 2,000 components or more. You could get constrained at some point. There could be things which cause bottlenecks, and then you could have overbuilds. There could be external shocks to the economy. There are macro risks. All that is possible at any given time. But am I confident that the compute we are planning for, we will have uses for it given enough time? Absolutely, yes. I was in the room when you said AI would be as profound as fire and electricity. If you believe that, shouldn’t you give researchers the biggest possible computer, even at the expense of today’s cloud profits? We’re all doing some form of that. Our cloud business is at any given moment in time forgoing short-term revenues because we are making compute trade-offs to support frontier AI development. What you’re seeing is true already. But you want to be thoughtful. Even our AI researchers have some constraints, which are also good. That’s what inspires creativity. That’s what helps you come up with a more efficient algorithmic breakthrough, maybe to use less memory. What we are committed to is to invest what’s needed to stay at the frontier. OpenAI President Greg Brockman told me a few months ago that the ideal AI product might have no interface at all—just as natural as conversation. What do you think the interface looks like in the age of AI? Computing historically has been humans adapting to computing. With AI, it’ll be the other way around. The technology will adapt seamlessly to how humans live and go about their day-to-day life. You want to be able to speak to that AI when you want to. It has context. It can be persistently running somewhere, give you continuity. You’re not dealing with a lot of complexity. It’s natural. All that makes sense to me. That doesn’t mean it won’t have any interface, however. Nowadays, even on my phone, Gemini voice models are so good. I can just press the microphone button. I can just ramble. That’s why we call it Rambler. I can ramble. It intelligently corrects with my personal context. Even my friends’ names, weird names, it gets it right, even if it first guesses it wrong. It’s amazing. I’m not going in, backspacing, editing once I finish saying it. It’s remarkable how good it’s gotten. But there are going to be moments you will need an interface. You may want to entertain yourself. You may want to watch a video. You may ask your AI to create something, and you want to watch it, consume it. The way I think about it is, you want it to be as seamless and as natural as possible, and that you are not doing the hard work adapting to whatever the interface is. The technology is adapting to how you’re working at any given moment in time. When I asked Sam Altman about competition from Google, he argued that OpenAI was building AI-native products and implied Google was bolting AI onto existing ones. Do you accept that framing, and can a “bolt-on” compete with an AI-native product? First of all, it’s a very dynamic moment. We are all pushing each other. OpenAI has pushed us as a company too. This is not just we are doing something to them. Far from it. Yes, Google went into its own version of a code red too. In the AI moment, all of us have to respond with varying levels of urgency at any given moment in time. We are constantly building new products as well. Gemini was built from the ground up as a product—the Gemini app I’m talking about. Or tomorrow, as all of us are adapting to agents, are you bolting agents onto ChatGPT, or are they building it natively? How would you think about that? You can always put these questions back. I don’t think of it that way. I think of it as you want to build world-class product experiences that are AI-native and they are winning. Some of it is adapting products, some of it is rethinking new experiences. They will all be part of how you approach this moment. How do you expect to make money from these AI interfaces? If an agent recommends a product and I buy it, should it take a cut of the sale, even if that’s disclosed? I’m going to put enterprise aside. Consumers are going to increasingly value this technology. Take software developers today. Developers are willing to pay for these coding models because it really helps make them more productive. Nobody would want to code the way they were coding three years ago, or even the same time last year. When you’re delivering value, there’s value to be had. I’ve never seen examples where consumer products really scale to billions of users, where people are using it intensely, and haven’t had monetization to go along with it. If you’re providing value to people, there is inherent value to be had. The nature of it may change. You’re asking a scenario in which agents are doing stuff for you. At the end of the day, you will have commercial needs. The agents need to fulfill those commercial needs. The providers on the other side are going to see value in that. There’s going to be some mechanism by which it gets fleshed out. That’s what an efficient system does. But what matters is that you’re building those world-class consumer experiences. I promised myself one product feature request: Can you make Gemini-powered search in Gmail as good as searching Gmail through ChatGPT? What’s going on here? One of the exciting things we announced recently is the unified Gemini agent for enterprise, and we’re bringing that to consumers. These are persistent agents which run 24/7 in the cloud. We’re spinning up a micro VM effectively for every query, and from the runtime, the harness, the model—it’s all a headless service. It’ll work everywhere, and it has your context with you. You can address this agent whether you’re in Workspace, in Search, in Gemini. Agents are going to drive this unification from a customer experience, which people are looking for. When you have a broad product portfolio and you’re moving forward fast, there are always areas where you’re moving forward faster than some other areas. That’s valid feedback. But we are putting a lot of time and effort into making sure, as we build out our agentic experience, that unification and that capability works well across all these services. People have asked where Google’s equivalent of Muse or Dots is. Do you anticipate creating a new product to compete with them? Our goal with Gemini is to be a personal, proactive, and powerful assistant. That’s what we are working toward. You saw us describe the early versions of it at Google I/O this year with Spark. Effectively, already, if you open the Gemini app as a Pro or an Ultra user, you see Chat and Task, and they start giving you that agentic experience. With Argon now, that capability is dramatically accelerated. What we announced recently in our enterprise event, we are bringing that to consumers as well. Soon in Gemini, you will be able to ask something, and it’s a continuum. You don’t need to think about whether you’re asking a chat or an agentic task. It’ll just automatically do the right thing, and it can work persistently in the background securely and get things done for you. Already, empirically, when I look at what’s coming into Gemini in our app, last year you would have tagged them and most of them would have been chats. But a steadily growing percentage is already agentic tasks in Gemini. Yes, we are going to bring personal agents to our consumers. It’ll work in the Gemini app. It’ll work in Google Search. It’ll work in Gmail. You can address it, call it, etc. We want to make sure we’re getting it right. It’ll be powered by Argon, and we are very excited to bring it to our users. Google has invested and committed tens of billions of dollars to Anthropic. What do you see in Dario Amodei that makes you want such a deep investment and partnership? Very early on, when we looked at the founding team of Anthropic, it was very clear to us this was a world-class team which was involved in some of the earliest GPT models. We were quickly able to assess that they would be at the frontier. We’ve always tried to identify and partner with companies which would be at the cutting edge of technology. We’ve done that with SpaceX. We’ve done that with Anthropic. That’s what we saw, and we had a chance to partner. Is your partnership with Apple to license Gemini for Siri expanding, contracting, or standing still? For both us and Apple, that partnership was about translating the capabilities of AI into better experiences for users. I thought Apple has done a phenomenal job of incorporating—they customize these models and build it into Siri—and it’s been great to see the reception. We are committed to doing more. You’re going to do more with them? There’s nothing specific to comment, but it’s a partnership. It’s working well. What have you learned so far about the feasibility of data centers in space? It’s a very exciting project. Projects like this remind me—I still recall that moment in the desert for the DARPA Grand Challenge, watching the Stanley car compete in it, which ended up being the basis for Waymo. With all these technologies, you’re at a moment in time where it looks a bit like sci-fi, but 10 to 20 years down the line, it’s actually working in a meaningful way. I feel the same in space. We’ve always believed in betting big on these technologies. Quantum has been a great example of it. We’re probably 10 years into the journey. Waymo, we are almost 20 years into the journey. Space, we are two to three years into the journey. It’s clear to me that it’s an approach which will make a lot of sense in the future. Being in sub-synchronous orbits with 24/7 access to the most abundant form of energy we have access to will make a lot of sense for this exponential increase in compute that we are going to need. Having said that, this is a challenging project. The team has disciplined milestones. It was a super exciting milestone to send the TPUs up to space. Sometime, as we are speaking, we are expecting a “hello world” moment from them. I look forward to seeing that happen, and we’ll continue investing here too. You previously said 25% of the code written within Google was written by AI. What’s the number today? That number in our world today is meaningless in the sense that every part of what we now touch is influenced by AI deeply. What we are tracking with Argon, for example, it’s been a point of inflection. People are using Argon to really go tackle complex projects, including projects which affect millions of lines of code. In the four weeks since we have launched Argon, we have seen a 30% increase in CLs, and more importantly, almost a similar increase—a one-third decrease in rollbacks. People are writing code more productively and better code with it. We are tracking it with newer metrics now. How are you dealing with opposition to data centers? If you can’t build in the U.S. the way you want to, where else do you go? We’re very committed to building our data centers very responsibly. When I look at our largest data centers in some of the communities where we have been the longest, I visit them, I talk to people. There are people who come and tell us their kids have gone to college because of our investments in these communities, and we’ve grown in those communities. To me, this is about doing it right. Engaging with the communities early, doing it transparently. We are committed to making sure we are bringing incremental power, and we cover 100% of the cost of our power needs for these projects. It’s understanding the impact on neighborhoods and making sure you’re working thoroughly to address them. Done rightly, you can continue making progress. But we are constantly taking feedback into account. I do worry not everyone operates the same way. People have been surprised by data centers showing up in their backyard. Those are all important. As an industry, we should continue to do better. But part of the reason, in a 20-year time frame, we are thinking about things like data centers in space is making sure we are able to put these data centers where maybe the trade-offs will be different. What unproven area of AI is Google working hardest to crack? Is it continual learning—getting models to keep updating after release—or something else? There are multiple parallel paths. You saw with Argon we have a million output tokens. With our Gemini models early on, we had a one-million context window on the input side. These are all examples of breakthroughs we are constantly working on. How do you think about synthetic data? How do you think about continual learning? How do you think about more efficient architectures? We do dedicate a meaningful percentage of our compute to these research bets too. It always goes hand in hand, including our world models. Those are all bets we are placing. Sometimes, as part of these bets, there are ideas that emerge which just make their way into the mainline models. Maybe we don’t fully talk about it, but we have multiple steps in these areas. Working in technology requires optimism that building something will improve the world, but those bets can leave real consequences when they fail. I recently returned to a town in El Salvador where enthusiasm for Bitcoin had given way to unfinished construction and something resembling a ghost town. How do you maintain the optimism you spoke about in your Stanford commencement address while accounting for those risks? I genuinely have a deep conviction that improving technology and improving access to technology makes people’s lives better. It’s been my personal experience in my life, out of which comes that deep conviction, optimism. But with that, I’ve always felt humanity’s journey is always about working hard to harness that technology to make it work well for society. The optimism is not an optimism that you just build whatever you want, whenever. You work hard to make sure you’re building technology in the service of humanity, to make the human condition better. That’s got to be the defining goal as we are approaching it. With that framework, I’m very, very optimistic. We are in very early stages of the kind of progress that we are going to see on many areas. We’re all talking about areas like healthcare. A lot of the progress is ahead. The ability to make things in a fundamentally cheaper way on many things where things have been getting expensive for a long, long time, discovering new materials, etc., which really bend the curve—it’s tough to step today and understand the impact of all that, but they’re going to be very profound. I’m very optimistic, but we have to be rightfully concerned about all the ways this technology can go wrong. Sundar, I know you don’t do many of these. Thank you. Thanks, Alex. Appreciate it.
TraceLink Announces $1 Million Challenge to Build Agentic Business Processes
/PRNewswire/ -- TraceLink, the world's largest Agentic Business Network, today announced the $1 million OPUS Agentic Business Process Challenge, a global...
TraceLink Announces $1 Million Challenge to Build Agentic Business Processes
/PRNewswire/ -- TraceLink, the world's largest Agentic Business Network, today announced the $1 million OPUS Agentic Business Process Challenge, a global...
TraceLink Expands Global Investment in Track & Trace and Supply Chain Agentic Business Processes
/PRNewswire/ -- TraceLink, the world's largest Agentic Business Network, today announced a major organizational investment, establishing dedicated global...
Veryon Unveils AI-enabled Bizjet Maintenance Platform
Integrated workflow aims to streamline and simplify troubleshooting
Ubisoft plans cuts at German studios, adding to wave of workforce reductions
Ubisoft is planning to reduce its workforce in Germany, where it operates a trio of studios, by about 100 workers, Game File has learned. The company wants to offer employees buyouts to, should enough people accept, avoid having to lay anyone off. “In Germany, we initiated discussions with the Works Council last week regarding a proposed voluntary departure plan across Ubisoft Dusseldorf, Ubisoft Berlin & Ubisoft Mainz,” a company rep told Game File. A “works council” is a union-like group representing employees. “The voluntary plan is the primary measure being proposed to support the intended organizational changes,” the Ubisoft rep added. The company declined to confirm the scale of the potential headcount cut. Of the impacted studios, Mainz might be the best known to the public. It leads development of the Anno series and shipped the well-reviewed strategy game Anno 117: Pax Romana in November 2025. Berlin and Dusseldorf operate as support studios within Ubisoft’s creative network. Berlin recently supported the development of Rainbow Six Siege and the Rayman Legends Retold remake. Dusseldorf also contributed to those projects, as well as The Settlers and Ubisoft-branded escape rooms. Dusseldorf is the biggest of the three studios, with some 400+ workers, compared to just over 100 at the other two, a source familiar with the reductions and the teams told Game File. More than half of these new reductions of about 100 people are expected to happen at Dusseldorf. The planned cuts were announced at a town hall for the German studios on Tuesday of last week. In an email follow-up, the studio group’s managing director, Benedikt Grindel, said the cuts were intended to “best position our German studios for the future... with a clearer focus on their role, capabilities and organizational setup needed for the work ahead.” The email was an echo of the past. In January 2025, Grindel had emailed the Ubisoft developers in his region (or studio “zone”) to explain that a round of 185 cuts at the time, including about 65 at Dusseldorf, were “essential to improve our operational efficiency and ensure the long-term sustainability of our studios, the zone and the company.” A half-decade of continued job cuts Ubisoft has been aggressively cutting its headcount for years. At the end of September 2022, it had 20,729 workers. By the end of March 2026, that total was down to 16,590. Studio management is trying to trim costs as it grapples with a changing industry and a drought of releases, let alone hits. The company announced earlier this year that it had reduced its costs since 2022 by “nearly €325 million” ($362m) and planned to lower them through 2028 by another €180 million ($200m). In June the company said it was closing down its Winnipeg and Belgrade studios and cutting roles in its Barcelona team. Even commercial and creative success hasn’t spared studios from this ongoing company-wide contraction. As Game File was nailing down details around the proposed cuts in Germany, Ubisoft management was informing workers of cuts in its Singapore studio, which shipped a hit remake of Assassin’s Creed IV: Black Flag this past July. Some 70 positions in Singapore are being eliminated “to align the studio with upcoming production needs,” per a company rep’s statement to Game File. In early 2024, a large team in the Singapore studio launched the multiplayer pirate game Skull & Bones. It performed poorly but has been supported with regular content updates since. Many of Ubisoft’s studios around the world are heavily subsidized via regional grants and tax breaks. Singapore had been among them for years, but received no such local assistance in the 12 months ending this past March. News of the Singapore cuts broke today alongside word that Ubisoft is also proposing the closure of RedLynx, pending negotiations with officials. The Finnish studio behind the motorbike series Trials was founded in 2000 and sold to Ubisoft in 2011. As reported by VGC (and corroborated by Game File), Ubisoft management told employees it could not come up with a new, sustainable assignment for RedLynx. The 2025 workforce at Ubisoft Dusseldorf were straight-up cuts. This time, Ubisoft is hoping the reductions at its trio of German studios will largely come from workers taking a payout to leave. The company offered such voluntary dismissals at its Massive Entertainment studio (The Division) a year ago. For the German studios’ buyouts this time around, Ubisoft and the Works Council had discussed a formula that’d pay a worker’s monthly salary multiplied by 1.5 and by the number of years they’ve worked at the company, according to Game File sources. The formula is still subject to negotiations with the Works Council and has not yet been part of an official offer to workers. It could be adjusted up or down, and it could be supplemented with additional bonuses. Asked about the floated payout formula, a Ubisoft rep told Game File, “The terms of a potential voluntary departure program remain subject to negotiation with employee representatives. No final agreement has been reached, and we won’t comment on specific proposals while discussions are ongoing.” Game File is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber. Subscribe Item 2: Epic’s AI adventures Epic Games (Fortnite, the Unreal engine) is all-in on using AI to try to transform the way it makes video games and game-making tools. It’s leaders appear to be undeterred by any resistance among gamers and some game developers around generative AI—and instead emboldened by advances, particularly in AI-assisted programming. A recap of the last three days: On Monday, Bloomberg reported that Epic management told workers last week that development of an upcoming game combining Star Wars and Marvel characters would be “AI first.” Employees would be expected to use AI models such as Claude to code the game. A manager told Epic’s developers there would be “[n]o more hand-written code” and that art for the game would be “generated with full AI workflows,” per the outlet. Just ahead of that story running, Epic founder Tim Sweeney tweeted: “We have a huge pivot underway here at Epic, moving developers around to retool for maximum AI acceleration. The aim isn’t to do the same stuff with fewer people, but to enable everyone here to build MORE and BETTER stuff, and to bring the benefits to all teams using Unreal.” In a follow-up, Sweeney predicted that most game development processes would be at least three times faster a year from now than they were in January 2026, thanks to AI. To a person on X who told him he was failing to read the room, Sweeney said: “Much of the room wants it to be 2016 again, but it’s not. Our job, as the provider of tools and technology to much of the industry, is to design for the present that is and the future that will soon be.” On Wednesday, Epic announced it was releasing a faster launcher for its Epic Games Store marketplace. It did so by stating in a blog post as clear as could be that the launcher is an AI-enabled, vibe-coded project: Programming is hard. So, instead of doing that, we vibe coded a new Epic Games Launcher that’s way faster and more responsive than the current one, and has a new look. Programming is hard? That was unexpected phrasing from a company founded by a legendarily great programmer. Back in 2011, Sweeney told me that he’d picked up programming quickly as a kid and rejoiced at the power it gave him. “I could look at any program that anybody else had written and, with enough effort, build something cooler.” Item 3: The Xbox Labs experiment (and one of the people behind it) Microsoft’s Xbox division, which is flooding the zone with announcements lately, yesterday revealed a new internal tech-tinkering team called Xbox Labs. In the words of Xbox boss Asha Sharma Labs is “a small group of game makers, engineers and scientists whose job is to turn emerging ideas into things our players and developers can experience.” Conventional wisdom coming out of the announcement is that Labs is a team that’ll focus on ideas driven by generative AI. In her announcement, Sharma cited some of the dizzying achievements in tech and gaming over the past month as AI agents and other AI tools have been deployed to swiftly figure out how popular games work, then rebuild them and mix them up. She mentioned seeing ”games rapidly decompiled and rebuilt to run on hardware they were never originally designed to support” and said “[p]erhaps more surprisingly, we’ve seen many of the biggest games in the world have been remixed into the source code of another.” (Minecraft plus Elden Ring, for example, or Fortnite + Super Monkey Ball + Overwatch + Diablo + others.) Sharma also cited the work of Windows PC games being ported to run in browsers, a seeming reference to a recent AI-assisted fan port of Halo. Or perhaps Call of Duty or GTA. Sharma didn’t use the specific term “AI” in her memo, but she came close, citing AI “agents” and their potential for game-making tools that Labs could help craft: I think XBOX has an important role to play here. Coding became the first major workload for agents because they can write, run, and test software. Games share many of those same properties, and people spend more than 72 billion hours a year playing XBOX games, roughly five times the estimated time spent writing code worldwide. We’ve spent 25 years learning how to make great games, and I want our game makers to help shape what these new technologies make possible. As we learn, we’ll share the tools we build with developers across the industry, just as we always have. When I asked an Xbox rep if Labs was an AI-centric team, they said genAI was not the focus, describing it as having a broader mandate for using new tech in new ways in gaming. As an example, they pointed to Xbox’s recently launched disc-to-digital gaming usage rights program. Microsoft didn’t specifically name anyone involved full-time in Xbox Labs yesterday, but on social media and LinkedIn, I spotted Adrian Mato IDing himself as leading the creatio of tools at Labs and talking up the launch. Mato first caught my attention in late September, when he said on X that he was “starting to build a new team at Xbox focused on products and tools for how games get built and distributed as agentic tools become more capable.” At the time, he framed AI-based tools as a supplement to traditional game dev tools. “I don’t think we need to replace them,” he said. Since then, he’s publicly shown interest in AI and in games. He’s recently spent some of his X feed marveling at the power of AI agents and praising the lighting in the latest Gears of War. Following the Labs announcement yesterday, he enthused about the Xbox Labs’ stylish landing page. He wrote: “We built some fun tools to art-direct shaders in real time, experimented with interactive ASCII art, and sprinkled in lots of tiny details. “Move your mouse to leave a colorful trail that reveals the original footage, or interact with the video to trigger bursts of particles. “A lot of love went into making it all feel just right. “More to come”
From Bedrock to Business: The Last Mile of Agentic AI - SPONSOR CONTENT WEBINAR FROM AWS AND WORKATO
Sponsor Content Webinar from AWS and Workato.
Andor Health Earns 2026 Frost & Sullivan Competitive Strategy Leadership Recognition for Agentic AI-Based Care Orchestration
/PRNewswire/ -- Andor Health, the leading AI-native clinical services company, today announced it has received Frost & Sullivan's 2026 Competitive Strategy...
OpenAI CFO Sarah Friar Sees Cybersecurity As Potential 'Big Driver' Of Business Amid Rising AI Demand
Friar also discussed increasing agentic AI usage in Europe, while acknowledging that the region trails somewhat in adoption.
Confido Health nabs $12.5M to scale agentic AI across PE-backed medical groups
Confido Health is expanding beyond AI voice agents to build out an agentic AI operating system that can execute administrative workflows across a medical practice. | The company secured a $12.5 million series A extension following its $10 million series A round announced in September to continue scaling an AI platform for end-to-end healthcare operations.
AFP.com
Reply[EXM, STAR: REY] introduces Symbiosis , anAI-driven operating modeldesigned for enterprises integrating generative AI and agentic systems into their operating processes. Supported by adedicated platform, Symbiosisconnects strategy, portfolio management and execution, helping organisations align priorities...
AI Agents and the Rise of the Lab-in-the-Loop
Automation and robotics enabled the self-driving lab, but these largely follow pre-defined workflows. In this Q&A, Mark Fish explains how AI agents enable the "Lab-in-the-Loop" approach, which could accelerate life science research.
NTT DATA Business Solutions launches "Elevate", its Agentic AI-powered delivery framework, to help accelerate SAP transformations
NTT DATA Business Solutions, a leading global IT service provider and specialist in SAP solutions, today announces the launch of "Elevate", its next-generation Agentic AI-powered delivery framework. Elevate brings together the company's secure AI-enabled platform NTT DATA aXet, Natuvion's data transformation capabilities and SAP technologies, including SAP's Agent-Led Toolchain and SAP Business AI Platform. Elevate is designed to maximize the value clients gain from their investments in SAP and
HSBC’s Synopsys Upgrade Bets on a Business Model Shift. Can It Justify $700?
HSBC upgraded Synopsys Inc. (NASDAQ:SNPS) to Buy from Hold on September 25 and raised its price target from $490 to $700, implying a multiple of nearly 35 times HSBC’s fiscal 2027 EPS estimates. According to analyst Frank Lee, the company is evolving beyond its traditional slow-growth software profile as its licensing-plus-royalty model and agentic AI […]
Weekly Rundown: Nextech taps Assort Health, EliseAI for agentic AI; Elsevier rolls out clinical AI tool for nurses
Stay up to date on the latest in health tech, digital health and health AI news with this weekly brief. Read below for news from the week of Oct. 5 to 9. | The latest news from Nextech, Assort Health, EliseAI, Elsevier, Arcadia, Solera Health and more.
Microsoft tries to spark new life into Windows
Windows is slowly morphing into an agentic OS.
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 […]
Google brings agentic AI to Gemini, starting with businesses
Google is turning Gemini into an AI agent that can plan, execute tasks, and work across business apps and systems. The agent can delegate work to subagents, use multiple AI models, and even gets its own workplace identity, complete with an email address.
Beyond hours saved: Building the business case for agentic automation
The RPA-era ROI model misses most of the value agentic automation creates. This post gives AI center of excellence leaders a framework to size the full value of agents across time savings, exception handling, decision quality, and maintenance economics, and to prioritize which workflows to automate first.
Charm Security Appoints Dr. Elizabeth Huppert as Behavioral Science Director
Former JPMorgan Chase Vice President and Behavioral Scientist joins to advance behavioral science in agentic AI for fraud, scams, and cybercrime.NEW YORK, NY / ACCESS Newswire / October 6, 2026 / Security today announced the appointment of Dr.
ASUS Announces New 64GB Configuration of the Ascent GX10
ASUS today announced a new 64 GB configuration of ASUS Ascent GX10, expanding the lineup alongside the existing 128 GB model. This compact AI supercomputer is designed to bring local agentic AI development within reach of more developers, researchers and data scientists. Based on the NVIDIA DGX...
Your AI Agents Can Write Code. Can They Understand Your Business?
AI agents need to understand how businesses work. Graphify’s rise shows why ontologies and knowledge graphs matter for the agentic enterprise and economy.
September Monthly Review: Health MCP Guy?
For all the time I spend talking about APIs and data exchange, this newsletter has had embarrassingly little interoperability of its own. It’s old-school SaaS, baby - UI or bust. Not all of that is my fault! Substack, for all its popularity, is (as mentioned before) really frustrating as my system of record. The primary user it optimizes towards is the reader, whose attention feeds the app, the Notes feed, and the recommendation network that Substack is actually selling. Health API Guy is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber. Subscribe They’re falling into the same trap that Medium once sprinted into, albeit at a much more glacial speed. Younger generations may not remember (or never knew) it, but Medium once lured publications like The Ringer and Signal v. Noise onto its platform, only to lose them as it rebuilt itself around a paywall and an algorithmic feed. Substack’s original pitch circa 2020 was the antidote (you own your list), but every new feed feature chips away at it. I get it: writers are generic supply in the eyes of the Substack product team. Generic supply follows demand, so the consistent pressure on the PMs is to generate demand, which leads to every product cycle going toward discovery (Notes, recommendations, the app, live video). The end result lackluster author tools (laggy and broken analytics are truly crippling) all despite the premium price point. The archive is where it crosses from authorial gripes to actual reader impact, though. Across Substack and LinkedIn, I’ve published 1,218 posts and nearly 900,000 words since 2018. Substack’s native search can see only the 464 posts that live on Substack, and even there it’s woefully inadequate at surfacing relevant content (even for the guy who wrote it)! Readers have the same problem, writ large. Ask the Archive Fixing search (for both my readers and my own use) turned out to be extremely non-trivial, unfortunately: There’s no public API for readers to access content RSS exists for new content notification, but it only carries the 20 most recent posts and cuts paid ones off at the preview The only Substack MCP is an admin tool for analytics The export exists for authors, but it’s painfully awkward and slow. So I built something to work around that: the Health API Guy MCP. It’s a private MCP server holding the full text of everything I’ve written across Substack and LinkedIn. Add it to Claude or ChatGPT as a connector and you can ask things like: It pairs keyword search (for acronyms and case names, of which we have many) with semantic search. Every post is also labeled by topic, form, and court case, so you can pull a lawsuit’s (or multiple cases’) entire docket across Substack and LinkedIn in one query: Hail, Healthapiguyhydra This is also a bit of me eating my own cooking. I’ve spent the year arguing that headless healthcare is arriving and that developers want primitives over abstractions. Last week in Hail, Benihydra, I watched Salesforce plug its platform into Claude and bet that the records underneath outlive whatever interface sits on top: Salesforce must believe that the Hydra’s immortal head is the records, permissions, and business logic (in that they will survive whatever happens to the interface). It felt a bit hypocritical to keep my own archive behind a Substack search bar. The archive is the body. Substack is one head, and now whatever tool you might use is another. Will the EHRs follow my lead? Probably not imminently, but one can hope! Why Founding Members Only The MCP is available to Founding Member subscribers. A few reasons: It costs money and time to run. Hosting, the database, Clerk usage and AI compute for search and labeling are significant operational costs (that increase with usage). Substack gives me no way to automate access. There’s no API to check whether someone is a paid subscriber, so every user is manual operations for now. It’s a beta. We’ll need to improve things and parts will break. I’d rather they break for a small group of dedicated readers who will tell me about it. The Founding tier deserved a real benefit. Until now it was mostly a generous tip. Now it comes with something. If you’re already a Founding Member, reply to this email with the address you’d like provisioned. If you’re not and want in, upgrade here. Larger teams (such as companies) can get access too: group subscriptions at the Founding tier come with the group discount. The admin for the subscription can reply with the list of addresses to provision. Get 20% off a group subscription Pricing Founding membership is now a flat $200 a year, replacing the old suggested $150 that readers could adjust. Existing Founding Members will have access grandfathered in at their previous rate. Month in Review Articles Published Make Networks Dumb Again (Sep 11): AI is making old analog networks cheaper to use and flooding them with machine traffic in the process. A plea for agent-native rails with the intelligence at the endpoints, and for Washington to drag them to ubiquity. Interoperability, Unbuttoned (Sep 29): Interoperability is a series of workflows in a trenchcoat, and this post finally takes the trenchcoat off. A color-coded map shows which provider workflows went digital, which are fragmented, and which still run on fax. Video Content Antitrust’s System of Record Problem (Sep 3): The FTC’s investigation into Epic prompted a ninety-minute webinar, now posted with chapters and a smoothed transcript. We get into why market definition, missing B2B share data, and the SSNIP test leave antitrust poorly suited to systems of record, plus why information blocking may be doing more than any pending lawsuit. The Information Exchange: Cerner Milking Edition (Sep 18): This was so fucking fun. Ryan Tucker, Ryan Brickner, Brad Thorson, and I try an Around the Horn format, complete with points, penalties, and a mute button. We debate whether Oracle Health is dead, who would actually buy Cerner (IBM enters the chat), p(Doom), and why the front desk still hands you a clipboard. Regulatory Ask Not What Your Network Can Do (Sep 23): CDC’s RFI on public health data intermediaries lists use cases already served, unevenly, by eCR, lab reporting, IZ Gateway, and syndromic surveillance. The real opportunity sits in one question about AI: a network that can handle the next emergency without another interface project. Court cases Particle v. Epic: The Naughty Schoolteacher (Sep 4): After a summer-long vigil, the answer on Epic’s early summary judgment bid is two paragraphs and a remedial assignment. Particle advances to the next grade, and market definition gets another year of discovery. The Epic Discovery Files (Sep 8): Texas’s motions to compel arrive with a 251-page appendix of requests, objections, and meet-and-confer letters with Cravath. Inside: Epic using generative AI for document review, denied interoperability requests in scope, and Noerr-Pennington raised against lobbying discovery. Epic v. Health Gorilla: “Working as Designed” (Sep 9): Epic wants nothing to do with the data breach MDL, and explaining why puts fresh discovery on the public docket. The spiciest allegation: Health Gorilla told GuardDog to scrub its law firm business from its website instead of cutting off access. Veeva v. Epic: Legal Bypass Surgery (Sep 15): Veeva asks the Wisconsin Supreme Court to skip the Court of Appeals, while law professors and three policy groups line up as amici. Meanwhile, Epic’s lawyers apologize for another round of citation errors, including a quote that never existed. Texas v. Epic: Behind the Black Bars (Sep 17): The court portal served up Epic’s unredacted filings, revealing a $5 million document review budget and the AI workflow behind it. Also inside: the 200-plus company names Texas wants searched, and one very expensive Zoom invitation. OpenEvidence v. Doximity: Consequences Will Follow (Sep 25): OpenEvidence tells the court it will not comply with a discovery order, and its own lawyers at Quinn Emanuel say they won’t defend the choice. Add years of Slack reported unrecoverable, and Rule 37 sanctions come into view. Health Tech Keeps Choosing Violence (Sep 28): Doximity moves to end OpenEvidence’s case, Epic asks Texas to show its work, and a sleepy Audacious Inquiry patent fight sprouts antitrust and information blocking counterclaims. The main event is a long-sealed Cognizant v. Infosys ruling that hints where system of record antitrust claims can survive. EHRs One Verb for Oracle Health (Sep 14): Oracle’s earnings call quantifies every data center to the decimal point, while Oracle Health gets a single verb: “accelerate.” Panning the few healthcare mentions turns up an agentic care management system, a research-to-care ledger play, and the question of whether healthcare gets Larry’s patience. Epic’s Forretress Under Siege (Sep 30): Judy Faulkner tells a panel Epic paused hundreds of projects to harden its software and calls it a “shame,” while a spokesperson insists the roadmap hasn’t changed. Project Glasswing, a Pokémon-branded QAN sprint, and a trove of Jodel grievances reconcile the two Epics and preview the Jevons paradox coming for all software. Industry Analysis Scanning for a Network (Sep 1): Scan.com raises $220M to build the imaging network labs never got, speedrunning Zocdoc’s consumer-to-B2B pivot. Its wedge to network density may be the most American one available: personal injury litigation. The Front Desk Wants the Damn Whole Building (Sep 16): Hello Patient buys Converse Health, pushing its AI front desk into referrals, chart work, and prior auth. The copilot categories keep blurring, and your distribution partner is still your final boss. Cross-industry Comparisons A Different Patchwork Quilt (Sep 10): Healthcare loves to envy open banking, but the US version has no mandate, a toll booth at Chase, and data quality problems of its own. The grass isn’t greener; it’s another American patchwork quilt, and perhaps a worse one. Hail, Benihydra (Sep 21): Salesforce brings its platform into Claude, Slack, and its own agent, inviting the interfaces that might replace it. Cut off one head and more grow back, so long as the records underneath stay put. Living on A Trade Secret Prayer (Sep 22): Six credit unions are suing Fiserv, the Epic of core banking, by claiming their own member records as trade secrets. Banking has no Cures Act, so its lawyers reach for legal alchemy that healthcare has already tried. Other News None this month External Media Nabla Accelerate NYC: I kicked off fall conference season at Nabla’s Accelerate event in New York, where Delphine Groll opened by asking the room to picture the company it wants to become and the one it doesn’t. I liked her framework of bureaucracy, fear, and ego as the blockers. I also love vendor conferences/events in general - different vibe than industry biggies. Healthcare 101 with Nikhil Krishnan: I briefly guest lectured Out-Of-Pocket’s Healthcare 101 class. Appreciated Nikhil’s takeaways cover whether AI doctors can query HIEs, why healthcare’s data-liberation rules beat most industries’, and the hi-res graphic is in the comments. HCN: Epic Probed by the FTC: I joined the HCN guys to talk through the FTC’s investigation into Epic. Cedric’s new hairdo was an unexpected plot twist. The episode also covers MFN drug pricing, healthcare M&A and Eli Lilly’s acquisition streak. Fall Conferences: At eHealth Exchange’s Annual Meeting in Austin on Oct 27, I’ll moderate a panel on the tensions between AI, HIEs, and timely patient access with Jean Ross (Primary Record), Therasa Bell (Kno2), Byron Crowe (Doctronic), and Michael Marchant (Freenome). Here’s the rest of my schedule if you want to meet up: CommonWell (Redwood Shores, CA): Oct 13-14 Will be participating in a fun Jeopardy event Open@Epic (Madison, WI): Oct 20-23 Will be enjoying Verona, Willy Street, and other local adventures eHealth Exchange and Sequoia Project (Austin, TX): Oct 27-29 Yeehaw HLTH (Vegas): Nov 15-18 Groundhog Day, Venetian Edition RSNA (Chicago): Nov 30-Dec 2 See me at “Governing AI in Breast Imaging: What Health Systems Are Actually Doing” Digital Health Counsel 2026 AI Summit (Seattle): December 2-3 Will likely be speaking to a bunch of lawyers about information blocking Posts I Liked ChatGPT’s Epic integration: game changer or nothing burger?: Joshua Liu, MD offers five thoughts on OpenAI’s Epic launch, starting from the point that it rides the same open APIs available to anyone. He argues read-only access leaves ordering to Art and the ambient startups, and asks why HCA is piloting it alongside Commure. He’s sorta a must-follow at this point - consistently excellent, thought-provoking takes. Epic, Lilly, and the Discovery pass: Seth Chaney reads Epic’s first major Discovery deal (with Eli Lilly), as well as its revised developer terms on commercial displays, as the pipe finally opening between EHRs and pharma. Life sciences and health continue to collide. Scott Rossignol on EHR payer platforms and ePA fees: Scott argues the top ambulatory EHR vendors are turning the prior auth mandate into a per-member-per-year toll on payers, for an API whose operating cost has nothing to do with covered lives. His fix is two lines of regulatory text: put the CRD/DTR/PAS certification criteria in the Base EHR definition, and bring payer-facing ePA connectivity under § 170.404’s cost-based fee rules. It’s a good solution. Health API Guy is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber. Subscribe
HP Work Relationship Index 2026: Workplace Health Rebounds as Workers Turn to Agentic AI, Skills-Building, and Better Tech Tools
Today, HP Inc. (NYSE: HPQ) released the fourth annual HP Work Relationship Index, a global study of more than 19,000 desk-based workers across 15 countries. This year’s study finds that workers with the healthiest relationship with work are becoming more confident in how they navigate uncertainty.
Microsoft rewires AI’s context layer for business
Presenting new capabilities for Fabric, Power BI, OneLake, and Azure SQL, the company took the stage in Barcelona to outline how its customers can connect data, business context, and AI to fuel the agentic enterprise.
Xsight Group Featured in Harvard Business Review on AI Agent Orchestration Across Silos
HONG KONG, Sept. 28, 2026 /PRNewswire/ -- Xsight Group (known as Xsight), a Hong Kong–headquartered cross-border technology group founded in 2004, has been featured as a research case in "How AI Agents Orchestrate Work Across Silos," a special report published in the September–October 2026 issue of Harvard Business Review. The article examines how a new generation of AI systems can connect functions and reduce the coordination friction that slows companies down, citing Xsight's practice of decomposing complex cross-departmental decisions into modular tasks and building a layered agentic AI orchestration system as a valuable model for cross-border organizational coordination.
Could AI Commerce Reshape Meta (META)’s Business Model?
Meta Platforms, Inc. (NASDAQ:META) is expanding the commercial capabilities of its Muse AI assistant through a strategic partnership with Shopify, giving investors another potential avenue for AI monetization beyond digital advertising. On September 21,Shopify CEO Tobias Lütke announced a partnership with Meta to bring Shop Pay agentic checkout to Muse, allowing users to shop from […]
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
Microsoft unveils Copilot super app, targeting business users with AI agents
Microsoft is launching its super app as a central interface through which customers can chat with AI bots, write code, and access a new "Autopilot" feature that brings agentic capabilities to its flagship AI app.
LRZ Hosts TPC Europe Hackathon on AI for Science and Supercomputing
Agentic AI is currently a hot topic in the scientific community: during the Trillion Parameter Consortium’s autumn hackathon at the LRZ, teams presented practical solutions designed to make supercomputers more accessible and research projects more efficient. Sept. 24, 2026 — In the seminar rooms of the Leibniz Supercomputing Centre (LRZ), small groups stand in front […]
LRZ Hosts TPC Europe Hackathon on AI for Science and Supercomputing
Agentic AI is currently a hot topic in the scientific community: during the Trillion Parameter Consortium’s autumn hackathon at the LRZ, teams presented practical solutions designed to make supercomputers more accessible and research projects more efficient. Sept. 24, 2026 — In the seminar rooms of the Leibniz Supercomputing Centre (LRZ), small groups stand in front […]
Wilson Sonsini Advises Ferry Health on $9 Million Fundraise as Company Emerges from Stealth
On September 16, 2026, Ferry Health, an agentic artificial intelligence (AI) platform supporting patients in administrative tasks, launched out of stealth, raising $9 million in seed funding from…
ChatGPT mobile app gets voice-based agentic features
Pro and Plus users will be able to use the Work tab on their phones
BizTrip AI Brings Business Travel Booking Directly Into Claude and ChatGPT
SAN FRANCISCO, September 23, 2026--BizTrip AI, makers of agentic AI for corporate travel management, today announced that its core trip-planning and booking experience is now available natively inside both Claude and ChatGPT. The integration uses the Model Context Protocol (MCP) — specifically MCP "apps," which support rich, interactive rendering directly in a chat conversation — allowing travelers to search, compare, and book business travel without switching to a separate app or website.
Why NuoData Believes Enterprises Need a Different Way to Connect Data, AI, and Automation — Not Another Place to Store It
NuoData’s leadership team sat down to answer six questions about the platform’s origins, its position alongside Snowflake, Databricks, Microsoft Fabric and Google Cloud, and where they believe enterprise data and AI architecture is headed next. The answers have been lightly edited for length. NuoData describes itself as an end-to-end enterprise Data, AI and Agentic orchestration […]
Meta (META)’s Stilla.ai Deal Could Turn Business Agent Into a New Monetization Engine
Meta Platforms, Inc. (NASDAQ:META) is taking another step toward monetizing artificial intelligence beyond its core advertising business. On September 9, the company acquired AI startup Stilla.ai. The acquisition will strengthen the company’s agentic AI capabilities as it moves to capitalize on growing demand for AI agents that can handle business transactions. Plans are underway to […]
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.
From Python and Statistics to Agentic AI: 5 Data Science Certificate Programs in 2026
Python and statistics still sit at the center of data science, but the work surrounding them has expanded. Professionals now move from cleaning data and testing hypotheses into predictive modeling, recommendation systems, deep learning, Generative AI, and applications that can retrieve information or act through external tools. That progression matters when comparing programs. A curriculum […]
APPLY Appoints Adam Steingart as Global General Manager, Entertainment
APPLY, the global Agentic Customer Experience (ACx) partner backed by Interlock Equity, today announced the appointment of Adam Steingart as Global General Manager, Entertainment.
Google Home Is Going Agentic Via Integration With The MCP Standard
Premium Advanced users will get an early access build "in the coming weeks."
Anthropic merges its Claude chat and agentic Cowork products into a single AI assistant as part of a push to build an AI superapp
Anthropic is folding its chat and autonomous task-execution products into one Claude interface, while adding collaborative document and presentation tools.
Gant Adds AI Assistant to Self-Service Platform
Tool uses agentic chat to take action on a traveler’s behalf
Focus on agentic commerce as customer data science firm dunnhumby wraps Retail Innovation Forum USA
dunnhumby held its annual Retail Innovation Forum in Bentonville, Arkansas, USA last week, bringing together retailers, brands, technology leaders, scale-ups, startups and investors. In a LinkedIn post, Dave Clements, Chief Strategy Officer at dunnhumby, said: "In many ways an eclectic group to
Accenture and Google Cloud establish Gemini Enterprise Business Group
The partnership aims to accelerate enterprise adoption of agentic AI, offering expanded industry expertise and technical resources.
Pamplin+ transforms undergraduate business curriculum into four-year experiential internship
Launching in fall 2026, Pamplin+ transforms the Pamplin College of Business undergraduate experience through hands-on learning, required internships, mandatory agentic AI coursework, and cross-disciplinary corporate projects.
Accenture and Google Cloud are creating a joint unit to embed AI engineers in enterprises
The Accenture Gemini Enterprise Business Group will train and deploy forward-deployed engineers to help enterprises build and scale agentic AI applications
AI Factories Turn Infrastructure Into a Business Advantage
AI factories are an industry must as enterprises move from experimentation to scaled agentic workloads, says Penguin Solutions’ Aaron Johnson.
Parakeet Health lands major AI partnership with national dermatology group Qualderm
Parakeet Health landed a major partnership with Qualderm, one of the nation's largest dermatology organizations, as the company looks to apply agentic artificial intelligence to a growing specialty | Parakeet Health landed a major partnership with Qualderm, one of the nation's largest dermatology organizations, as the company looks to apply agentic artificial intelligence to a growing specialty care access problem.
Accenture and Google Cloud Deepen Partnership with Formation of New Accenture Gemini Enterprise Business Group
Accenture and Google Cloud today launched the Accenture Gemini Enterprise Business Group, a global group designed to help clients scale Gemini Enterprise outcomes in the agentic AI era.
Logitech launches $99 MX Keypad for coding and AI workflows
Programmable physical controls are becoming key to power user workflows, including agentic programming. Now Logitech is bringing to market a...
Presidio’s Resonate uses agentic AI to turn sports archives into searchable inventory
Presidio is pairing Captivate and Resonate to automate tagging, retrieval, and clip creation on AWS Media Lake, aiming to speed monetization and production work
Nvidia Wants to Turn Your Idle PCs Into a Personal Home Data Center With 'PAIR'
Nvidia's PAIR links your unused laptops, desktops, and Macs together to run agentic AI locally without subscription fees or cloud privacy risks.
August Monthly Review: ChatGPT In Epic
Death, taxes, another frontier lab healthcare launch inspiring truly insane LinkedIn takes. As much as I want to stay away, I repeatedly am the crewman unplugging his ears whenever we pass these sirens. So it’s not the first and probably not the last time we dial up the typewriter rather than tie ourselves to the mast. The full blow-by-blow timeline: Health API Guy is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber. Subscribe The End of the Standalone PHR (Jan 8): ChatGPT Health arrives for consumers, at the time powered by b.well. I was bullish on the approach in that the best way for the PHR to finally solve distribution was to meet consumers were they already are (i.e. When Horizontal Meets Healthcare (Jan 9): OpenAI for Healthcare puts out the shingle for enterprises, with SharePoint ingestion standing in for the EHR integration the product actually needed. Another One: Anthropic’s Healthcare Debut (Jan 16): Claude's version, on HealthEx rails instead of b.well, with Agent Skills for FHIR development as the one true vertical-specific investment. ChatGPT for Clinicians: The Trap Sprung (Apr 24): OpenAI filled in the missing GTM middle with a free product for verified clinicians, effectively peeling the enterprise wrapper off ChatGPT for Healthcare and taking the PLG fight directly to OpenEvidence. OpenAI’s Second Attempt at Health (Jul 31): a relaunch of their patient specific app, now branded Health in ChatGPT, where the privacy partition came out, the connectors thinned, and b.well disappeared in favor of first-party FHIR work against a smaller network Beyond a Shingle So what was announced now? Today, we’re introducing a new electronic health record integration that brings authorized patient context from Epic into ChatGPT for Healthcare, along with the Healthcare Public Data plugin for direct, structured access to official healthcare datasets like PubMed, DailyMed, and CMS Coverage. Together, these capabilities bring ChatGPT closer to the systems and sources healthcare teams trust, while supporting the controls and compliance healthcare work requires. The core portfolio is unchanged from prior announcements we delineated and discussed (aside from the relaunch of patient-facing). Since trade publications still seem to be confusing them by using the wrong names: Enterprise sales motion: ChatGPT for Healthcare Protecting their core product / PLG: ChatGPT for Clinicians Consumer/patient-facing product: Health in ChatGPT So really, this is a feature release (specifically for ChatGPT for Healthcare when it comes to EHR integration). What’s cool about it, though, is that we are moving beyond just putting out a shingle. Everything in the January enterprise launch was horizontalizable: Horizontal tech companies thus generally start verticalization with their “putting out a shingle phase”. Telling the world you’re open for business as a horizontal tech company reliably attracts early inbound interest from buyers who are already trying to force-fit horizontal tools into domain-specific workflows and are eager for any signal that the vendor intends to support their use case more directly. Nothing was really vertical specific, which is exactly what you'd expect from the ultimate hypergrowth horizontal company dipping their toes into specific industries. But eventually you have to stop changing the sign on the door and start changing the product. Can you guess where the horizontal product was going to run out of road? This is the core dilemma writ large: verticalization requires differentiation, while scale economics of consumer and horizontal push toward unification. I’ve buried the lede so deep here it will be painful to some readers, but what should be shocking to no one is that my perspective is that EHR integration is the only path that meaningfully resolves this tension for ChatGPT for Healthcare. Integration with Microsoft Sharepoint is fine, but it is categorically insufficient for any product that hopes to influence clinical decision-making. SMART Money Well here we are, eight months later! Real vertical investment has begun, which surprised a few people. One friend in the industry messaged me: “Absolutely no way Epic gave ChatGPT API access.” Looking forensically at the announcements, the integration is notably read-only across appointment notes, laboratory results, medications, and specialist documentation. So while I’m not a betting man, I'd wager the house this is a SMART on FHIR launch: That sounds a lot like USCDI! The recommended path for the workflow they outline would be SMART on FHIR Their Health AI lead’s post is evocative of both standalone and EHR launch SMART, which another termed “ChartGPT” and “EHR Plugin” The implementation burden for SMART on FHIR is the lowest of the available integrative paths This is the team who literally did patient-facing SMART on FHIR with Epic last month Most importantly, if you’re a horizontal company, you generally still haven’t taken the full plunge and thus (overly) value reusability. SMART on FHIR is the logical compromise: healthcare-specific enough to matter, but standardized enough to reuse. Nobody Asked Judy Another friend asked “I thought they had to go to each health system, or did they go direct to Epic and now health systems opt in?” Given the trust paradigms in healthcare, this functionally cannot be Epic hoovering up all the data for OpenAI and shoving ChatGPT in their customers’ faces (or they would revolt). However, I do believe the aforementioned terrible coverage by TechCrunch and other outlets is probably responsible for that question, given the 325 million reference. When using FHIR as a business associate to an Epic customer, there is no EHR gatekeeping to speak of, as that isn’t how fhir.epic.com works: You register an app You pick your APIs You test against the sandbox APIs You list as “Ready for Production” Hospitals pick your app You test and go live with them Bluntly, nobody had to say yes or no in Verona (nor were they given the chance). Vendor Services (their next developer tier up that I doubt OpenAI is using quite yet in this initial release) certainly has contractual paperwork that some resent, but even there, the twin pressures of mounting antitrust and information blocking make outright gatekeeping increasingly fraught. My friend’s sentiment is, to me, representative of a broader industry neurosis, perhaps a sort of scar tissue of prior eras: people assume the gate is still there and never actually try the door. There are certainly other ways Epic (and any EHR) can put its thumb on the scale, but they are not so dumb as to stand in front of a federally mandated API and play bouncer at this exact moment in time. So I think they can and probably should go deeper, as this release is at best parity and at worst behind vertical-specific competition. OpenEvidence did a basic SMART launch with Sutter Health in February, allowing for more convenient evidence search by providers. They then add patient-context (the equivalent SMART on FHIR flow to what we see here) when Cedars-Sinai went enterprise-wide in May. UpToDate has four distinct applications across deeper workflows like patient engagement listed in Epic Showroom and announced a partnership with Epic to power Art at UGM Abridge and other ambient scribes have invested deeply into the deepest bidirectional clinical copilot workflow with integrations well beyond SMART on FHIR. In that light, the “Who is OpenAI primarily targeting here?” is clear. This release brings them into striking distance of OpenEvidence, but not the others quite yet. They are the only one of the three you can reach without building something truly Epic-specific. When will they go further? Competition is a great motivator to overcome the horizontal demons and start building things that can’t be reused: not across industries, not across EHRs, maybe not past the customer you built them for. That’s the price of actually verticalizing. The question is if and when OpenAI will be willing to pay it. Month in Review Here is the monthly review. As a reminder, this is a regular round-up of the month’s posts and other content to surface things you may have missed across regulation, litigation, interoperability, and beyond. AI assistance is used in these bullet summaries so I can focus on articles. Articles Published: None this month Video Content: The Information Exchange: Standards-based Thruple Edition (Aug 14): Back to school for us too, with Brad reporting in from the CMS Health Tech Ecosystem’s one year anniversary in DC. We get into the January CMS-0057 deadline, the fall rulemaking reading list, and why enrollment still keeps most apps off FHIR. The Information Exchange: Epic Dúnadan Edition (Aug 25): A UGM roundup with Ryan Brickner joining for the first time to check our takes against what the building actually thinks. Rangers, Ergo sitting on top of EHI, and a down-market lineup that needs some cuts. Regulatory: Stacked Deck, Bad Hand (Aug 05): ONC’s website refresh took the entire HIT Policy Committee record with it, so I rebuilt the archive and went looking for the regulatory capture story everyone assumes is buried in there. Stacking the deck and winning the hand turn out to be very different things. The Sixth Generation of Patient Access (Aug 19): A quick catalog of the five generations of patient access we’ve layered on since HIPAA, and the two candidates now competing to be the sixth. One is planned. The other is the market routing around the plan entirely. Court cases: Three Cases Walk Into a Docket (Aug 06): Two surprise settlements cleared the board in a single week, and then Judge Maddox dropped ninety-four pages on Vyne v. Henry Schein. The sleeper is a DMCA holding that makes direct-to-database a considerably riskier business model. Epic v. Health Gorilla: Into the MDL (Aug 10): Nine class actions are headed to Miami, and the Panel signaled it wants to bring the case that spawned them along too. That would leave Epic arguing the requests were obviously fraudulent in one courtroom and unknowable in the other. Amazon v. Perplexity: Agents Are Legalized! (Aug 11): The Ninth Circuit vacated the injunction against Comet, holding that the user is the one accessing the servers rather than the company that built the agent. A real win for agentic access, and a much narrower one than the headline suggests. Veeva v. Epic: Come At Me, Bro (Aug 26): Epic’s response brief wants the dismissal affirmed and, unusually, wants the opinion published as precedent. Buried in it is Epic’s own description of what its non-competes actually prohibit, which current and former employees should read closely. EHRs: The Contract Epic Would Never Sign Today (Aug 04): A 1999 SEC exhibit catches Epic licensing nearly its entire product line, Tapestry included, to the company that became TriZetto. The marketing services menu attached to it is the part that will make you blink. Forecast from Verona (Aug 07): The Epic Almanac makes one argument six different ways: the AI is only as good as the networks behind it. Emmie headlines, Art’s context stack is the most ambitious part, and Penny finally gets an autonomous coding date on the calendar. Works With Epic MyChart (Aug 17): Epic’s first new Showroom category since the death of Workshop certifies hardware instead of software, badge on the box and all. Made for iPhone, but for blood pressure cuffs, and rough news for anyone selling the RPM stack sitting in between. UGM Hot Takes 2026 (Aug 20): Everyone else covered the AI announcements, so I went after Savvy deleting the payment gateway, Rangers as Boost with the timer removed, and a down-market lineup with too many entries. Plus the Health Grid tidbits I cannot help myself on. MyChart Central Grows Up (Aug 24): Device data and Emmie turn Epic’s identity hub into a full consumer platform, which is both a logical answer to ChatGPT and a bit of a mistake. It also happens to be the best scraping target Epic has ever shipped. Industry Analysis: The Wrong Yardstick (Aug 12): Vertical software keeps getting judged against Superhuman and Notion, which is the wrong bar entirely. Your user is comparing your product to a whiteboard and forty phone calls before lunch. Primitives vs. Abstractions (Aug 27): Developers want building blocks, systems of record prefer to hand out business logic, and both sides have a real case. There’s no test from the outside that separates the engineering reason from the competitive one, which is why this keeps ending up in court. Cross-industry Comparisons: The Two Kinds of Platform Power (Aug 18): Attention power and record power are different problems that keep getting handed the same regulatory toolkit. Congress’s latest swing at Big Tech shows both what’s possible and where it falls apart. The Dogs of (Platform) War (Aug 21): A CourtListener alert turned up a property management fight with healthcare’s exact shape and none of the Cures Act. Shell prospects, ghost accounts, a notetaker bot that came back to haunt someone, and a lawyer arguing both sides of the same theory in two states. Fractals All the Way Down (Aug 31): Yardi dominates property management right up until you zoom into one segment, where AppFolio owns it outright and Yardi doesn’t appear at all. Where you draw the line around a B2B software market is about to decide an awful lot of cases. Other News: If Judy Had the Courage (Aug 13): A short eulogy for the era when software was allowed to look insane: WinAMP, bold colors, and everything the grey chatbot era has taken from us. External Media: Portland Monthly Health Tech Meetup: Erin O’Brien channeled my own feelings - maybe it’s just because it’s August and beautiful, but the monthly edition of the PDX Health community meetup was a ripper. Make sure to reach out if interested to join for the next one. STAT’s Coverage of Epic This Month: Brittany Trang of STAT did a fantastic job of breaking the FTC investigation that’s been lurking, as well as UGM related coverage, so I had to give her her wish of a meme. Fall Conferences: I’m pumped to be kicking off my fall conference season in NYC at Nabla Accelerate next month, which Chrissy provided the link to apply for. Here’s the rest of my schedule in case you want to meet up: Commonwell (Redwood Shores, CA): Oct 13-14 Open@Epic (Madison, WI): Oct 21-22 eHealthExchange (Austin, TX): Oct 27 Sequoia Project (Austin, TX): Oct 28-29 HLTH (Vegas): Nov 15-18 RSNA (Chicago): Nov 29-Dec 3 Posts I Liked: HTI-6 Should Unbundle API Certification: Josh Mandel makes the case that HTI-6 should split (g)(10) into separate authorization and data certifications, so a PACS or a genomics platform can certify only the role it actually performs. Imaging is the urgent example, but the structure solves a much bigger problem. On implementing ePrior Auth with Epic: Scott Rossignol’s field notes from a live ePrior Auth build on Epic, including the two app registrations nobody warns you about and the CPT mapping problem waiting at the end. CMS-0057 implementations are going to be such a beast. Healthcare point solutions are starting to look a lot like streaming services: Spencer Dorn runs the cable-to-streaming arc against health IT and lands squarely on the bundling half of the cycle. Bundling can be good! It can also be bad! Life is nuanced. Health API Guy is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber. Subscribe
Sonos Ace Ultra, Beam Ultra, Sonos Fabric, and a New App: Everything Sonos Just Announced
Sonos is cramming AI into its software because it’s “very hot these days.” The new features, which include agentic automation, are opt-in.
Digital Science's 2026 Catalyst Grant Seeks "Agentic AI Workflows You Can Trust"
Up to £25,000 to be won for early-stage solutions building research agents - with audit, governance and provenanceLondon, United Kingdom--(Newsfile Corp. - September 1, 2026) - Digital Science, a leading technology company serving stakeholders across the research ecosystem, today launches its Catalyst Grant round for 2026 to support innovative technology ideas. This year's theme is: Agentic Workflows You Can Trust.Up to £25,000 will be awarded equity-free to individuals, startups or research...
Digital Science's 2026 Catalyst Grant seeks agentic AI workflows
Up to £25,000 to be won for early-stage solutions building research agents – with audit, governance and provenance
Digital Science's 2026 Catalyst Grant seeks agentic AI workflows you can trust
Digital Science today launches its Catalyst Grant round for 2026 to support innovative technology ideas. This year's theme is: Agentic Workflows You Can Trust.
Run NVIDIA BioNeMo NIM Microservices for Protein Structure Prediction in Claude Science
Agentic AI is changing how research is done. AI scientists can read papers, propose hypotheses, call models, and determine which experiments to prioritize next. First proving their value in software…
Colorado Sues EarnIn For “Immoral, Unethical, Oppressive and Unscrupulous” Business Practices
Hey all, Jason here. When this hits your inbox, I should be en route to the airport — not to go anywhere myself, just to pick up my better half. I must admit, it’s been strange to be home alone (well, with two dogs) for two weeks! Really lets me “optimize” to spending more time on reading/writing/working, whether that’s healthy or not! Looking forward to getting back into a (somewhat) more balance routine. Subscribe or Support by Upgrading Compare Notes With the People Who Built Modern Fintech Partner content: Remitly CEO Sebastian Gunningham. Increase founder Darragh Buckley. Linda Du, Co-founder & President at Valon. They’ll unpack the decisions behind building and scaling major financial products: what they got right, what broke, what they would do differently, and how AI and new infrastructure are changing the operator playbook. Most events tell you where fintech is going. NerdCon puts the people who built the present in the room and asks them what comes next. Expect an operator-led conversation grounded in real products, real trade-offs and lessons earned the hard way. Fintech Business Weekly readers save 20% on Regular tickets with code FBW20 before standard pricing ends September 11. Meet the Builders Things To Know & Other Good Reads Advent and Stripe Abandon $50 Billion Pursuit of PayPal (Bloomberg) Trump Family’s New Crypto Bank Is Backed by Abu Dhabi Sheikh (Wall Street Journal) The multiplying risks of financing data centres (FT) Let the Bond Market Speak (Wall Street Journal) Did we waste a crisis? Modest proposals to reform deposit insurance. (Fintech Takes Banking) Untangling Guggenheim: How Private Credit Built Its Own Universe (Net Interest) The Hater’s Guide To Circular Financing — Part One (Ed Zitron) Fiat Ventures Rebrands to FGV Capital, Announces $35M Oversubscribed Fund II (BusinessWire) Socure Announces Strategic Growth Investment at $5.2B Valuation and Acquires Agentic Operations Platform Fravity (Socure) Listen: When Consumer Protection Disclosures Work Too Well (Consumer Finance Monitor Podcast) Colorado Sues EarnIn For “Immoral, Unethical, Oppressive and Unscrupulous” Business Practices When Andreessen Horowitz led EarnIn’s $39 million Series B in 2017, Alex Rampell, a partner at the storied venture capital firm, described EarnIn’s small-dollar advance product as “free,” relying on “entirely on voluntary contributions instead of fees.” Rampell favorably compared EarnIn’s approach to typical payday loans, writing, “Payday loans historically have had the potential to be a slippery slope for consumers into financial distress: opaque systems and steep fees that are hard to repay and set individuals back further than where they started. For many Americans living paycheck to paycheck, that kind of slide into debt can be extremely difficult to recover from.” But EarnIn’s product — which, the company says, isn’t a “loan,” legally speaking — ends up being anything but free for most users. While it is technically possible to take a no-fee advance from the company, most users incur “Lightning Speed” (instant funding) fees and/or ostensibly optional tips that combined can equate to annualized percent rates that can reach over 1,000%. With an average term of 9.77 days, even fees and tips that are small on an absolute dollar basis are equivalent to high APRs on an annualized basis. A lawsuit filed last week by Colorado Attorney General Phil Weiser against EarnIn describes the company and its product and business practices quite differently than Andreessen’s Rampell; “EarnIn’s business practices are unfair because they are immoral, unethical, oppressive and unscrupulous,” the complaint, filed in District Court in Denver, Colorado, says. EarnIn’s marquee product, Cash Out, isn’t a loan, the company says, as Cash Out is not “[f]orward [l]ooking” because “[t]he money is transferred based on earnings to date,” does not carry a mandatory fee to access funds, and is “non-recourse,” meaning EarnIn cannot pursue recovery from users who don’t repay advanced funds. EarnIn is sometimes referred to as an “Earned Wage Advance” service. Though, unlike some others in the broader category, EarnIn does not integrate with employers’ payroll or time and attendance systems. Rather, EarnIn attempts to verify a user’s income and employment, by examining direct deposit transaction data in their linked external bank account, by using GPS location and Bluetooth data to estimate a user’s time spent at work, if they work in a fixed location, and/or by having a user provided a work email address. While repayment is theoretically optional, as EarnIn would have no recourse, such as reporting non-payment to the credit bureaus or a collections agency, this is anything but clear to users of the EarnIn app, Colorado’s complaint argues. During the process of taking a Cash Out, users must agree to a preauthorized debit of the amount advanced, plus any expedited funding fee and/or tips. Though users technically can revoke this ACH authorization, the fact that they have this right isn’t presented to users during the Cash Out process, but rather is included in linked terms and conditions few users are likely to actually read. And users who do wish to revoke their ACH authorization must do so three or more days before the scheduled transaction date by emailing EarnIn’s customer support. Similarly, EarnIn’s marketing claims that it carries “no interest” and “no mandatory fees” is technically true, but is not the reality for many of the app’s users, the Colorado complaint argues. Further, EarnIn has extensively marketed to users the ability to “access your pay today,” to get “instant” funds, and to get funds “the same day you work” — when, in reality, this speed of access was only possible by paying the additional, undisclosed Lightning Speed fee. The Colorado complaint highlights this contradiction specifically, noting: EarnIn tells consumers that they can ‘Make any day payday with EarnIn’ and that there no ‘hidden’ or ‘mandatory fees.’ For a consumer to access their pay on the same day requested, they would have to pay the Lightning Speed fee. In the fine print of the advertisement EarnIn discloses that ‘[f]ees apply to use Lightning Speed,’ a direct contradiction to the larger text of the advertisement, and a fact that would not be immediately understood by a consumer as ‘Lightning Speed’ is not defined nor explained. Like other small-dollar lending services, rather than explicitly charge interest, EarnIn collects a markup on the optional Lighting Speed fee, which, given the nature of the transactions, many users opt for, as well as offering users the option to leave “tips.” EarnIn leverages either Real-Time Payments (RTP) rails or push-to-card via debit rails if a user opts for the optional Lightning Speed. But while these capabilities cost EarnIn, on average, $0.075 or $0.20 per transaction, respectively, EarnIn charges users significantly more. Initially, EarnIn charged as little as $1.99 for Cash Outs up to $24. But EarnIn has hiked these fees multiple times, and now charges $4.99 for Cash Outs up to $75 and $6.99 for those over $75. According to the Colorado complaint, EarnIn leverages “dark patterns” to manipulate users and to make it more difficult to avoid tipping. Examples of these “road blocks” highlighted in the complaint include: As of 2023, in the EarnIn app, a consumer attempting to get a loan under the default settings had to make at least eighteen separate taps to complete the transaction and reduce the tip to $0. There is an alternative method to leave a $0 tip, via the “Custom tip” feature, but EarnIn made this option difficult to access by visually deemphasizing it and positioning it near devices’ “home” button. Using language and imagery such as “Tip to pay it forward,” “Your generosity supports the service + helps,” “Your tips make a difference,” and “[t]he APR for this cash out is 0%. Your tips help support us.” While Lightning Speed and tipping are, in theory, both optional, in practice, users nearly always ended up paying something. According to the Colorado suit, users were charged either a tip or an expedite fee for more than 92% of transactions, with an average APR of nearly 388%. And, although the advances are “non-recourse,” EarnIn successfully collects on nearly all transactions — 99.18%, according to the complaint, which, Colorado argues, “demonstrat[es] that EarnIn’s advances function as high cost loans, not voluntary payments.” The share of EarnIn users incurring fees — despite the product being marketed as 0% APR — is hard to reconcile with Andreessen investor Rampell’s favorable description of EarnIn vs. “opaque systems and steep fees” associated with payday loans, which, Rampell says, can be “extremely difficult” for borrowers to extricate themselves from. According to Colorado’s suit, EarnIn’s Cash Out product “trapped many consumers in extreme cycles of high-cost reborrowing.” The suit gives an example of one consumer in the state, who took out a whopping 1,151 loans, paying a total of $4,038.50 in Lightning Speed expedited funding fees on loans that averaged the equivalent of 1,421% APR. Another Colorado user, the suit says, took out 1,033 loans, paying $8,561.22 in tips and fees at an average APR of 1,539% . These types of usage patterns demonstrate “the severe and repeated financial harm caused by EarnIn’s illegal lending model,” the Colorado complaint argues. In aggregate, from January 2023 through July 2025, EarnIn extended 3.1 million loans to 56,778 users in Colorado — an average of nearly 55 transactions per person during the time period. EarnIn lent more than $300 million and collected more than $16 million from Colorado users in Lightning Speed fees and in tips. Ultimately, the Colorado complaint argues that “EarnIn’s Cash Outs are loans under Colorado law, and EarnIn’s contention that its product is not a loan based on its disclaimer of any legal repayment obligation lacks any real-world significance given how the Cash Out transactions actually operate in practice.” Colorado further argues that “EarnIn was not working with companies here in providing consumers with funds but acted as a third-party lender and charged illegally high rates, used deceptive design strategies to extract some charges, and trapped consumers in repeat borrowing. Colorado will continue to stand up for consumers and hold companies accountable when they violate our credit laws or attempt to evade them through misleading practices.” Colorado’s suit alleges EarnIn assesses finance charge in excess of that permitted by Colorado law, that EarnIn failed to make legally required disclosures to Colorado borrowers, that EarnIn made supervised loans to consumers without obtaining the requisite license, that EarnIn violated the state’s Deferred Deposit Loan Act, that EarnIn engage in unfair and deceptive trade practices, and that EarnIn made false or misleading statements concerning price. The suit seeks a court order enjoining EarnIn from violating relevant Colorado laws, for EarnIn to refund amounts charged in excess of state law, fees, costs, and penalties, and other relief the court deems to be just. Fintech Business Weekly is made possible by the generous support of paying subscribers — bringing you independent analysis of banking, fintech, and crypto without fear or favor. You can support my work by becoming a paying subscriber if you aren’t already. Paying subscribers enjoy: access to the entire archive of nearly six years of newsletters extended versions of the weekly newsletter, with additional content and analysis and (for founding member tier) quarterly personal 1:1 fintech Q&A / strategy calls with me, tapping into my unparalleled knowledge of the intersection of banking and fintech and experience working in the sector, including helping to launch Goldman Sachs’ retail bank Marcus Support Fintech Business Weekly You can also support Fintech Business Weekly by sponsoring a newsletter or podcast, putting you in front of 93,000+ of the most influential decision makers in banking, fintech, and crypto. Learn more about sponsor opportunities or request a media kit by dropping me an email. Q2 Quarterly Banking Profile: Five Quick Takeaways The FDIC released its quarterly banking profile, which is always worth a read, for the second quarter. The profile provides an aggregate snapshot about the overall health of the 4,238 institutions whose deposits are insured by the FDIC, who, collectively, hold more than $26 trillion in assets and saw a total net income of more than $90 billion in the second quarter. Net income was up 12% vs. Q1 2026, with lower provisions and realized gains on securities helping to drive the improvement, somewhat offset by increases in non-interest expense and applicable income taxes. Overall net interest margin was basically flat quarter over quarter (up 1 basis point vs. Q1 2026), though community banks saw NIM grow by 10 basis points, to 3.81%. Unsurprisingly, given that interest rates haven’t changed, unrealized losses on securities changed little, edging up slightly to $326.7 billion vs. $325.1 billion in Q1 2026. Aggregate credit quality has remained benign, with asset quality actually improving slightly in Q2. Overall, institutions reported a 9 basis point decline in their past-due and non-accrual rate, to 1.44%. Net charge off rate also declined by 2 basis points, to 0.57%. And finally, the FDIC’s deposit insurance fund stands at $161.1 billion, making the DIF reserve ratio 1.48% — above the statutory minimum of 1.35%, but still below the designated target reserve ratio of 2%. [Paid Subscriber Exclusive] Nissan Withdraws ILC App, State Associations Plan BankChain Alliance, FDIC Defines “Unsafe or Unsound Practice” For First Time Automaker Nissan withdrew its application to charter a Utah industrial loan company and its corresponding application to the FDIC for deposit insurance last week, on August 19th. Nissan’s application had been outstanding for more than a year, as it originally applied in June 2025. Recent applications to the OCC to form national banks have been decisioned, one way or the other, substantially more quickly than has historically been the case.
The Eureka Machine: Richard Socher Breaks Down Science In The AI Age
Richard Socher’s The Eureka Machine argues agentic superintelligence could transform science through simulation, integration, experimentation, and discovery.
Plaud's new earphones come with an eSIM-enabled case for talking to AI agents
Plaud's new 'agentic' earbuds are priced at $249.
Perplexity launches a local AI agent with zero token costs
Perplexity has launched "Portable Computer" — a new agentic AI that runs locally on NVIDIA hardware and saves token costs.
Intel Crescent Island GPUs Pack Up To 32 Xe3P Cores, Optimized For Agentic AI With Low-Cost LPDDR5X That Reaches Up To 480 GB Capacity
Intel has given the rundown on its next-gen AI inference accelerator, Crescent Island, which packs 32 Xe3 cores and 480 GB of LPDDR5X memory.
Autonomy and Innovation
Incentives favor offense when it comes to agentic cybersecurity; it’s the same dynamic that will limit incumbents and fuel startups in the long run.
Why Agentic AI Could Expose the Leadership Gaps Holding Businesses Back
Agentic AI can boost productivity while exposing leadership gaps that require stronger guidance, accountability, and decision-making.
SAP Completes Dremio Acquisition to Expand Business Data Cloud for Agentic AI
SAP has completed its acquisition of Dremio, bringing Iceberg-native lakehouse, federated data access, and semantic capabilities into SAP Business Data Cloud to support enterprise analytics and agentic AI.
OpenAI targets heavy users with premium ChatGPT Business seats
Analysts say enterprise AI vendors are redesigning pricing models to better monetise high-intensity and agentic workloads.
How Cohere Health digitizes clinical policies using Amazon Bedrock AgentCore
In this post, you learn how Cohere Health built a multi-tenant agentic architecture on AgentCore using AgentCore Runtime’s secure MicroVM isolation, unified tool access through AgentCore Gateway, AgentCore Memory, and the Agent Skills open standard to rapidly scale policy digitization capabilities, while preserving transparency, version control, and human oversight.
Google Maps adds agentic features, including food ordering and hotel bookings
The launch of these new features reflects Google’s ambitions to transform Google Maps from a navigation tool into an assistant that's capable of helping users complete real-world tasks.
How to manage token costs while keeping agentic goals on track
Learn about AI tokenomics and how Dell AI Factory with NVIDIA helps reduce token costs by running AI agents locally.
The Real Reason AI Investments Fail To Transform The Business
Why AI strategies stall, and how leaders can improve AI ROI, redesign workflows and operating models, and turn agentic AI into lasting business transformation.
[Galaxy Unpacked July 2026] A First Look at Galaxy Z Fold8 Ultra, Galaxy Z Fold8 and Galaxy Z Flip8
At Galaxy Unpacked July 2026 in London, Samsung Electronics unveiled the eighth-generation Galaxy Z series. Combining agentic AI that understands users with
Microsoft launches its first cyber model and a new agentic cybersecurity system
Microsoft bolstered its AI cybersecurity offerings this week with the launch of its first AI security model and a new security platform.
How To Get Your Business Data Ready For AI Agents
Discover four practical steps for making business data ready for agentic AI while reducing risks around governance, compliance and accountability.
Dell Deskside Agentic AI cuts token costs by up to 87%
Dell, with Nvidia, introduces Deskside Agentic AI to manage token costs, enabling local trillion-parameter model runs and reducing cloud expenses.
AI web browsers 'aren't ready for the public,' scientists warn as they highlight massive security red flags
By enhancing the functionality of agentic browsers, they have become insecure and could be sharing personal information.
Google Is Expanding Access To Its Gemini Spark Agentic AI Assistant
Gemini Spark is now available to more users.
Agentic coding goes hands free as OpenAI brings GPT-Live's full duplex voice control to Codex and ChatGPT on the desktop
This architecture creates a pair-programming dynamic where developers talk through problems conversationally while agents execute tasks asynchronously.
Microsoft 2.5: New security business chief Hayete Gallot on the company’s push into the agentic era
Hayete Gallot, who rejoined Microsoft in February to lead its security business, is pushing the company toward "agentic security" — systems that spot threats and act on them without waiting for a human. She's doing it while building a new leadership team in a division where several VPs have exited in recent months.
Agentic AI in biomedical research: What is it and can it expedite science?
Researchers are venturing into a new area of collaboration that relies on agentic AI, co-scientists that can act independently and help humans ideate, hypothesize and speed up scientific exploration.
Bluehost Expands AI Business Platform with New Simple-to-Use AI & Agentic Solutions Purpose-Built for Small Business
Today, Bluehost, the all-in-one platform trusted by millions of small businesses to build, grow, and run online, today unveiled its next-generation AI business platform for small businesses, introducing intelligent AI agents that can build websites, run online stores, capture leads, answer customer questions, and book appointments, all without requiring technical expertise. The solutions are purpose-built for small business needs, giving entrepreneurs the capabilities of a much larger business w
Ray Veeraraghavan joins L7 Informatics as VP science and AI
Indian American health executive to advance agentic AI platform across scientific workflows at life sciences firm.
Amid hardware legal battle, OpenAI releases a $230 keyboard for Codex
OpenAI, which is in the middle of a legal battle with Apple over hardware trade theft allegations, just released a light-up keyboard designed to be paired with its agentic coding app.
OpenAI's first branded hardware is... a light-up keyboard?
The Codex Micro is designed to monitor multiple agentic threads at a glance.
How Clean, Accessible Data Powers Agentic AI Workflows and Better Business Decisions
Over the past few months, conversations with customers, partners, and peers — including at the recent Databricks Data + AI Summit — have surfaced a consistent set of shifts happening across the analytics and data landscape.
Business Value of AI Is Spiking, Driven by Increased Adoption and Agentic Expectations, SAP Finds
A new SAP and Oxford Economics study shows businesses are increasingly driving value from AI, even as challenges continue to accrue.
Meta's Chief Data Officer Says Agentic Commerce is the "Next Tier of Business"
Alex Schultz says stablecoins are assumed inside Meta. The harder problem is getting the rest of the world there.
OpenAI is shutting down Atlas, but its AI browser ambitions are still growing
OpenAI is sunsetting its AI-powered browser after less than a year. But it's moving some agentic browsing features to its desktop app and a Chrome extension.
Genpact Partners with Nestlé Business Solutions to Establish New Global Capability Center
/PRNewswire/ -- Genpact (NYSE: G), an agentic and advanced technology solutions company, today announced a partnership with Nestlé Business Solutions (NBS),...
Security firm finds 'first documented' AI agentic ransomware attack
Researchers at Sysdig say they found "a warning sign" of where ransomware attacks are heading.
With ERP.net’s New App Builder, Businesses Can Create Applications Directly Inside Their ERP System
Bulgaria, July 3, 2026 -- European enterprise software company ERP.net has introduced App Builder - a new Agentic AI-native platform that enabl...
Gemini Spark, Google's agentic assistant, is now available on Mac
Google's 24/7 agentic assistant, Gemini Spark, comes to Mac alongside other improvements, like real-time tracking and support for more apps.
Content Independence Day, one year on: building the business model for the agentic Internet
One year after declaring Content Independence Day, a dynamic market for monetized content has officially emerged. In this report, we examine how the rise of autonomous AI agents is upending traditional search referrals and detail the new infrastructure required to support a sustainable web economy.
Gemini Spark Comes To Google's Gemini App For macOS
The Gemini macOS app now has access to Google’s Spark agentic AI assistant.
OpenClaw is finally available on Android and iOS
The free open source agentic program is finally invading your phone.
There's Now An OpenClaw App For iOS And Android Phones
Smartphones are welcoming the agentic AI overlords.
Cursor releases iPhone and iPad app following recent acquisition by SpaceX
Earlier this month, SpaceX (which includes xAI) purchased the agentic coding firm Cursor. Today, Cursor has released its first iOS...
Microsoft Discovery Aims To Advance The Era Of Agentic Science
The cloud-based agentic AI platform aims to help human researchers overcome resource constraints and complex data challenges to accelerate science at scale.