20VC: Taking on NVIDIA: Fractile CEO on the Future of Chips | What Everyone Gets Wrong About Capex Spending | The Winners and Losers in Neo Clouds with Walter Goodwin

Harry Stebbings: We're so traumatized by the dot-com bubble that we've priced in exuberance right from the outset. We will be spending far more on running models in 10 years' time than we are today. It feels like we need thousands of times more compute than we have today. Frontier intelligence is not yet commoditizing. I think we'll have a couple of multi-hundred-billion-dollar companies built in that space. One neocloud that is competent is much like another neocloud that is competent.

Harry Stebbings: This is 20VC with me, Harry Stebbings. Now, joining me in the hot seat today is one of the leading minds when it comes to Frontier Chips. Walter Goodwin, founder and CEO of Fractile. The business has just raised a $720 million Series C led by Thrive, Founders Fund, Lightspeed and Redpoint, valuing the company at a $7 billion price. As I said, Walter is one of the leading minds in the future of chips, and we get into that and so much more in the discussion today.

[Ad break for 8sleep]

Harry Stebbings: But before we dive into the show today, founders and investors sleep on 8sleep. It is the intelligent sleep system trusted by Mr. Mark Zuckerberg, Dr. Andrew Huberman, Charles Leclerc, and the Aston Martin F1 team. The Pod is a smart mattress cover that fits on any bed. It cools each side to 55 degrees, heats to 110 degrees. It even tracks your heart rate, HRV, and sleep stages without a wearable, thank God. The new Pod 6 has a hub 50% smaller than the last one and starts at just $1,999. Visit 8sleep.com/20VC. That's E-I-G-H-T-S-L-E-E-P dot com slash 20VC.

[Ad break for MongoDB]

Harry Stebbings: While 8sleep helps you sleep better, MongoDB helps you build better. MongoDB has always been the database developers love. Well, now is the data platform AI agents need. Agents need accurate context, fast. MongoDB stores, searches, and reasons over your data in real time. JSON-native database, vector search, and Voyage AI embeddings all in one place. One system instead of 10. No data pipelines to maintain. Build and scale from your first user to billions of vectors. Run on any cloud, on-prem, or your laptop even. That's why 75% of the Fortune 100 run their most critical apps on MongoDB, moving trillions of dollars every single day. And it's why AI-native companies like Eleven Labs run 40 million agents on MongoDB. So if you're building an AI, MongoDB for Startups helps you move faster with Atlas and Voyage AI credits and dedicated support. Don't build agents that answer once and forget. Build agents that remember and learn from your real-time data. Go to mongodb.com/agents. That's m-o-n-g-o-d-b.com/agents.

[Ad break for Framer]

Harry Stebbings: While MongoDB scales the product, Framer scales the story. So if your team wants a website that looks and feels handcrafted, but is still fast to ship, Framer is built for exactly that. Framer is the AI website builder that helps creators, teams, and businesses ship production-ready sites faster than ever while getting every detail right. Prompt, inspect, edit, and publish in one place at a whole new pace. Agents and humans work in tandem. Agents bring speed and scale. You bring taste, judgment, and control. The work lands on the canvas and stays editable. Build custom code components, manage CMS content, optimize SEO, and audit for issues all in one place. Enterprise grade hosting, security and 99.99% uptime SLAs, trusted by leading brands like Perplexity and Miro. Learn how you can get more out of your site from a Framer specialist or get started building for free today at Framer.com/20VC for 30% off a Framer Pro annual plan. That's Framer.com/20VC. Rules and restrictions may apply.

Harry Stebbings: You have now arrived at your destination. Walter, I am so excited for this, dude. It's lovely to be able to do it in person. And I'm so thrilled that you could join me in the studio, man.

Walter Goodwin: Very good to be here with you, Harry.

Harry Stebbings: Now, I want to start, for those that don't know, 60 seconds, like what is Fractile? What do we need to know?

Walter Goodwin: We're a frontier AI company of sorts. But we make chips. We make these big rack scale systems that are fully focused on advancing the frontier of AI capabilities. From a team composition point of view, this means that we have ML researchers down to process experts that fly out to Taiwan every few weeks, down to people that actually look at the thermals and the mechanicals of putting together one of these rack scale systems.

Harry Stebbings: I was listening to you on NoPrize, which is talking about Sarah, and you were saying unequivocally that you were a chip company. And I was listening to this thinking, great. But then I'm looking at it going, everyone's moving full stack. We're seeing OpenAI have Jalapeno. We're seeing DeepSeek have their own. We're seeing Anthropic build their own. Obviously, Meta and Mark has his own. My question is, is the future won by vertical integration? And can you be a chip company alone these days?

Walter Goodwin: It's a very good question. I think there's a verticalization in that as all of these different components come together, and especially as you look into the future and you work out what are these models going to look like, how do the workloads evolve, how does the silicon evolve, there is this absolute necessity to have a vision on every part of that. So Fractile is a chip company, but in a sense, we're also a Frontier AI company. What that means for us is we monetize by selling chips. We monetize by selling racks of compute to our customers. But it's completely essential that we have a bunch of folks internally who are ML researchers, who are looking at how model architectures might evolve, who are maybe even running smaller scale experiments than the ones that you would see in the leading labs.

Walter Goodwin: I think there is a distinction here between the expertise that is needed internally to win, and then the nature of the business, what the business model is. We never intend to compete with our customers. We will never monetize internal models because that's how we've chosen to build this company. We've chosen to compete in the chip layer. But you can't step back from challenging yourself to have expertise in all of these other places because that would mean you'd place the wrong bets. You have to synthesize everything that you're learning across the ecosystem. And you have to bring that back to be obsessive about the part of this space that you're building. If you do this right, you've ideally found yourself an enormous piece of green space where you're incredibly helpful to your partners, to your customers, and to your suppliers upstream. You've got to know parts of their business as well. But of course, you never seek to conflict with them.

Harry Stebbings: Would you argue that everyone's moving to vertical integration because everyone's fighting against commoditization of one layer? If you think about what OpenAI did with Jalapeno in nine months, it suggests some form of commoditization of the layer. Do you think everyone is desperately expanding to run against commoditization?

Walter Goodwin: I think there's definitely the fight to keep your margins expanded. If you are serving frontier models, it is important that you're not squeezed all the way down by very high margins on incumbent chip players. That's where it's very useful to have a first party play or another third party play that can be a counterpoint to that. I think there is this confluence of forces that mean that if you're a frontier lab, if you're a hyperscaler, it's very important to have your own retort to various parts of this. I do also think that there's this pretty crucial distinction today still between what is perhaps becoming commoditized and what we still see at the frontier.

Harry Stebbings: What is commoditized? Help me, I'm an ambassador. I look for value in amazing entrepreneurs and businesses. What is commoditized that most people think isn't and what isn't that maybe people think is?

Walter Goodwin: That's a great question. There is this sort of sometimes concern, sometimes excitement about the idea of intelligence saturation. As models get smarter and smarter, there will become a kind of waterline that they exceed where we don't really need greater intelligence for a whole set of tasks. I think it is true that there are certain tasks for which we see this kind of intelligence saturation. This is where you then see an open model is good enough and these models become commoditized at that layer. Where that becomes more nuanced, though, and where models even at that level still fight against commoditization, is that there is always economic pressure to reduce the cost of intelligence. Perhaps counterintuitively, this is why I think the frontier labs and the frontier efforts remain so vital. Frontier intelligence is not yet commoditizing. We see large margins there. We see big deltas against everybody else. These learnings from the frontier trickle down into building better smaller models as well. If you look at that fight across the entire gamut of frontier LLM capability, as well as behind the frontier LLM capability, there's still this ferocious battle. I would characterize that as not commoditized in so far as if you are a market leader in something, you can still extract high margins.

Walter Goodwin: A good example of this is DeepSeek released a few days ago their entire inference stack, their entire software stack. It had really remarkable results. People very quickly characterized it and realized that DeepSeek had done so well in building an inference stack to run their models on GPUs that they were getting much higher throughputs, much better interactivity, better speeds than everybody else who had built a DeepSeek inference server. Something that you might have thought would have been by this point commoditized, which is the serving of an open model. In their case, by going full stack, they built a model which was more efficient and they built an inference server for that model that was more efficient. They carved themselves out the right to have really healthy margins on that. Even in this open source space, we don't yet see this kind of complete race to the bottom, at least for those folks that try to own a lot of vertical understanding up the space.

Walter Goodwin: I think what is commoditizing, and I don't think anybody has ever argued it won't be, is the whole space of cloud compute. One neocloud that is competent is much like another neocloud that is competent. The reason it's so exciting today is we don't yet know which ones are competent. I think there's still a lot of space there for there to be huge winners and maybe huge losers as well. I think that's why you see all the neoclouds desperately trying to vertically integrate too, because they're very aware of the commoditization of their own layer.

Harry Stebbings: We had the founder of Nebius on the show who was talking about the movement into model routing and inference for them as a core business and a core strategic, just because I think they're aware of that.

Walter Goodwin: I think that's true. You definitely see every neocloud is trying to build an inference server. A lot of these folks who built an advantage in inference serving, which is a software advantage ultimately, are also thinking about how they can have more control of their destiny on the compute side. Obviously, I'm biased, but I think this is where the neocloud story and the differentiated silicon can actually become a really core differentiator for those players as well.

Harry Stebbings: Can you unpack that? Why is that?

Walter Goodwin: We're still in the very early innings of the rollout of silicon that is not the mainstream conventional GPUs and TPUs and other XPUs. It is still much bolder, especially if you look at the way that these build-outs get financed, to lean into a new chip platform like Fractile, a Cerebras, and so on. We need these new chip platforms because workloads are becoming more and more complex and intense. The specialization of these chips will allow for the execution of these workloads to be much more efficient. These new platforms have a belief that in certain workloads you have this kind of more than 10x advantage, whether it's on cost or in our case, cost and speed. In a sense, they are contrarian bets still. We still don't see many gigawatts of new silicon built out, but they do have this potential to fundamentally alter the economics. If you imagine you are a neocloud and you're the first neocloud to work out how to break out of a particular bond with a particular GPU vendor, potentially you get to reap this enormous benefit. We've seen this with FluidStack. They were, in some ways, the first neocloud to decide that they were going to step bodily away from NVIDIA. They deployed a bunch of Google TPUs. They've had an extraordinary year as they crossed that Rubicon and earned the right to start deploying other chips. I think there are opportunities to, as you go more vertical, also take some of these slightly higher risk bets and reap huge benefits from that.

Harry Stebbings: What does that do to the NVIDIA business? If we think about NVIDIA as a horizontal provider and you as a specialized provider, is the future of chips specialized according to workloads? How do we think about that market composition, horizontal versus specialist?

Walter Goodwin: The way that I think about this: NVIDIA is an extraordinary business and they build by many metrics the best AI chips in the world. The best systems, and they have extraordinary work on data center scale inference. In the inference space, I think they've done some brilliant work as well. What I think is different about this moment in time and core to Fractile's identity is this idea of an evolving frontier. The frontier necessitates slightly different bets and a slightly different way of thinking, because you are necessarily setting up a business that is able to look around corners, is able to make bold predictions, and is then able to steer the entire tech product towards those bold predictions. If you look at NVIDIA's business, what they excel at is delivering high quality products at extraordinary volumes that are, in a sense, unopinionated. These are powerful general purpose processors that are very good at various parts of the AI compute world. If you look at where Fractile's bet is on inference, which is building chips that are orders of magnitude faster than these GPUs, that's a particular focus. For us architecturally, it's meant this focus on finding paths to huge memory bandwidth. It's potentially taken us away from some of that complete generality.

Walter Goodwin: What's interesting today, and the reason we're so excited about this space, is the frontier is also an overwhelming portion of the total market. If you look at what's happened just over the last few years, we have this enormous build out of specifically inference compute to serve LLMs at the frontier. What is in some sense specialist is also by no means a niche. I think that's this real transformation. If you look back 10 years, even five years, it would have been slightly crazy to build a specialist chip. We didn't have any single workload that made up more than, let's say, 10% of the total compute burden. Suddenly we do. I think that demands slightly different company structures that can have that kind of predictive power built in that try to orchestrate the entire org around what it means to chase after that.

Harry Stebbings: You mentioned memory and HBM there. Why is the memory wall such a prominent problem still today?

Walter Goodwin: It's one of these interesting facts about chip physics. If you look at the two things...

Harry Stebbings: All great days start with that line.

Walter Goodwin: [Laughs] So if you look across the last 20 years in the silicon world, there are two things that really make up the top level specs, the top trumps of your chip. One of them, certainly for AI chips, is the flop count. How many floating point operations can you do per second? That's your mathematical power on that chip. The other piece is the properties of the memory on that chip, the capacity, and in particular, the bandwidth to that memory. That's how quickly you can shuttle data back and forth. If you look at a GPU, you have that silicon in the middle that is running the maths with that high flop count. Then there are other pieces of silicon around it, the HBM, that are coming in on this highway, bringing data in. We've been extraordinarily good with Moore's law at scaling flop count. We've scaled it about a million fold in the last 25 years on these chips. Bandwidth, by contrast, is much more intrinsically physics limited. We've scaled it about 40 fold in the same time. It's lagged behind the flop scaling about 25,000 times. We naturally found ourselves more and more bandwidth limited in a lot of the operations that we do. For something like LLM inference, this has become a critical bottleneck because of this interplay between the generation of text, which is this auto-regressive, sequential thing. One word after another is generated out of these models. Every time I go and generate one of those words, I'm loading from that memory over that limited bandwidth all of the weights of my model. Today, also all of the context that I'm operating on as well. All of the past states of things that I've seen. That is my bottleneck on generation speed. The reason that if you take a very large frontier model, it might generate at 80 tokens per second, 80 words per second, is going to be not because of the limited maths capability on the chip, but because you don't have nearly enough bandwidth to that memory. I cannot load the weights of the model from memory into that processor more than, let's say, 80 times per second. The reason I started with the sexy line about physics here is if you zoom in at the interface, there's a bunch of essentially copper wires that are going from this DRAM memory into the processor. We've been pretty good at making them smaller and smaller, but there's only so much you can scale that. You have, let's say, a few thousand wires.

Harry Stebbings: Is that why... sorry, you have SRAM. Am I right in saying that moving memory closer to it?

Walter Goodwin: So SRAM, yeah. There's this sort of... DRAM is like often a different building, basically. You're carrying these boxes down the stairs and then up the stairs again into the processor building where the maths is done. You have to carry them back again, unfortunately, because there's not enough space in that building to store all the boxes. Maybe it's a warehouse and then you've got the office where you're doing the work. SRAM is like having a bunch of boxes that are permanently in the processor. It's a small amount of memory that you cram onto the same silicon where you're doing your processing. But that's the snag as well. The reason that over the last 30 years, we've externalized memory into this thing called DRAM and put it on different silicon is that these are built in completely different foundries. They've been specialized for very different things. DRAM is completely focused on density. The DRAM bit cell is an incredibly simple structure. You can cram many billions of these into a very small area of silicon. SRAM, by contrast, is built conveniently on the same exact process nodes. It's built on the silicon that you're building your processor by TSMC, for example, but it isn't optimized for density. You might have less than 10 times as much SRAM capacity per unit area as you can get gigabytes of DRAM on that same area. By extension per dollar as well, because your silicon wafer costs you a certain amount.

Harry Stebbings: When you look at inference and the future of inference, which is then best suited? Is it DRAM or SRAM? Because latency is important.

Walter Goodwin: That's right. Latency and bandwidth as well. How much can I just cram through this? This is one of the things that we've been working on over the past couple of years at Fractile: the rejection of this trade-off. SRAM as a memory is the only one today that can be built on the same processor. It's the only one that can be incredibly close and have incredibly high bandwidth. DRAM, much cheaper, much denser, much more capacity, but fundamentally limited in the way that we access it. If you look at that physics, the fact that it's coming in over wires, it's because we've got it in the wrong place. We've got the DRAM sitting in plane with the chip and we've got this limited... it's called a beachfront. It's literally the limited real estate on the edge of the chip. What we've been working on at Fractile over the past couple of years is essentially stacking that memory on top of the logic. Looking for ways to, rather than have this limited beachfront, everybody gets to have a seaside property, basically. The entire memory layer sits on top of that logic die.

Harry Stebbings: How is that possible?

Walter Goodwin: 3D in the chip world is a dimension that people have been scaling for a long time, but it's got all of these complexities. If you look at an HBM, a high bandwidth memory, it is a stack of DRAM chips, but it's a stack that has certain properties that don't make it amenable to just being taken and plonked on top of your logic die. We don't do that today. But there are these process technologies. This is something that Fractile's team has a long hinterland in around 3D stacking.

Harry Stebbings: So you're building condos on the beach instead of beach houses to allow for more capacity.

Walter Goodwin: I guess we're building one of those new built archipelagos where everybody has a bit of beachfront. Strips of land. Everybody's right next to a piece of coast. When you put this on top, suddenly you have this whole two-dimensional area where you can have essentially those copper wires are now not going laterally. You have millions of them coming straight down.

Harry Stebbings: When you look at differentiation between other providers, is that your differentiating factor? And how do you analyze the specialized providers that also compete with you in this from like an Etched? What is the differentiation?

Walter Goodwin: I think there's a few ways to look at this. One is philosophical, and the other is technology today. Philosophically, I do think that Fractile has a mission. The mission is to advance Frontier AI capability through building better hardware. Today, that's taken us over the past couple of years to this particular solution. I think this extraordinary work that the team has done in unlocking a future where you can have the bandwidth of SRAM, but with the capacity and the scalability of DRAM, it is a world-changing innovation. It's going to have longevity as well. We're very enthused about taking this forward over multiple generations and compounding a technology advantage there. At the same time, with a mission, if the workload changes, you also want to be able to change your response. One thing which we're cautious not to become is the company that cracked 3D DRAM. We want to be challenged to continue to innovate on whatever comes next.

Walter Goodwin: If you look at mechanistically what it does mean for us over the next few years, there is this challenge today in the chip space where on the one hand you have all of these HBM based chips. At some level, it's very hard for them to diversify, whether it's the GPUs, TPUs, the Jalapeno chip from OpenAI, Etched, MatX. They have often really brilliant architectures on the chip which allow them to do things slightly more efficiently than one another. But they are fundamentally boxed in by the top trump specs. You can get only so many flops and you can get only so much bandwidth. The HBM that sits off to the side of them is exactly the same piece of silicon. It's the same module from the same vendors for each and every one of those chips. On the other side, you've got these SRAM-based chips that have high memory bandwidth, but also have this huge capacity constraint. If you look at Groq or you look at Cerebras, what they can do today, which is extraordinary, is allow you to run powerful models much, much faster. One of the things that I've always been excited about with Fractile, it was what I set the company up to do four years ago, is build chips that are better at LLM inference, but in particular, build chips that allow us to run that inference much, much faster because we always believed that we would get to this space where we had what we actually do have today: test time compute scaling. Reasoning models, long running agents that run for longer and longer to get better and better at what they're doing, a better answer for you. This is the profile of all of the frontier workload that's changing the world today.

Walter Goodwin: If you look at what happens when you turn a frontier LLM to the problem of solving a Millennium Prize problem, it will go off and generate many millions of tokens sequentially. You scale up this swarm of them as well. There's a parallel element of this. But the many, many millions of sequential token generation means you're waiting days and days to get one of these responses. It's a space which somewhat tantalizingly hasn't been cracked at all by any of the silicon, any of the chips that we have today. There's this tantalizing mismatch between the fast inference chips, the SRAM-based chips that have the ability to access memory quickly, generate tokens quickly, and the needs of this new workload, the long-running agent, which needs a huge amount of capacity because you have these really long contexts that are emerging. The world urgently needs this kind of up and to the right thing where you have simultaneously really high bandwidth and high speed, but also high capacity. The ability to scale to longer contexts and larger models.

Harry Stebbings: Is this one of those where it always takes longer than you think? Or is this one where you see dramatic inflection points in the acceleration of technology?

Walter Goodwin: I think what's exciting about this one is, and this is true often of technological breakthroughs, you are always standing on the shoulders of giants. In our case, some of the technology that we're leveraging to do this 3D integration has actually been around for quite a long time. It's just not been turned to this exact problem, to the memory stacking challenge. It's a space where if you're willing to do the architectural effort and redesign your chips to make use of this, and you're also willing to do the manufacturing operations legwork, and you're able to persuade your partners that you're not crazy along the way, there is actually the possibility here for these kind of overnight transformations.

Harry Stebbings: Can I be blunt? I was talking to a founder of a brilliant chip company last night, and he said there's a huge amount of hot air and BS in the industry, and there's a lot of marketing fluff, but there's not much actually in production. When you look at your Etched or your Fractiles, are we actually in production? How do we actually think about on the ground being in production? And is he right with that statement?

Walter Goodwin: It's a great sentiment. There's two parts of this. Yes, Fractile has silicon that's gone through the foundries and has come back to us. Yes, we're running further trial runs and we'll be doing full product tape outs in a few months as well, which is quite exciting. I think the scale of production is something which is very much worth challenging, right?

Walter Goodwin: There is a real challenge, especially for new chip entrants at the frontier in accessing TSMC capacity and HBM capacity. In particular, for some of these efforts, there is a race for exactly the same resource. A lot of these chips today are built with the same backend partner. For example, whether it's the TPU or Jalapeno, MTIA from Meta, Matex or Etched, all of those are rumored to be using Broadcom. Same or very similar TSMC process nodes, same HBM technology. So you can see why there might be a real gold rush for that particular finite resource. That means it can be challenging to get to really meaningful scales in a short space of time.

One of the things we've always been quite focused on at Fractile is that we've had this sort of unfair reach into the stack. The point of a backend partner is they do the last mile, the physical design. They interact with the foundries, with the manufacturing side of the business. Fractile has always had that in-house. We have this extraordinary team that has always worked that way, even before Fractile, for decades in how they've delivered their chips. It has become a core strategic advantage for us because it means we've always had direct communion with our manufacturing partners. We've had to do that legwork from the outset.

Harry Stebbings: Just so I understand, the alternative is you go through Broadcom who then have the manufacturing partner relationship.

Walter Goodwin: That's right, yeah.

Harry Stebbings: Got you. So Broadcom is the intermediary layer that most people work with that you don't.

Walter Goodwin: And that we don't.

Harry Stebbings: What does that enable you to do then?

Walter Goodwin: One thing is build completely different building blocks for these chips. If you look at a Broadcom, I was at a conference this summer where Charlie Kawas, who's one of the leaders there, held up a Lego model of the next gen platform. He said this is going to be in production with six major customers in 2028, which is a strange thing to say, because what he means is something that looks identical to this is going to be made with six of the major players and given different names.

Harry Stebbings: It will be the TPU, MTIA, Jalapeno. Dorito 2.

Walter Goodwin: Exactly. So it reveals something about the structure of this market. Each of those chips will have a different name. They will also be different. There are incredible chip teams inside all of those companies that come up with insanely clever architectural optimizations. But in the end, it's going to be exactly that same platform. It's going to be the same stack of HBMs, the same number of chips. A lot of the fundamental building blocks are dictated by the backend partner because it's a platform technology play. You get to change the color of the paint, basically.

For us, what it has allowed us to do is maneuver and build these new building blocks. The memory stacking work is incredibly complex and painful and requires convening multiple, much larger supply chain partners to get them to do something that is highly custom. But once you crack through that, as we have, you then have an asset that nobody else has.

This was a slightly long-winded way of getting to your question about volume production. That also more likely than not has meant that you've gone and found a way to use resources that others haven't. In our case, a different process node, maybe a slightly different DRAM node or DRAM generation than might be going into Frontier HBM. You sort of find the opportunity to build gigawatts of capacity down the back of the sofa with things that other people and this set of Frontier platforms have already moved away from.

Harry Stebbings: Can I ask you, when we think about the production line, a lot of people told me that I need to ask about Tape Out. What is Tape Out? What do we need to know? And how do you think about it?

Walter Goodwin: Tape out is the moment where the final file from your chip design process, which is basically an enormous image we call a GDS2 file, gets sent to TSMC or to your logic foundry partner. The image dictates where all of the copper traces are, where the transistors are, where the logic is. They commence the process of building the mask sets, the stencils that will ultimately produce the chip. Then you start doing your first trial runs, your first production runs as well, potentially.

It is a moment in time that is a seminal handoff for a chip designer. At the end of this long process of architecture, front end design, verification, physical design layout, you send the file. If there is a bug in it at this point, it's too late. The mask is getting printed. That's why it's a major date on the calendar.

It is also just step one in quite a long process. That silicon then goes through the fab and you wait months for it to come back. You have to be poised and ready to receive that and run your bring up process. When the package silicon returns to you in the lab, you then sprint to get a workload running on a single chip, on a tray, eventually on a rack.

At Fractile, we track progress. There are tape out dates in the middle of it, but the thing we are continually shooting for, which we call Task Zero, is the day that a particular reference workload runs on an entire rack of Fractile chips.

Harry Stebbings: What does no one see about this process that is most prominent to you? If we think about tape out being a seminal handoff, where is the difficulty, the momentousness of the post tape out date?

Walter Goodwin: I think that's it actually. It is what comes next. There is an awful lot that you can fail to prepare yourself for if you are ultra fixated on that one thing, the handoff of the first chip to the fab. For us, it's continually tracking what it is we are actually trying to build here. We are trying to build systems that yield well, so the underlying economics are great because we've designed silicon that has the right levels of repair and redundancy. It is highly manufacturable. It makes a lot of sense as a product and then fits into a system that again yields well, is reliable.

Building a product that is manufacturable in volume, where you can then find a path to ramping as quickly as possible, that is the goal. Tape Out is a... I almost feel like over the last few years, there has been this accidental shared obsession between chip startup founders and the venture capitalists backing them on this phrase, because it's such a tangible moment, such a tangible date. But a lot of the work we do at Fractile, all of the work on building software, all of the work on systems, and a lot of the silicon team as well, is focused on those parts of the chip that are intangible, ineffable, but are about what it looks like to be able to get this stuff up and running and reliable and to volume as quickly as possible.

Harry Stebbings: When we were downstairs, you said that silicon may no longer be the bottleneck in the same way that it was before, and maybe data centers were or data center capacity was. Can you talk to me about that relationship and the transition of shortage versus not?

Walter Goodwin: This is why the last four years have been so exhilarating. It feels like every quarter, the thing that is most constraining our ability to run more frontier models at scale in the data center has shifted. There are these underlying properties that we urgently need, like the ability to run inference much faster, the ability to run at longer contexts with better economics.

If you think back three years ago, there was no question of other chips, it was an NVIDIA GPU shortage. Then we found ourselves in an underlying silicon and supply chain shortage, acute memory shortages for HBM which stretch years into the future, logic silicon shortages with TSMC as well. If you look at the space today, land and power is definitely another limitation. We've had all of the low-hanging fruit for industrial sites that can be retrofitted to be data centers, places where in the US you can get permitting to bring gas turbines close to spin these things up in 90 days. Those one-time tricks have all been exhausted. We are now exploring the longer tail of harder ways to find land and power to install these chips.

On the one hand, it is data center capacity that is our fundamental bottleneck on how much compute we can build out. If you want to be completely agnostic and invest in your long AI bet, invest in energy, because that is obviously the most fundamental. But at the same time, it is precisely that constraining limit that is why it is so vital that we have the best possible chips for any given job.

If you look at Frontier Labs and Hyperscalers today that have a fast mode on their models, if you are running that on an HBM based chip, you are doing so at an extraordinarily inefficient point on a tradeoff curve. On these chips, you can either be slow and efficient in generating tokens, or fast but deeply inefficient, far fewer tokens per second per megawatt of power.

What happens if I want to take a multi-trillion parameter model and run it at thousands of tokens per second, but I have a 500 megawatt data center and I don't have enough aggregate throughput? I can't serve my user base on those chips. There is a real pressure to install a set of chips that do retain really good aggregate throughputs, even as you squeeze them to much higher speeds. Fractile would be an example of that. So although it is land and power that may be the most acute bottleneck, even ahead of supply chain constraints, it is exactly that which incentivizes chips that are much more fit for purpose.

Harry Stebbings: How much of an energy deficit do we have? I know it's an incredibly macro and bold question. But when you look at energy requirements versus energy supply, what do you think the deficit looks like?

Walter Goodwin: When you look at backlogs in the US and the UK, the queues for grid connection are an order of magnitude more than there is energy in the grid. The extent to which all of those are legitimate projects that would get off the ground is something we can question, but it does look like there is this staggering level of demand. Energy is one of these real fundamentals.

Going back to what we said earlier about DeepSeek and their incredible innovation on their inference stack, when you look at China's policy and regulatory environment towards energy, it is green light anything. Including moving millions of people out of their homes because they need the sites for energy provisioning. In the US, you have NIMBYism. In the UK, you have bats and butterflies preventing sites of scientific or natural beauty.

Harry Stebbings: Are we not just inherently disadvantaged in a world of energy requirements versus China?

Walter Goodwin: I think this is one of the places where China has stolen the march on the West. In particular, these mega projects, huge hydro projects, for example. They are ideal for AI compute build-outs because they have this particular signature where it might have once been inconvenient to have 15 gigawatts of power all arriving at a single site that then needs to be distributed. An AI data center doesn't really care where it is located. So why not go put 15 gigawatts of AI compute right next to this hydro plant in the middle of nowhere?

In the US and in particular in the UK and in Europe, there is a lot that we can do to look a little more like that. The good news here, and this is something the government in particular should be doing more to play up, is that a lot of the power sources that could be brought to bear on this are renewables. They can be a huge part of this story. You can locate a data center in Scotland next to an enormous wind farm where previously there just would not have been a story as to why you wanted to have energy generation there.

Harry Stebbings: Do you know what's really interesting for me? We often do a lot of Instagram now in short form. The minute we talk about data centers and them not happening, we did it the other day with Microsoft pulling out of the N-Scale project. Everyone goes, "Good riddance, good riddance, you tech bros, good riddance." It was astonishing to see the universal backlash from bluntly moronic civilians.

Can I ask you, you saw inference as a market before so many other people, which I'd listened to years ago. The thing that I know now that I wish I'd known 12 months ago is just the infinite demand for inference. What do you know now about the market for inference that you think isn't that obvious or you hear other people say?

Walter Goodwin: I think the thing that remains is an even more pronounced version of my original bet. Four years ago, the insight was twofold. One is as you start to get these foundation models that have been trained on everything and then generalized to anything you might turn them to, you have this hard logic where training becomes a fixed cost and the deployment, inference, is just the use of these models to do useful things. It is the only part that matters. If you were spending more on training in the long run, why are we even building these things? You could pack up and go home.

I always describe it to my mother like the Olympics. You train in the gym. Great. And then you decide you are not going to actually do the race.

Harry Stebbings: What's the point?

Walter Goodwin: Exactly. This is why we are here. Imagine you spent far more money building a factory than you ever actually recouped from generating something out of it. The hard logic of building something with a high fixed cost is that you must believe you are going to be doing enough deployment to justify it.

That was part one. Then this bet that we would have what we call test time compute scaling, the advent of reasoning models. They didn't arrive until two years later, for which I am very grateful to Noam Brown and the team at OpenAI that was the first to publish one of these.

What is happening today is we continue to see this vector of test time compute scaling laws. It is this stack of different ways that we pursue greater capability. One is 10xing and 10xing the amount of compute we use to train these models. But the other today is how do we pour more compute into deploying these models into the inference side? It is a huge part of how you get to greater capability.

If you look at the example of solving a Millennium Prize problem, that Navier Stokes solution was 80 hours of execution from 10,000 agents in a swarm. You have these two axes that trace out this enormous volume of token generation, where you have 10,000 parallel agents all generating their own chains of thought, and they are running for 80 hours. It is so different to the intuition we might have about an LLM, where we think we ask a question and the single threaded thing will generate almost immediately an answer.

When we bring these models to bear on the hardest problems, and this is a bit of good news, they are also the ones that have by far the best, most positive societal impact. Generally, it is solving a science problem or something in medicine that we haven't been able to solve before. To do this where you have this incredibly economically and societally beneficial thing to solve, which is beyond the reach of just applying the model and asking it one shot, inference compute is the axis that we have. The Navier Stokes example is evidence of that.

Harry Stebbings: What do you mean inference compute is the axis that we have?

Walter Goodwin: It is the knob that we have to turn to get greater capability. Imagine that at any moment in time, we have in the world the best frontier model. It will come from one or other lab at any point in time. The best frontier model is extraordinarily capable. You talk to it and it's smarter than any you've spoken to before. But it turns out when you ask it, "I am working on this challenge. We are really trying to find a cure for this rare disease," we are not at a place where it is just going to say, "Oh, here it is. I've thought of this. In my neurons, this has fired out immediately."

The thing we also discover is if you do throw enough compute at it, if you say, "How about rather than just trying to give me an answer, you propose many hundreds of candidate potential solutions that you could go and explore. You create multiple copies of yourself. You go and traverse the internet. You read the literature. You debate with copies of yourself and you exchange information. Come back to me in a week's time, once you've generated billions of tokens in pursuit of finding an answer." Then, a little bit like if we gave a bunch of smart people a lot of time to solve a problem, you do have a chance that you've come back with a solution. This is the key. It is what every single application of a frontier model to solve a hard problem is going to look like.

Harry Stebbings: My question to you is, I am a venture capitalist. I am an investor. I see the crunch, the debt cycles, the credit, the problems that we are hitting now and about to hit. There is not enough money in the world for the requirements of this cycle and everything that you just said there, which I agree with. Then I look and go, and we haven't addressed the physical world and physical data and robotics and everything in between that's required there. I am holding these two seemingly opposing thoughts in my mind going, what the fuck happens now?

Walter Goodwin: This is why we need continuous innovation in every part of the stack that leads to serving intelligence. The good news is, for all that we are focused over here on chips, if you just look at the algorithmic side, model efficiency, the cost assuming no hardware change to serve a given level of intelligence is coming down precipitously quickly.

Harry Stebbings: Improvements in efficiency with no hardware change.

Walter Goodwin: Yeah. As we see longer context becoming a challenge, we see incredible innovations on how you can compress context and do sparse attention and other ways of handling this. There is sustained innovation on what you might call the compute efficiency of the models. With the same chip at ISO compute, how much more intelligence, how many more solutions can I arrive at? Then, of course, we do have the same on hardware, and we are working very hard to make sure that we have that hardware effect as well.

What is really thrilling, and the reason that I think on the whole this build out is entirely justified and we won't see a collapse in this market, is that far from the fiber build out of 2000, 2001, we have complete full utilization of the GPUs that are powered on in data centers. We don't have dark GPU fleets because we have this enormous number of problems as a species that we still have to solve. We now have tools that we can throw at those problems. We have many orders of magnitude more thinking that we could be doing at inference time to solve the hardest challenges if we had the compute.

I agree with you on the one hand, it feels like we need thousands of times more compute than we have today. We are not going to get thousands of times more data centers. I think that is probably a good thing for the world. What we will be able to do is take roughly that kind of land and power from data center sites that exist today, and over the next few years, through both model innovations and hardware innovations coming together, squeeze dramatically more intelligence per megawatt, per rack in one of those data centers.

Harry Stebbings: So I have Mike Cannon-Brooke on the show from Atlassian, and he said to me, "I think the cost of inference is going to come down dramatically over the next three to five years." That's pretty good at you.

Walter Goodwin: That's why I'm not an actor. You should do me in the next one.

Harry Stebbings: You in the next one is relatively close to just me. You want me doing Russian, that really is a good one. But with the Mike Cannon-Brooke theme in mind, what happens to the cost of inference?

Walter Goodwin: If you look at the cost, what will a token cost in 10 years time? It is one of those questions that begs another question. What is the model that we are running? It is probably orders of magnitude larger than the model that we have today. That token is also delivering substantially more capability. The margins in inference is something that will fluctuate and they will accrue in different places. We will be spending far more on running models in 10 years time than we are today. That part I think is clear.

Who is going to win in this space? You have chips, you have clouds, you have Frontier Labs building frontier models, you have folks that are building and serving open models. One of the reasons why I am very long on chips in terms of where that value accrues is because of the nature of the challenge of placing the right chip bet. It has this intrinsic lag, the lag to design the chip, to get the chip out, to have it built, and the degree to which it has to have a useful lifetime in a data center means that you don't have quite that same kind of fully efficient race, potentially a race to the bottom, in the chip space as you do elsewhere. If you have a better idea, if you can build some sustained advantage in some way, I think you see a lot of the value accrue there.

I think you also see a lot of it accrue at the frontier, the construction of frontier models, which have this sustained advantage over some of the others. It is a costly process. It requires incredibly talented teams working with huge amounts of compute. But it is something which delivers... who doesn't want the greatest intelligence on Earth? People will continue to pay for that. The overall prize pool expands. There will be relative winners and losers within that, but it is a positive sum game for inference.

Harry Stebbings: We just had Sacher on the show the other day, and he said that we would see a future where every company has their own specialized model with their own unique and proprietary data. Actually, a cool value segment would be evals and benchmarks for themselves and for that company and for their own objectives. Do you agree with that world of many specialized models on a per company basis? Do you think it is relatively techno-optimist and wrong? How do you think about that?

Walter Goodwin: I think there is absolutely space today. If you look at the technology, there is positive impact in taking a model and fine-tuning it for your use cases. But it is a transient thing. Generally, the advantage of having something that is internally fine-tuned will disappear as soon as the next release of the next open model arrives. I think that can still become a bit of a "yes and" situation. As long as you have a model where this stacks on top of the best frontier capability, then I can absolutely believe that there is an advantage to having an internal, specialized variant of something that is otherwise constructed at the frontier.

But I do think this is one of the hardest questions to answer. Over time, it may well be that you can only do this with the open models, for example. If that is the case, what does it mean for this trade-off? Is it better for me to specialize an open model or just go all in on the thing that has been released at the frontier that is still pursuing these better compute efficiencies and by extension might actually be very cost-effective for my challenge?

Harry Stebbings: When you look at that value stack, you said obviously chips, no shit, where you think is the most valuable, where is the least valuable segment?

Walter Goodwin: I think there is a danger for anybody that is working purely in the software layer. I have seen a few plays recently on orchestration. There is a coming wave of heterogeneity in the chips that we are going to have in the world: GPUs, other XPUs, these new chip efforts. Some clouds and some software providers have set themselves up to be the orchestration layer over this coming wave. I think there is a world in which that kind of heterogeneity is a slightly transient thing.

Today, we have a world where NVIDIA having acquired Groq will deploy two chips alongside one another, Rubin GPUs and Groq chips, and they will do what is called prefill decode disaggregation and attention MOE disaggregation where these pieces of silicon are doing different parts of the LLM inference workload. There is this sophisticated orchestration between these two.

Harry Stebbings: It's like model routing, but chip routing.

Walter Goodwin: Exactly. Chip routing. There is a kind of technologist view that maybe, hey, why stop there? Why not have seven different pieces of silicon? Maybe they are from different vendors and they are each specialized in different parts of this. There is a belief that this might be a huge wave of value creation, being specialist in how you orchestrate over these. I think a lot of that could collapse back down over time. There isn't a fundamental existential reason why we desperately need to have seven different platforms to run a workload. There isn't really a strong economic case for why we would be able to sustain that.

If you imagine even just the NVIDIA Groq data center deployment, how many NVIDIA GPU chips do I put in there versus Groq chips? The bet today would be this attention feed forward disaggregation. If you will give me a minute on the tech here.

Harry Stebbings: I will.

Walter Goodwin: I put the weights of my model on the Groq chip because I access them really fast on that high bandwidth. Because every user that I am cramming through that data center, they all want the same model weights. I can kind of amortize, I can share the cost of that much more expensive memory. I can recoup it and make it up on volume. Attention, which is the part that operates over the context, isn't shared across users. My context is different to your context. For that piece, I am incredibly exposed to the cost of my memory. We talked earlier about how SRAM is far too expensive, DRAM is much cheaper. So I need to fall back to the GPU and the GPU memory to run that piece. That is the logic of an NVIDIA data center with these two chips today.

But as I spec that data center, how many Groq chips do I put? What if my context lengths get longer over time? Suddenly, I probably wanted to have more NVIDIA GPUs than I did last year, and the context lengths have got longer. How much longer will they get? Maybe the model gets larger suddenly, and actually, I wish I had more Groq chips. You are very likely to end up in a place where some portion of that capacity in your data center is stranded. You are just not filling it because you have the wrong mix.

Harry Stebbings: That's unacceptable.

Walter Goodwin: It is completely unacceptable. You have a certain number of megawatts. You are trying to juice as much revenue as you possibly can out of them. This idea that you are placing a fixed blend and you are not going to be able to change that data center for the next few years, right?

Walter Goodwin: That's fixed. It's capex. It's being amortized over time. Heterogeneity is actually a really challenging thing. And I think it's a transient response to the fact that you don't today have a chip that can kind of do it all. These SRAM-based chips solve this fast inference challenge a bit, but they don't have the memory capacity to do attention. So we still need to bring that GPU in. They also don't have the flops density back to the top trumps to run the compute-intensive part of LLM inference, the pre-fill portion, the prompt processing, because you've got this one piece of silicon, which cost you a lot, and you've had to share the area that you might put maths in with the area that you put memory in. So you definitely haven't got as much maths as your counterparty that used HBM, who could use 100% of that chip area for the tensor cores that run matrix multiplies.

If you look at what Fractal is doing, you suck the SRAM upstairs, basically, and you have a DRAM layer stacked on top. All of that compute die on the base is now free to put tensor cores down. You can have as many flops as anybody. You can be at TopTrump's maxer on that. But you also get the SRAM level bandwidth from the 3D access and you get the capacity from the DRAM. Suddenly you have this holy trinity where you've got all of these things in one. And it becomes clear that you don't need heterogeneity and all of this complex orchestration and running the risk of stranded capacity.

Harry Stebbings: Was NVIDIA's acquisition of Groq an insurance buy? There are so many fun rumors about this.

Walter Goodwin: I think the Groq team is an incredibly smart team, and Jensen is a smart guy that wants smart people on his team. They also built a platform that I think answers a big challenge that NVIDIA faced at the time, which is how do we move into the fast inference space? I think Jensen is quite a positive sum player in this space. He said some nice things about Fractal last week when he was in London, and is in some ways an ecosystem player. The way that NVIDIA captures the bulk of this market is that they will sometimes be deployed next to other platforms that augment parts of their capability.

Harry Stebbings: There is another rumor that Cerebras got a deal with OpenAI. I think that can be an alarming moment if you're running a multi-trillion dollar GPU company and one of your largest customers suddenly signs on for something that you perhaps hadn't previously thought was existentially necessary, which is a fast inference platform. What's your response? And I think both deals were signed on the 24th of December, which must have been a fun time to be a lawyer.

Walter Goodwin: I think it's an acknowledgement that some parts of this space at the time required a different solution. And this SRAM play from Groq was the right solution at that moment to be able to run a lot of them faster. Today, though, we've already moved on from this. You cannot solve LLM inference with these SRAM-based chips. There isn't enough capacity. The cost to achieve speed is too high. The constraint on the context lengths you can operate on is very stark. I'd characterize that solution as giving you snappier chatbots, but it doesn't give you the ability to take overwhelming frontier intelligence, these agents that run for days on end, and truncate that to be orders of magnitude faster. That is where the huge prize is. Today, that's not something that provides an answer to.

Harry Stebbings: The different players in the inference space that people know are your Modals, your Fireworks, your Baseten, your Together. You would bet on one. Which one would you bet on?

Walter Goodwin: I'm friends with founders of multiple of those companies, so I'm bullish on them all. But it's like children, you always have a favorite. What I like about all of these guys is they still have the capacity to dodge and weave. They haven't yet locked up huge amounts of their own capacity. They're not a NeoCloud. They have the potential to bet on a new chip as it comes up. They have the potential to carve out new portions of the market. I think there's a world in which this piece doesn't fully commoditize. You can be a specialist in serving multimodal models, text models, fast inference. I won't place a specific bet, but I will say that I would rather be an inference provider today, a specialist in running these models, than I would a NeoCloud. I think there is more mileage in being able to create the business flywheel of signing up these enterprise customers that consume these tokens.

Harry Stebbings: Will they be multi-hundred billion dollar companies? Is it an Uber lift market where winner takes most and significantly less valuable second and third place? Or is it a GCP as your AWS market? Even that's only three players.

Walter Goodwin: I think you're not going to have 10. Today you probably do have a long tail that takes you towards that. If you add the NeoClouds into that, you definitely have tens of players, all of whom are maneuvering around billions of dollars. Are they multi-hundred billion dollar companies? I think we'll have a couple of multi-hundred billion dollar companies built in that space.

Harry Stebbings: Can you talk to me about the new funding round? How much did we raise?

Walter Goodwin: It was a $720 million Series C.

Harry Stebbings: Wow. That's quite a lot of money. How was the process? Was it fun? Because now you're the belle of the ball.

Walter Goodwin: If you go back two years, we were still teaching people about the word inference. That makes for a trickier investor pitch. When you educate an investor, you've got 30 minutes and you have to spend 20 minutes explaining. Four years ago, you had to spend 10 minutes explaining what an LLM was, then another 10 explaining why you felt that training these things wasn't going to be the be-all and end-all. I then spent five saying why I think speed is going to matter. We hadn't yet seen reasoning models. Compared to then, it's been very fun to be able to go straight in and talk about how we have a completely unique and extraordinary team that can do things differently, how we have a very unique position in the supply chain and can maneuver differently. A different set of philosophies on how you build to win in this space. That's what I get to spend my time over coffee talking about. There is a real advantage to now being in the eye of the storm in a space where there is an appreciation of why this is such a crucial problem to solve.

Harry Stebbings: $720 million. Is that enough? Suddenly you have that in your account. What changes? What do you do that you couldn't do before?

Walter Goodwin: Suddenly there's a weight of responsibility. There are a number of folks who have staked a lot of capital on us. One of our responsibilities is to ensure that we are always putting multiple shots on goal. Capital allows us to distribute ourselves to look at multiple things at the same time. We're now able to start looking more generations ahead. We're able to explore more parts of the stack than we previously would have. More innovation at the systems layer, at the networking layer, building this entire roadmap of future systems is one thing that capital allows us to do.

This is an expensive space, but the underlying capital needed to get that chip out, to get those first systems back, has been a fixed cost for the last few years. That's not what most of this capital is used for. What a lot of this is important for is the ability to maneuver faster in the supply chain, the placing of long-term agreements over certain components that we need to have years into the future, and the ability to build more of that muscle of predictive power. We can now start to think about how we run more of our own experiments, work out what might be important in the future. Fractal will always be a chip company, but we need to have strong insights so that we can be more helpful to our customers about what scales well, where model architectures might head next, how you can handle very long contexts, what it looks like to run an agent swarm versus one very long serial agent. All of these ML research challenges we get to go and tackle.

We're also expanding geographically. We've got folks in Taiwan, a team in the States. I'm very excited that we were joined just a few weeks ago by Dan Rabinovich from Meta, who worked on a lot of their AI compute systems, to drive our systems build out. Everything from the chip upwards, the rack scale systems, the networking, the integration with the data center. It is a very long stack, and we get to go and build parts of the business where it makes sense to, and pull in the world's best talent as we go after that.

Harry Stebbings: If you could invest in any one of your suppliers, who would it be and why?

Walter Goodwin: TSMC is an extraordinary business. If you look at everything that is happening today, everything traces back to them. This is not a scoop for your podcast, but it's pretty well known that TSMC underpins all of the world's chip production. It is cheap at the price. I would buy that stock every day of the week. All of the world's chips at the frontier, or the majority of them, are built there. They are an extraordinary business and an extraordinary team.

Harry Stebbings: If you could invest in any of your customers, who would it be and why?

Walter Goodwin: I really like some of these scrappy Neo Labs. I'm thinking like an investor here. They're a little cheaper. I think we are going to see some...

Harry Stebbings: Which one?

Walter Goodwin: Friends of mine have started Inherent in London. I don't know if you've come across them. They're looking at the RSI challenge, but also in particular what it takes to build systems that automate science and automate research. I think there is a lot of mileage there. It's not quite the same as pursuing the best frontier model that goes and tackles everything, but there are so many interesting questions in what it looks like to orchestrate and deploy models, build systems that are better at solving these challenges.

Harry Stebbings: If you were forced to invest in one of your competitors, who would it be and why?

Walter Goodwin: I'm a big believer in 3D DRAM. One other company called Dematrix is pursuing a different architectural direction in the same space. I think both companies are going to be very successful.

Harry Stebbings: Are you ready for a quick fire?

Walter Goodwin: Maybe I'm not quite ready for a quick fire. Do you want to read me the quick fires? Sorry, and I'll just...

Harry Stebbings: Is that not an honest Quickfire? Please can we keep that in? No one's ever ready for the Quickfire. What have you changed your mind on in the last 12 months?

Walter Goodwin: I think the underlying pace of capability advance in models has been extraordinary. These math results that we saw this summer, people talk about the truncation of their timelines. What has become clear is that anything that is verifiable, maths and coding tends to fall into this, will be solved in the next 12 months. We will be able to tackle any of these hard verifiable challenges with frontier models.

Harry Stebbings: Who do you not have on your board that you would love to have on your board?

Walter Goodwin: This is an easy one. I think Demis Hassabis is both an academic hero of mine and a bit of a British hero. I think he should come aboard.

Harry Stebbings: Have you asked him?

Walter Goodwin: Not yet.

Harry Stebbings: Well, this is a good way to do it. Demis. What's the biggest misnomer about the chip industry today?

Walter Goodwin: The misnomer is about technology businesses in general. They're very technological, but it's a very, very human space. So much of it is about relationships. So much of it in our case is about having this extraordinary team that can collaborate across the entire stack. It's powered by humans.

Harry Stebbings: Will NVIDIA be a $10 trillion business in five years?

Walter Goodwin: Yes.

Harry Stebbings: You said that pretty confidently.

Walter Goodwin: It's a very positive sum game. NVIDIA today trades at a lower forward multiple than Cisco after the crash, after the dot-com bubble burst. We're so traumatized by the dot-com bubble that we've priced in exuberance right from the outset. I think we're not yet appreciating how sustained this build out will be. You don't have to have incremental data centers added necessarily. If I have finite numbers of megawatts, I need to replace my chips every few years with better chips. That's a huge advantage for Fractile as a market entrant because every chip that you see in the data center today will definitely be out of those data centers in six or seven years tops. Many of them won't last more than a couple of years. That's an enormous market opportunity for the market leader. They have extraordinary supply chain relationships. They're going to do very well. There is then this enormous new opportunity in being the chip vendor that serves the needs of the frontier. I'm very long on Fractal's chances over that horizon as well.

Harry Stebbings: You are in charge of the UK. Congratulations. What do you change to make the UK a superpower in technology?

Walter Goodwin: So much of what unlocks growth is unlocking energy. It is doing everything that we can to bring online more renewables, more nuclear, more sources of power that allow us to have build out of industries that matter here. I love Alan Chang at Fuse.

Harry Stebbings: I love Alan. Yeah, phenomenal. Final one, what are you most excited for about the next 10 years?

Walter Goodwin: We're using these models now. What are we using them for? We're using them to solve problems we couldn't solve before. I think we're only just starting to see that pick up and take off in terms of what it means for the real world. It is going to be all of the ways in which life gets a bit better. We find cures for diseases we couldn't previously tackle. We have education that is high quality for everybody. These things are unequivocally going to come. We have the technology to make it happen now.

Harry Stebbings: I totally agree with you. I think it's so awesome. You look at class sizes traditionally, 40 children in a class. With AI, you have personalized education for some that like multiple choice, some that like visual learning at their own speed. I was shit at maths, probably unlike you. I was always too scared to put my hand up and say I don't understand because I'd look stupid. Having personalized learning to that point is one tiny example. I so agree with you. Listen, Walter, I love doing the show because I really get to learn from amazing people like you. Thank you for doing it with me. This has been so much fun.

Walter Goodwin: Thanks very much, Harry. It's been very fun.

Harry Stebbings: But before we leave you today, founders and investors sleep on 8sleep. It is the intelligent sleep system trusted by Mr. Mark Zuckerberg, Dr. Andrew Huberman, Charles Leclerc and the Aston Martin F1 team. The pod is a smart mattress cover that fits on any bed. It cools each side to 55 degrees, heats to 110 degrees. It even tracks your heart rate, HRV, and sleep stages without a wearable. Thank God. The new Pod 6 has a hub 50% smaller than the last one and starts at just $1,999. Visit 8sleep.com slash 20VC. That's E-I-G-H-T-S-L-E-E-P dot com slash 20VC.

While 8sleep helps you sleep better, MongoDB helps you build better. Every serious AI company right now has the same problem. Agents need accurate context fast. Most infrastructure wasn't built for it. Well, MongoDB is. 11 Labs runs 40 million agents on MongoDB. 75% of the Fortune 100 run their most critical apps on MongoDB. Trillions of dollars moving through MongoDB every single day. Instead of stitching together 10 different tools, you get one JSON-native database, VectorSearch, and Voyage AI embeddings in the same system. And it runs anywhere, no lock-in. If you're building an AI, MongoDB for startups helps you move faster with Atlas and Voyage AI credits and technical support to help you ship. Don't build agents that answer once and forget. Build agents that remember and learn from your real-time data. Go to mongodb.com slash agents. That's m-o-n-g-o-d-b dot com slash agents.

While MongoDB scales the product, Framer scales the story. Framer is this incredible, complete website platform, not just a builder, so teams can launch and keep improving their sites in one place. Framer is the AI website builder that helps creators, teams, and businesses ship production-ready sites faster than ever while getting every detail right. Prompt, inspect, edit, and publish in one place at a whole new pace. Agents and humans work in tandem. Agents bring speed and scale. You bring taste, judgment, and control. The work lands on the canvas and stays editable. Build custom code components, manage CMS content, optimize SEO, and audit for issues all in one place. Enterprise-grade hosting, security, and 99.99% uptime SLAs trusted by leading brands like Perplexity and Miro. Learn how you can get more out of your site from a Framer specialist, or get started building for free today at Framer.com slash 20VC for 30% off a Framer Pro annual plan. That's Framer.com slash 20VC. Rules and restrictions may apply.