A Push Up Contest With Pat Gelsinger 2026 Ian Interviews 49

TITLE: An Push-Up Contest with Pat Gelsinger (2026) // Ian Interviews #49 CHANNEL: TechTechPotato DATE: 2026-04-09 ---TRANSCRIPT--- So, next up in our interview series, well, it’s Pat Gell Singer. Do I need to say more?

Good to see you, Pat. And great to see you. Thanks for having me on. We’re here at Playground Global, which is Pat’s new home. So, I just have to ask, how’s it going? Well, you know, as I’d say at this phase of my career, uh, just turned 65, just had my 65th birthday. Uh, you know, you don’t look a day over 40. Oh, thank you so much. You’re the man. Uh, but, uh, you know, I want to do things that matter, right? you know that you know if they succeed make a difference with people I enjoy right you know pretty simple at this point uh in life and playground’s a place I can do that you know I have about 10 companies that I’m deeply involved in now right on the board of most of those and uh I think uh some of these are going to be industryshaping companies and if I get to help build great companies with great leadership uh teams at this phase of my career okay this is good and so far the year you know I’m a rookie I’ venture capital, but uh the team’s embraced me and we’re making a difference. Did you ever see yourself eventually migrating into this sort of role? It’s a bit hard to go from a sustained, you know, big company sort of position into, you know, VC. Well, remember 45 years, right? Not a day was I, you know, when I left Intel the first time, the EMC, literally I retired at night, started the next morning. when I went retirement. Yeah. Anyway, you know, officially I retired. Uh and uh when I went from uh EMC to VMware, literally I did both jobs for two weeks. When I left VMware to come back to Intel, I did both jobs for four weeks, right? You know, this has been 45 incredible years. And then all of a sudden, yeah, my wife walks into my study literally the next day after the uh announcement uh departing Intel and she said, “You’re not done yet.” Right. I think that meant don’t be home too much. Yeah. Yeah. Yeah. Yeah. I love you, but there was there were limits. There a little bit too much energy there. So, please. And uh that began a 100 day journey for me, right, of figuring out, okay, what do I want to do next? And I interviewed a lot of venture firms, private equity firms, uh a bunch of CEO roles, you know, government roles. It was when the administration was coming in, you know, right? University roles. And at the end of that, it was sort of like, okay, you know, I want to make a difference. And Playground came to the top of that uh list. Uh and then also my glue uh company, the faith technology company. So I’m splitting my time between the two of those. you know, making a difference eternally and making a difference in the future of science and bringing those two together. Okay, that’s a pretty good position of life. It’s I can’t imagine how many people, you know, I use a phrase came out the woodwork, but to offer you advice on the next stage, of course, and a number of those, hey, I wanted their advice, you know, I wanted but you know, 100 meetings in a 100 days to figure out what was next and here we are a year later and it’s going pretty well. And still semis uh you know a lot semis but you know I’m doing alva you know nuclear operating I didn’t know anything about nuclear now I’m doing that you know I wasn’t surprised you know everything right yeah well I am learning a lot right you know I’m doing uh superconducting right Joseph’s and junctions and you know okay it’s sort of incomputee but wow those are like really different looking transistors right you know from what you know we thought about uh worrying about cryogenics stretching my mind into quantum computing and for a digital boolean I my entire life, you know, it’s like you got to turn your brain, you know, to a 90 degree angle to think about cubits and how they’re functioning and tangled and how you use them to compute. So, a lot of learning, you know, for me personally, but you know, as I say, you know, the only disappointment I have at this phase of my career is I’m not 35 years younger. This is the greatest time to be a technologist in human history. So in my mind I’m seeing you at a desk pouring through papers of potential startups trying to find the one that’s you’ve got something a bit unique is well is that you know I’m reading a lot right so lots of papers so that’s part of it you know a lot of that work gets done by our associates right so you know hm that one sounds pretty interesting and they’ll seek my advice on how it would fit into the industry they’ll go do work you know and we’ll meet you know I like reading but even more so I like interacting okay you To me, if I I could read for 10 hours or meet with the company for one hour, the one hour is more valuable. You’d rather have the college lecture than the Well, it’s not it’s not the college lecture. It’s the questioning, the interaction. I want a few smart people at the table. I want to ask questions. I want to hear their questions, how they respond, you know, and that’s my that’s my best learning modality, you know. But yeah, I mean, my brain is being stretched in so many dimensions. I’m learning about bioengineering. You know, how semiconductors are being applied to bioscience as well. You know, what I call the trinity of computing, you know, the fusion of classical AI and quantum effects, right? Coming together. You know, I think that will be, you know, literally, you know, the problems that we will now be able to compute, right? Will be, you know, almost everything that we’ve been able to express mathematically will now become computable, right? So science, biology and of course you know areas like some not NP hard I guess. Yeah. But you know even so I mean you just have this powerful new tool coming over the horizon with quantum computing you know where many of those uh you know quadratic functions that were uncomputable before. Okay you now they’re sort of like yeah great let’s go uh you know attack those portions of the problem as well. So how do you marry the fact that you’re not you know not completely green and brand new but it is you know these are new uh new degrees of freedom for you to explore but then also evaluating whether a startup is actually worth the investment right there is something to be said for the person who spent 40 years in a space and their uh addin to to that discussion. Yeah. And you sort of come down to sort of three questions, right? You know, one is okay, can we make the technology work, right? You know, just is there, you know, real proof and we do hard tech things, right? So there really is this question, does the tech work, right, at that and then you’re sort of saying, okay, how does it get to market, right? How do we bring it, you know, into an at scale deployment? So, you know, combining the business side and then, you know, the third is do we have the team to do it? Yeah. Right. the leadership team and obviously you know for each one of those okay the tech you know I mean you’re a playground right we have more brilliant technologist to ask those hard questions you know what are going to be the milestones in the way to show that we get there you know the market insertion and by the way that’s part of my unique value right you know I’m able to call up CEOs throughout the industry hey you know don’t let this one get stuck down here as some entry level person as you evaluate this one matters right let’s work together on how differentiate your offerings, you know, your GPUs, your networking components, right? You know, your business using these and so that connectivity, you know, helps quite a lot with number two, right? You know, driving the business uh development and accelerate it, you know, into the marketplace. And then obviously with 45 years of uh uh major leadership roles, okay, my job is to shape leaders into great CEOs, great, you know, CXO teams uh as well. And I give them some pretty tough feedback, you know, sometimes. And you know, some of them are, you know, great leaders, you know, couldn’t present to save their life. Y okay, we’re going to make you at least a mediocre presenter, right? You’re going to be able to do that. In other cases, they’re great presenters, you know, but they’re not great leaders. Okay? You know, how do we build a leadership team around you, you know? In other cases, well, you know, right, you know, we got to tell you how to make a sale, right? You know, I mean, there’s all these things associated, you know, with building a great company. And since I’ve been in these big leadership roles, you know, I really enjoy bringing that into the CEOs. But the hardest thing, Ian, okay, I don’t have my hands on the wheel. Every once in a while it’s sort of like Yeah. Well, you can lead a horse to water, right? Yeah. For some of those, but uh you know, it also does give me a lot more flexibility as well. So, tell me how you solved this one. By Monday morning, I’m taking the weekend and taking my grandkids skiing. So, you got that veteran status that you can take the weekend off. You’re a startup. You still got to work. Yes, absolutely. If you like this content, there are multiple ways to support the channel. You can like and subscribe this video and many thanks for doing so. There’s also Patreon which gives you access to our Discord. There’s a merchandise store and a newsletter. Links in the description. For all of you who do contribute, thank you. You are keeping me wellfed. I’ve been seeing some of the magic you guys have done with um our good friend Mark Wade over at IR Labs. They just did their series E, you know, mighty impressive and they seem to have funding from everyone now. Yeah. Well, you know, we also I’ll say there’s this this characteristic where, you know, you get a term sheet, you know, somebody ready to lead the round, price the round, you know, and then okay, people are ready to jump in, right? And there’s this sort of, you know, interesting thing, you know, what’s going to cause that dam to break? Yeah. Right. You know, for obviously that was Did you mean that in the sense of people don’t want to fund or it just may end up being overs subscribed? Um, everybody is sort of around the table. Who’s going to put the lead term sheet down for this round? Who’s going to price the round? Yeah. Who’s going to say, “I’m going to put the biggest check into this round.” And as soon as quote, you know, them or, you know, maybe it’s two who are sort of co-leading, you know, the round, then others are ready to come into it. Okay, the round is going to get formed, right? You know, we’re going to see it. Now, we want to be part of that as well. And that was very much the case with Mark and the you know series E once we had that lead term sheet sort of broke the the dyke a bit then the energy unleashed of many other people wanting to come into the round and quite a few strategics uh as well. So you know to me as I’m learning the venture uh game uh if we could call it that you know how we you know having that network of people that I am confident that they’re going to be my leading with me you know to form rounds in timely fashions because if it takes us three months to form a round okay that’s three months later in the market as well so you know and I don’t want to let some of the hottest companies you know languish in competitiveness by not moving them forward effectively as well It’s speaking to a few people in the last couple of months since the uh Nvidia’s acquisition of Grock a lot of people who are raising have suddenly found it’s getting easier in 26 are you finding similar the you know clearly that was catalytic yes right you know for now Nvidia for it yeah you know uh Nvidia is not going to acquire 10 companies in that space you know they were clearly deficient in having optimized inference and you know uh as I said at GTC uh last year in the pre-show we need to make inferencing 10,000 times better you know not 10x not you know 10,000 times that’s a great headline the question is how yeah right you know which you know when you think about you know it’s replacing search right we’re going to you know now you know with open claw something none of us uh you know uh quite predicted even though everybody was predicting a gentic right? You know, it’s just demonstrating that, okay, we got to make inferencing a lot better, right? uh you know for it and you know my 10,000x was sort of you know right a number that I pulled out you know based on some math of where you know search was and so on in terms of energy compute right uh cost but uh that led Nvidia right as proud of they uh as they are and should be yeah uh around the uh incredible progression of the GPU saying okay the GPU is great for training you know it’s great for some of the waterfall training into inferencing but it’s not an optimized inference chip, right? And that led them to do Grock, obviously, but now, you know, there’s 20 companies pursuing that assignment to say, how can we be 10x better or 100x better, you know, than where Nvidia just described, you know, the LPU with Rock. I had to tell you, but I’m tracking 150 these days. It’s anything can precede, you know, all the way up to I got I got to see your list here. So, yeah. Yeah. you know, the variety of SRAMM focusbased and uh, you know, data flow RAM machines and you know, yeah, people talking about HM5 and high flash and um well, so so so that that brings me on to the fact that when I look at the companies that often put you in their press releases when hanging around take, you know, Snow Cap for example, this uh, you know, sub-zero, you know, almost absolute zero type of computing. Um, it seems like a lot of bets you’re making are non vonoyman these days. despite, you know, you literally just saying you’ve spent so long in digital logic, um, and, you know, traditional compute, for lack of a better term, you know, x86 is still high and mighty, and I know you’re very proud of that. So, to turn around and say, well, hey, let’s look at, you know, the analog side of the business. Let’s look at different ways of computing. Is that where you see the future going or is this just a we have to optimize for every workload sort of play? Yeah. Well, I do think, you know, well, part of that is, you know, the language I’ve used called the trinity of computing where, you know, we’ve always believed in accelerators. I’ve always believed in accelerators and whether they were SIMD accelerators, SIMT accelerators, right? Uh, you know, MIMD, right? You know, there’s other, you know, forms of how you attack workloads. Yeah. You know, to me, you know, architecture in that sense, you know, trying to fit the world through a vonoyman lens, you know, means that all workloads look vonoyman, right? Is the workload defining the architecture or is the architecture defining the workload? Yeah. And in this sense, right, you know, what we’re now and you know, I think clearly, you know, the surge in AI, right, you know, has clearly said, hey, different architectures enable different workloads to emerge at scale, you know, and that’s what’s that, you know, that’s what’s made this period exciting that all of a sudden, you know, datacentric workloads, right? You know, with, you know, matrix functions against them, right, are able to do really interesting things. But it’s the workload that matters right and in this sense you know I think you know I was saying that before even Jensen was saying that uh in some of our debates so it’s the workload that matters in this so which workload right that’s interesting solving unique and wonderful problems and then what’s the best architecture to run it on you know there’s nothing that you saw yesterday at GTC that can’t run on the CPU right yeah right yeah right that’s the beauty of vonoman you know type machines can you run Jensen on the CPU Maybe he’s got a little bit of magic there, but I know what you mean. I know what you mean. You know, because, you know, it may not run efficiently. Yeah. Right. It may take a long time, you know, and as we said, you know, there’s algorithms that only can run on the quantum machine. No, I can run them on a, you know, AI or classical machine. They just take a billion years and run out of memory. Yeah. Yeah. Right. You know, I can’t get there from here using that computing architecture. So I do think in that sense you know we’re going to see this multiplicity or heterogeneity right of what the compute architectures are most of the interesting problems will combine them. You know yesterday Jensen announced the CPU right you know what a backward thing for him to do. Why are we going back to the CPU? It it was funny I because at at the beginning of the presentation I’m not sure if you caught it he said you know the CPU is old. It’s the legacy. Yeah. And then as you’re right he then announced the CPU. Yeah. Right. So sort of you know of course I noticed that. Right. I’m a CPU guy. Of course I noticed it. It’s like, so so first we’re going to spend half the keynote throwing it under the bus, right? We don’t need it anymore. And then we’re going to announce that we now have the best one because we need it. Yeah. Right. Which is it? Well, the answer is the workloads need both. Yeah. Right. There are things that run like crap on a GPU, right? You know, if I have control flow, I need my code. Bunch of my code, you know, but you know, they, you know, if I have control flow related, right? You know, functions, they’re terrible in a GPU. Anything predicated and Yeah. Right. You know, right? you know your basic if then else is a terrible thing to do on a long pipeline GPU you know may be able to run you know six different dimensions of parallelism you know but you know right you know if then else is not a parallel function right so the you know in that sense you know I really do see it as a heterogeneous view right that’s why I call it the trinity of computing classical you know that’s going to be control flow you know toolbased you know analytics uh you know based uh operating systems all of those things you know more in the CP CPU, right? I’m going to have this raft of things that are datacentric uh algorithmic that are going to be great on the AI and then I’m going to have these things that only work because of a quantum machine, you know, and being able to apply, you know, entangled cubits, right? You know, against those problems. And with that, okay, now we can open up a whole lot of workloads that are not computable today. It it’s I know people are going to be shouting at the camera because again we we discussed this back in the day and uh you know your good friend Greg Lavender was you know on top of this it’s the software strategy right you can’t have a heterogeneous uh installation without having the software that attacks and I’ve always been a vocal critic of the write once compile to target because it’s never as efficient as you need it and we know working with hyperscalers they’re going to extract the what every percentage point of power and efficiency out of code and you can’t do that at a high level where you might have this heterogeneous. Um so what it’s it it sounds mar to ask but what can we do here? It it’s we we’ve had standards in machine learning you know for example on you know Windows ML and uh you know Intel tried to do it with one API and um everything else but there just doesn’t seem to be any consolidation right now like we had in GPUs. Yeah. And and you know I do think you know we sort of abstract abstract abstract right and then all of a sudden we need to collapse. Yeah. Yeah. Right. You know when there is that and you know I think everybody was you know uh a little bit over a year ago was the deepseek moment. Mh. Right. Well what did they do? Right. You know I’ll say they collapsed. Yeah. Right. They went to really understanding what the machine was doing you know and you know getting very targeted on how to use the machines to be much more efficient. Stick people in a box and they’ll engineer their way out. Yeah. Right. And uh you know I mean the you know what does an engineer do right? He produces great results in the constraints that he has right you know and to me you know that was part of what made DeepC such a defining moment because they sort of tunnneled through the stack said we really know what’s going on in hardware. We’re going to align our algorithmic functions against what hardware we have available and is capable. And I think we’re going to you know see some of that happening again. You know, I also think, you know, one of our portfolio companies, next silicon, you know, building a programmable data flow machine saying, you know, hey, this has all gotten too complicated. We have to create an abstraction sitting beneath a more programmable hardware embodiment, right? You know, that uh is able to statically and dynamically reconfigure itself both in network topology and in compute, you know, resource to the most efficient way of the current running workload. And as we’ve all seen, you know, in AI workloads, the characteristics of the workload change dramatically over, you know, am I in the prefill phase, right? You know, and then, you know, and you know, as you go through the different phases of computation, right? You know, big GPUs are turning on and then they’re turning off and power rails are bouncing across the scale. You know, networks are being overloaded and then they’re beingow, right? So, you know, for that, you know, I think there has to be this layer of programmability. Part of the reason we’re quite excited about, you know, next silicon, but you know, clearly, you know, we have DRAM shortages today. You know, imagine that, you know, four years of great DRAM business. This ain’t ever happened in industries history, right? You know, now you have extraordinary logic and DRAM demand. That never happens in the industry. Everybody says we’re so used to these bubbles in in these commodity markets. And the question I get asked a lot by investors, and I’m sure you do, is when is it going to end? Yeah. And and I’m not sure about you, but I struggle to predict that end because um realistically the only way it’s going to end from my perspective is if somehow the bottom comes out of the market. Yeah. Well, you know the uh uh to some degree, you know, the gas law, Devon’s law, right? You know, whichever way you view it, you know, you know, we’re in a you know, Prometheian period of, you know, compute expansion, right? And in this sense, you know, and I think OpenClaw, you know, clearly has been sort of that, you know, that next accelerant, you know, that all of a sudden, you know, how many tokens per day are you going to use as an engineer, right? You know, because, you know, hey, I’m going to spend $100,000 on a token. I’m going to give you a $200,000, you know, uh, salary, right? You know, okay, you know, what point does that become? That’s the argument I’ve made. It’s the how much do you want to put in into an engineer? Yeah. And you know there doesn’t seem to be any end in that value proposition right now. Now obviously as I said we have to make inference 10,000 times better. I do think some of these compute memory architectures are going to have meaningful breakthroughs right you know that are going to you know make inferencing dramatically cheaper uh and you know not 10x like we saw from grock yesterday but thousandx or 10,000 uh x better. Did did you see the talis announcement earlier in the year? No, I didn’t follow that one. Uh, so they’re doing what looks like a structured ASIC design. So you bake in in the communication metal layers your model. Okay. Um, and they’re they’re seeing 10 14,000 tokens per second. Okay. Impressive. Yeah. So, so pure ASIC style, you know, traditional ASIC by definition. Mhm. Um I’m not sure how you feel about people calling things GPUs versus A6. You being the being the traditionalist you are, you know, uh you know, tomato tomato, right? You know, some investors sometimes don’t realize. Yeah. Um you know, but you know, at the end of the day, you know, the you know, market acceptance will be based on, you know, tokens per second, tokens per second per watt, you know, aggregate throughput, uh capabilities, latency, you know, it will deduce to real engineering measurable results uh over time. So my argument against that has always been it doesn’t matter how many tokens you produce if your tokens are useless right the value of a token right and I’m sure you’re seeing the same thing I am that right now the biggest workload that has value in the token for the output is code people are willing to pay you know if you accelerate your engineer um or I often site IBM because they have a consultancy business where they generate tokens for their clients and they can upsell and the client just sees you know the reduction in cost. Um, is there anything we can do, you know, to stop talking about just garbage token? Well, you know, let let me let me disagree with you a little bit. You know, I’m, you know, I’m now starting to see very meaningful agentic business process workflows, right? You know, where people are turning, you know, I’ll say, you know, nominally, you know, low-end white collar business process flows into highly agentic workflows, right? where I truly am putting agents to solve things, you know, for me, right, that make me more productive, more scalable as well. And by the way, that’s sort of what led to the whole SAS implosion. Yeah. Right. You know, as people started to see, oh, what’s going to happen to my Salesforce, my, you know, my Oracle, you know, all of those start changing as well. And those are very real, you know, and I think that probably becomes the AI pathway into most enterprise, you know, workloads as well. Once we solve security problems and people get comfortable, right, you know, with where they’re running and whether that’s on prem or in the cloud or, you know, what the security model is, I think that’s powerful, right, associated with it. You know, I do think there’s always going to be a little bit of my token is different than your token, right? You know, so I do think there’s going to be, okay, what really is the token? But good benchmarking is going to help us sort that out over time. And you know I’ve done more benchmarks in my career than most humans ever should consider uh you know for it. But you know bench you know there’s lies lies and benchmarks. But you still are always seeking you know the best metrics that allows us to deduce differences in hardware. That’s going to get harder as we’ve already talked about with regard to heterogene as architectures uh as well. So you know the the role of good benchmarking is going to get more important in this next phase. And by the way, it’s going to be critical because people are going to be making billions or tens of billions of dollars of capital decision, you know, based on that work. So, it’s going to get to be more, not less important because otherwise I’m going to build data centers that produce crappy tokens, right? Okay. Yeah. It’s I I I think we started calling it benchmarking. Well, depends on the graph you get from the Remember, my code is in the spec benchmark, right? One of the most venerable benchmarks of all time. You know, my code is still in there. in in six or 17. Uh yes, all you know all right you know from the very earliest days of spec right and my code was there. So which test which subtest? Uh well if you go into the spec benchmark you look right you know the uh you know like the espresso flow that was mine right you know associated with it you know we had the compiler flow that was my you know code that we were using in the compiler when we initially did it you know I think I have a third one in there as well. Okay. So, would you trust yourself to do that again? Oh my, you know, at the time it was great, right? Of course, you know, when people give me those numbers, I think they suck today, right? You know, it’s like everybody was trying to break cash sizes and so on with them at the time. Uh, and uh, you know, today, you know, it’s just not, you know, it’s no longer a valid view, right? You know, so I would just yell at our engineers in my last four years at Intel when they would show me spec numbers or spec rate numbers because, you know, it’s just like, okay, that’s code is so old. It’s been so tortured into cache footprints. You know, there’s no longer meaningful view of system performance. It’s and there’s a small argument here because I’ll speak to ARM and they still value spec 2006 because of the embedded market. The embedded market still relies on it. But no, you know, I I I completely understand your point. Going back a little bit for a second and when we have the you know these agentic setups um I often see that especially if you you know look into the wider industry a lot of people are playing with it and it seems a very personal implementation um on on on people improving their workflows and I struggle to really see where it’s going to offer it at scale and the the only sort of workload I’m seeing where it is actually being applied at scale you know you buy your software program and it offers your agentic flow is chip design. Mhm. Because our good friends at Synopsis and Cadence are leaning on it heavily than almost anyone else. Mhm. Um so so h how do we bridge that gap? Is that just a time thing or is that just a familiarity thing or Well, you know, I do think you know and you know that’s a great place to start, right? So let’s just say that’s wonderful, right? there’s a whole lot of other engineering workflows other than that you know in material science and you know CFT and you know uh you know all of the you know now we’re worried about hypersonics and so you know there’s just all sorts of wonderful things that open up you know across the greater EDA space not just chip design right you know and the explosions that are happening and you know you know bioengineering right you know okay now you’ve you know created orders of magnitude of additional complexity as you start looking at you know threedimensional molecular models. You can’t make nice 2D simplifications like you can in a lot of the chip design flows, you know. So, I do think in the in this sense, you know, right, the you know, you know, we’re at the beginning of these processes, not anywhere close to the end. And I do think that, you know, every one of these as you open one up, somebody’s going to be the one that sort of jumps ahead and then, you know, what happens is it’s sort of like cockroaches. You know, there’s a little bit of food in that corner. You know, you know, a lot more industry innovation will run in that direction. So, so gi given your background and it relates a little bit to my background on the HPC side um I speak a lot with the people over there um on almost on a weekly basis and they’re getting frustrated by the whole march of machine learning um you know especially as we look at 64-bit precision you know being reduced in you know traditional accelerators and uh I know that a lot of them are wary of you know 64-bit emulation in in in in in the 8bit side. Do you have any good news for him? Well, you know, a couple whatever your opinion is on a couple of thoughts. I mean, when people started talking to me about BF4, right? I was like, what are you talking about? You know, I was worried if like FP1. Yeah. You know, sort of, you know, I was worried if 64 bits was good enough, right? We needed the 80 bit modes that we put into, right? Uh, you know, I e54, right? you know for precision thing you know and now we’re saying that four bits is good enough right and I think even there you know four bits is you know people are doing uh you know interesting uh uh work but what they find is models actually don’t settle very well when you get to precision that low so you know useful uh you know bits seem to be more in the 8 to 16 bits for most model work you know I don’t you know I’m not a model expert in that regard but you know right you know they sort of run some of those you know benchmarks right? You know, at 4 bits, but the reality is most machines aren’t actually operating, you know, there. And, you know, emulating 64 bits in a, you know, 16- bit environment is very different than saying, hey, we’re going to do it in the 4-bit environment, you know, but I do think as we, you know, move to the next phases of science, right, you know, and not just doing, you know, large language model, you know, performance, as we move to the next phase of science where the LLM almost becomes a leaf node, you know, not the core computational node, right? and you’re gonna be back to worrying very much about precision. So I think there will be over the next couple of years the revenge of the HPC guys. Okay, because you know okay now when I want to start using LLMs in the context of you know a CFD right you know for you know my airplane or hypersonic design you know I’m going to be fusing these together and I’m not going to be possible to give up precision for this portion of the workload even though this portion of the LLM you know may be perfectly happy running on a BF8 machine right I’m going to need them both you know so I do think that the next phase of application of AI you know is going to be increasingly in the science domain will be increasingly in three dimensions you know not in flatter you know language representations and as a result you know machine you know all this pursuit right of getting to you know 4bit 2 bit etc you know I think some of that’s going to be for not right because the real machines are going to be combining those together you know in a in a much more central way in workloads so now maybe maybe I’m just trying to convince myself that all that hard work I did on 64 and 32-bit floating point is going to come back you know reality but I I think it’s based on the view of workloads the next phase is not doing more language modeling it’s much more science modeling as you get to science I need real precision to look at many of those algorithmic domains again it comes back to the workload yeah it’s um speaking to a few people who run um you know university uh supercomputers they’re saying because of the price of accelerators these days and the high demand they’re almost seeing cloud be more cost effective. Yeah. Right. And and and if there’s one thing you don’t want, it’s a student accidentally clogging up a thousand GPUs in the cloud with bad code. Um so yeah, I I think they’re struggling and and you mentioned them before, you know, our friends at Next Silicon are trying to solve part of that. Yeah. Um I I I think there’s a community who really want more love. Yeah, more love. I think so as well. And you know remember you know uh Nvidia began its journey into AI through the HPC door. So you know I do think there’s going to be good science being done there. I think the national labs matter a lot. I think some of the you know government programs matter a lot here and you know people like next silicon I think will be helpful you know but I do think uh hey there needs to be more networking work right you know right you know I I joke that NVL72 is an engineering marvel and a manufacturing nightmare right? you know, it’s just how do you build those things, scale them? You’re hitting just the limits of copper. You know, we saw a bunch of optical conversations. You know, I think we need uh, you know, new models that become resilient networks work like we’re doing with Delos data, right? You know, so that we uh create more flexibility in the underlying network because we can’t have failure rates, you know, that we’re seeing today for these super large, you know, configurations. Then you’re just spending all your time either restarting and redoing or checkpointing, right? you know, and uh that that’s not a computing architecture suitable for a lot of the workloads that we want to, you know, get to. So, you know, I do think some of these things are going to move us more rapidly to optical, move us more radically to resilient networks, move us more radically, you know, back toward data flow machines with the full gamut of precision, you know, for it. And then, of course, we’re going to do radical things like Snowcap, which are just, you know, you know, thousand times better. It’s it’s so I I’ll go speak to people like you know IBM with the Z processors and then some of the automotive players and they speak about resilience as a very hardware focused element yet you know with the lost data and others we’re seeing it more as a software application on top of the hardware is there a right balance uh you know I I think the workload again is the determinant right you know in the sense that hey if the software has assumed faulty hardware right you know and has built resilience into the software layer, right? Okay, you know, then software is the right answer, right? You know, but if in fact a whole lot of workloads, right? Uh right, and the sort of the difference between a virtual machine and a Kubernetes uh uh machine, you know, the difference is okay, who is taking care of resilience, right? In the VM, I’m assuming the hardware is right and I’ll run any software including containers, right? In the opposite view, I can only run containerized software because I’m not presuming resilience in the hardware. And for a lot of things, hardware is better for resilience, right? You know, I can see memory errors, I can correct memory errors, I can, you know, see link failures. ST area that I could spend on another ALOU. You you you certainly could. And I think now as we’re looking at machine learning and failure rates that we’re seeing in machine learning, we need more in the hardware, right? it’s not resilient enough anymore, you know, for some of these incredibly large cluster sizes that we’re seeing emerge to as well. So, I’d say, you know, I don’t think there’s a one answer to that question, right? And, you know, I I do like uh Jensen’s, you know, code design perspective this way because hey, you know, there are certain things that I think are done better in software resilience, but there’s a whole lot of things that, okay, don’t let the hardware guys off the hook, right? You’re going to build, you know, right? You know, 59’s hardware. you’re going to be able to prove, you know, 10 to the 12th, 10 to the 14th, you know, uh, error rates, right? And, you know, until your material structures have proven that, you know, get back in the lab and finish your freaking work, right? To, you know, uh, you know, prove something that I can really build on into the next generation. It’s Last time we recorded one of these, uh, it was at, uh, Intel Foundry Connect. I’m sure you remember. Yes. Back in the, uh, five nodes in four years days. How much do you spend these days thinking about next generation foundry and packaging technology? Well, uh obviously Xite, you know, I’m spending a lot of time on how we build uh you know, next generation, you know, as I call it, you know, waking Moore’s law up from its nap, right? So spending a lot it’s not dead. No, absolutely not. Absolutely not. you know, I think we’ve put it on an economic uh uh uh pause uh you know, for it, you know, because transistors haven’t gotten cheaper. I can still build more of them, but they’re not cheaper anymore. Um and they’re not uh right you I’ll say as effective powerful tool as they have been in the past you know that said right you know okay Xite you know bringing you know a better light source to EUV you know enabling us to think beyond today’s 13.5 nanometer you know of light to moving uh you know solving stochastics polarization you really think there’s going to be something less than 13.5 nm absolutely absolutely right you know I think I disagree oh why do Do you disagree? It’s I don’t think I’ll see anything beyond the UV in my lifetime. Oh. Oh, man. Because of the stochastics, because of the energy. Oh, but but we could take, you know, with a free electron laser, I can give you so many more photons, right? Yes. But in terms of making it commercially viable at scale, absolutely not not the not the slightest hesitation in my mind. Yeah. You know, that we’re going to pull that off uh over the next decade. absolutely the case, you know, and whether it ends up being, you know, in the 7 nanometer range, in the 4nm, you know, range, absolutely, there will be a next generation wavelength of light, uh, as well, right? It’s going to be at much, much higher power levels. Free electron lasers can produce, you know, 2, three, 4,000 watts, you know, of delivered energy, right? Which allow you to radically reduce the stoiccastics. Even if you apply that only to, you know, 13.5, you get a big win in the yield, you know, characteristics. I’m going to be able to go from double patterning to single patterning. So, I get more productive in my capital efficiency. But but you can’t reflect and it means you have to scan the wafer vertically in order to do it. And hey, there’s going to be new material structures that, you know, we’re going to innovate. So, absolutely, you know, uh you know, you you you skeptic for the advancement of science and new materials. This is going to be I’m skeptic. I’m I’m I’m I’m pulling on my material science background. Okay. For you know, for your audience here, let’s you and I are going to take a bet right right now, right? You know, that we will see wavelengths below 13.5 nanometers in the next decade in production deployment. Okay. In mass production, yeah, a dollar is always a good bet. I always bet a dollar. Yeah. Yeah. You know, pick pick your favorite bottle of wine, right? Your favorite, you know, thing that you like, you know, whatever it is because you’re paying me, buddy. Yeah, it’s we’ll see. It’s um we recently saw a startup promising um what they called X-ray. Mhm. Um, and I I was I was doing the mathematics and uh the video on it might actually be out before this um because we’ve just done the final cut and I I spent an hour going through the math of you know how many um how many photons you need to have uh you know compared to EUV you need for an effective use and in order to scan that and how you produce that and you know an EUV machine will do you know 150 wafers per or you know layers and the best I could get um a good sign to do across all the beam lines was uh 40 wafers per day. Okay. In my maths. Okay. Well, you know, free electron lasers, you know, we believe we’re going to be able to, you know, deliver, you know, dose at full productivity, right? you know, an equivalent, you know, of uh on the order of, you know, in excess of 2,000 watts, you know, compared to today’s 5 or 600 watts in the EV machine. They just announced a,000 watts. So, you know, we believe we can go much, much higher, right? Uh, in that also the spectral purity of a free electron laser, right? You know, I mean, it’s, you know, it’s the most, you know, instead of splattering tin, right? We’re producing finely precision light sources, uh, as well. You know, this is a pretty magic technology. So that’s one example. Uh you know the other a small story I found the supplier who provides tin for ASML and they wouldn’t tell me anything. I reached out to the supplier and and and you know eventually worked out that you’re you’re using like less than grams per day. Yeah. In an EV machine, you know, it’s the cheapest part of the machine. Yeah. Is the tin. So but yeah so you know I definitely think that next generation light you know we see that light also enabling many other aspects of semiconductors you know new forms of metrology new forms of packaging uh as well I think it will initiate a whole set of equipment development as well as new materials development you know for a mass refraction uh uh technologies new forms of chemistry as well can be improved if I give you more photons you know I’ll be able to move to harder resists which you know natural benefits in and of themsel you know not just next chemistry when I go to 4 and a half nanometer you know wavelength of light you know so yeah I’m I’m a optimist there right and of course you know as we’re seeing just extraordinary you know capital investments going into today’s low and a you know level of machines you know I think people will start to see oh there’s real value right in some of the critical layers moving beyond that well so so so here’s another one of my critical points and I want to cite the um the work that was done on 450 mil wafers, you know, the 18-in transition. Um, you know, as well as I do between IBM, Intel, and others, this whole consortium came together, solved it, and then the industry turned around and said, yeah, but it’s going to cost us a trillion dollars to transition, so we’re not going to. Yeah. Um, I have a great picture of me holding one of those 18inch wafers, and I asked IBM, can I have it? And then we realized there may be export restrictions on it. Um so so do you think about that when you’re you know considering next phase of Oh absolutely uh absolutely you have to think about the you know I mean the adoption you know where’s the insertion point you know in volume supply chains you know what’s the capital requirements to move it into volume de uh uh deployment you know what will be the business models associated with it you know one of the things with the xlite business model is we’re going to move to photons as a service you know where it will be oh no You know, essentially that sounds terrible. Well, you know, I call it God said, “Let there be light, right?” You know, light as a service and uh just like today you have chemical, you know, uh supplies sitting outside of the fab and energy substations out of the fab. You know, you will have light substations out of the fab, right? It’ll become a utility, you know, going into the fab. And that has so many benefits, right? You know, I use different pools of capital to go accomplish that. I’m not burdening the fab with the capital requirements. Uh also that’s very long life. I can keep upgrading the free electron laser. You know I can attach it to many different types of equipment, you know, metrology equipment, packaging equipment, lithography equipment. You can bend it. Oh yeah. Oh yeah. Yeah. Yeah. Not not that hard, right? Grazing incidents, mirrors, other things, you know, that are pretty uh robust technologies. Uh so you know we we see this entering into a next phase of how you build much much higher productivity into the most expensive capital you know uh equipment industry on earth of semiconductor we have to bring capital efficiency to that uh industry if we’re going to continue to have this dramatic expansion that Moore’s law you know was enabling took a nap and now we’re going to you know wake it from a slumber. It’s I’m seeing this interesting dichotomy here because on the one hand, you know, you you you you paint a very vivid picture about the future of, you know, where we are in semis, whether that’s architecture, whether it’s manufacturing, what have you. But then also, it’s the solve the problems of the day, right? That’s where the business is, right? It’s all very well having your blue sky investments, but realistically, you still need to make money today, not just tomorrow. Um, do you find you you uh now in in this new stage in your career, you’re favoring one more than the other? Well, I’ve always been a tech guy, deep tech guy. So, I’m sort of drawn to those 10-year projects, but I also realize I have to have a portfolio that investors can look at and we have limited partners as well. So, I need some that, okay, this is going to be a two or threeyear win solving a today problem. and uh you know you know a company like power lattice that we announced okay that one’s going to be much faster because that is a today problem right you know as well and some of the other power you know related uh companies hey I think those will have two to threeyear cycles that’ll give me some time to work on my 10year cycles right uh as well you know but even there like a company like snowcap you know I think we’re going to have real commercial use cases around that in two to three years okay right as uh you know as dramatic as superconducting logic is um you know the fact that uh h you know satellites happen to operate in you know 4 Kelvin space okay so saying you know I have a two or four Kelvin superconducting you know it’s sort of a native ambient environment and I don’t produce any space so I don’t have to radiate any heat you know it actually that’s pretty good or or anything you do radiate is a larger proportion of what could affect it yeah so so you know that that’s pretty clever uh in that regard uh also uh every uh quantum project, you know, will want more things sitting in the cryogenic uh temperature range. So, you know, all of those projects will want more of what we’re doing with snow cap, you know, and there are some nice problems that fit beautifully, you know, into a smaller logic, all the signal intelligence problems, you know, give me smaller, you know, uh, you know, footprints of compute where I don’t need as much memory. They’re going to fit nicely into it. So I think there’s going to be great commercial applications even before we hit the, you know, the big winds of, you know, creating AI inferencing centers that are a thousand times more energy efficient than today’s, you know, okay, you know, that’s that’s the holy grail, but there’s some like really good milestones along the way for commercialization. So what does it take to get on Pat Gellzinger’s radar? Well, being super smart, right? You know people that really uh you know but you know I mean people who have really worked on problems you know like the snow cap team they have worked in that problem for 20 30 years deeply academia for it. Yeah. Or national land you know they have real you know really deep expertise uh on problems that are meaningful you know if we solve that problem. Wow that matters right uh you know for it. And then you know having real views uh you know fusion you know not one that I’m particularly excited about you know because you know I need 10 billion dollars to prove right right you know versus okay you know I can you know get you know $50 million to prove critical technology milestones th those are dramatically different problems. So, we like those kind of problems that we can say, boy, we have tangible milestones that allow us to get to the point where I have to go raise billions of dollars, but I have something that I’ve proven that now I’m ready to go raise the billions of dollars. I’m not afraid of the billions of dollars, but I want to have proof points along the way. And and that’s that that’s obviously a point where governments start mattering. Absolutely. So, how often these days are you having to interact with the administration and and and such? Yeah, quite regularly. um uh you know for it and obviously Xite was uh the first recipient of the CHIPS act right under this administration. So quite excited uh about that. uh we have a number of other BAAS and process the new you know process for the chips act from our portfolio uh companies but also other governments as well you know interacting you know Japan hey they do a whole lot in the semiconductor industry so working with them a number of our companies leverage Australia you know we have a number of okay right you know uh so you know quantum you know half the team was from Australia right I have a packaging company that we’re forming uh out of Australia doing 3D packaging you know technology some of them out of the UK uh we have you know presence uh for one of our companies in Germany so you know it really is across the spectrum you know leverage non-dilutive capital you know where governments see you know critical roles for them so there’s a good amount of that and I say you know the network I built with the you know CEOs of the industry combined with many of the governments that somewhat gives playground from you know a pretty unique you know perspective as we’re in this next phase of taking some of these really cool companies and scaling them to be things that change the world. Again, I know back at Intel, one of your things was reinvigorate the America semiconductor industry. Um, I assume it still matters, but it sounds like you’re being a bit more holistic now. Yeah, I’d say yes to that, but also yes that I’m still deeply committed to reinvigorating the US semiconductor uh industry. You know, that’s you know, I’m a US citizen. I consider myself, you know, an American dream, you know, farm kid becoming uh, you know, CEO, right? You know, leading technology in, you know, that’s that’s the American dream, right? Uh, coming to reality. I feel a deep loyalty, you know, to the nation in that regard. I helped to bring the CHIPS Act into existence. Uh, and I do think the absence of explicit industrial policy by our government, you know, renders us at risk, right? uh as a nation, you know, a brownout in Taiwan, you know, has a economic impact. Yeah. You know, that is twice as great as the Great Depression. Y right. You know, a brownout, right? Remember, Taiwan has three weeks of energy reserves on the island, right? This is not a stable situation for the world’s supply chains of technology, right? Yeah. I want more of that in the US. Yeah. I want more manufacturing in the US. I want more of our supply chains for critical minerals, semic uh conductors. You know, we’ve radically underinvested in the energy capacity of the nation. You know, for a decade, you know, one one and a half% uh increase in the energy capacity of our nation. You know, you know, we spent so much time preoccupied on renewables, we forgot that the only thing that really matters is how much. Right? You know, there are 39 nuclear reactors being built in China today. How many in the US Ian? Zero. Okay. 39 to zero. That’s terrible, right? Just terrible. Or is it they they stood up last year in March 90 gawatt of solar? Yeah. Yeah. But again, they’re investing in their energy infrastructure. In the AI age, energy capacity equals economic capacity. Okay. Right. It is critical that we accelerate that. You know, that’s why uh Alva, our nuclear operating company, is so important. You know, that we just brought that one out of ST stealth. How do we get more from our current fleet as well as restart the industry to build a lot more uh as well? Well, you you should come to Europe because what we have more than anyone else is regulation. It’s it’s that interesting dynamic because obviously, you know, I’m Europe based, but I spend a lot of time here um dealing with the companies in in especially in the Bay Area, right? Um it’s I know I know you’ve highlighted in the past for example you know optical packaging up in Scotland um and then you know there’s you know other facilities in you know you said Germany and Germany is quite a hub for that um but I often find that even with all the investment over there all the startups I deal with if even if they’re based there they will slowly migrate to the US. Yeah, you know, and I think there’s two, you know, I think of Europe as having two problems that way. You know, one is I think it’s difficult. You know, Europe has a a lot of low capital startups, but that mid capital range where you need, you know, you move from tens of millions to hundreds of millions. Very little of that happens in Europe. So, I think you know that capital formation and that critical middle phase of a company, right, is very hard in Europe. And then I think the regulatory domain is uh you know ridiculously challenging uh there. So you know I think as a result of that you know there are the occasional arm right that pop through but they’re occasional right uh in that sense extraordinary you know uh support but then also hurdles you make it hard for them to pop through and that’s why so many companies you know and by the way you know some of these are coming to playground right you know and they sort of say hey let’s keep a footprint in Europe but I want my you know doicile to be in the US or in the Bay Area because the vibrancy of the community the research community but also just the speed that we can get things done and capital formation. Well, so so it’s good that you bring up ARM because whenever people say, well, what does the UK bring to the table, right? Okay, we have ARM, we had Graph Core, um, you know, for whatever that architecture was worth in the end. Um, but I’ve always said the one thing we’ve done good at is produced CPU architects and and and engineers. And uh, I I know you’ve probably had the same conversations I’ve had with people like uh, Philip Wong at Stanford. Um part of part of the difficulty of this industry is bringing in new chip designers. Um partly because the US is such a software driven economy compared to a Taiwan or a South Korea. Um even though Playground is very much a you know startup incubator accelerator, do you do more holistic stuff on the how do we get people to um become engineers? So you know the uh you know clearly you know UK UK Europe has a lot of great things going for it right uh in that sense and with that you know I’m quite excited you know about many of those uh you know for it and you know they continue to produce that you know I was also meeting with one of the you know ministers in the UK recently sorry um but you know I was also highlighting that there’s more quantum based work emanating from Europe and UK than maybe anywhere else in the world. You know, why should Quantum Valley, if I analogize that to Silicon Valley, you know, be presumed to be in the US, right? You know, I do think there’s a lot of these major new technologies that are going to be impactful for the entirety of the world, right? You know, that are yet to be developed and where will those centers be, right? in the world and I think many of them could be there. You know, I think it takes capital formation. You know, as I was uh, you know, recently having a conversation, you know, with uh, you know, one of the ministers in that regard, they, you know, gave me their initial proposal, right? You know, for how much capital they were going to put behind it and I said 10x, right? I think they I think they just announced two billion this morning or something. Yeah. Anyway, right. You know, they need to 10x that, right? I said, you know, 10x what they originally said to me. And I said, “Then you’re demonstrating to the world you’re serious.” At that proposal, it’s sort of like, “Oh, okay. That’s nice.” It’s it’s I was at a Ministry of Defense um sort of conference and they asked me to keynote and I did, you know, the wave of new AI hardware and we had, you know, somebody from Adiabatic Computing, somebody computing and then at the end of the day there was a Q&A in the room and they said, “Well, the government’s just issued, you know, a billion dollars for new um, you know, semiconductor investment. Where should we spend it?” and you a billion dollars just isn’t enough to fund startups. It isn’t enough to restart any sort of manufacturing or away from production. And so my response is invest in people to which the response was well we do invest in people and then they move to the US. I said well what if you don’t invest in people and they stay and and and and yeah so so so my question is what can we do to invest in invest in people? I mean it’s all very well having you AI and chip design meaning you can design a chip with 15 people rather than 500 but we still need chip designers. Yeah. Well and I do think there’s you know there’s a lot of great raw talent you know there and I I really right you know when you think about you know the Oxford Cambridge cor just speaking UK I’m just you know in in the US getting more chip designers. Yeah it um but you know I mean the reality is there is a wave back toward hardware. Yeah right. You know there is this enthusiasm again. you know everybody was you know right how many you know AI dating apps do we need right you know right you know it’s just I mean there’s just such a move to software right uh SAS you know different application things where all of a sudden hardware is getting cool again and I think that’s being recognized uh in our universities you know uh in the startup community as well you know some of my friends uh leading some of the big uh venture firms you know one of them confided in me and says I think we forgot how to do hard, right? You know, and to me, you know, it’s like like but that’s where you started. How did you lose your heritage, your foundation? And I think now you’re seeing that whole resurgence, you know, come back, you know, and whether that’s physics, whether that’s material science, chemistry, biology, right? And, you know, chip development, manufacturing sciences as well. You know, I think that renaissance, right, of activity in that space is starting to emerge. And I’m super excited, you know, some of the universities that have never walked away from it. You know, like we just uh, you know, Purdue is now, you know, viewing this as a West Coast, Bay Area location. You know, for them, they never walked away. Some of the great schools, you know, Stanford, Berkeley in the area, MIT on the East Coast, you know, to me, it’s the resurgence of the hard, right? The material science, the hard engineering sciences. You know, this is our day again. So, what does 2026 look like for Pat Gellinger? Well, I want to have, you know, several good exits of our companies. Uh, this Have you had an exit yet? Um, not one that I’ve led, you know, but obviously, you know, the the recent announcement with Higher Labs was, you know, you know, on, you know, a nice step forward, but a couple of good exits of our portfolio companies, you know, want to do six to eight, you know, foundational investments this year, right? you know, companies that really matter and then key milestones of some of the next rounds of our companies, uh, you know, that we’re, you know, bringing them to that next phase, uh, of, uh, reality. Um, and, uh, you know, clearly on my glue, you know, side, you know, putting a platform in place that really starts to scale the faith ecosystem. And then, uh, I got eight great grandkids, so, you know, we’re, uh, you know, enjoying time with them and our family. I have more freedom to invest in them and their lives. And then of course I have about a dozen philanthropies that I spend a lot of time on. So you know seeing some of those milestones as well. You know I have a lot of wonderful things happening in my life. So you’re getting the grandkids into semis. That’s what you’re saying. We’re working on it. Working on it. We’re working on it. Awesome. Thank you so much for your time sir as always. It’s great. Pleasure chatting. We’ll put a link in the description to my previous interview with Pat uh so you can see what it was like a couple of years ago when we were only limited to 20 minutes. So, I’m really glad that these guys are giving me time. We’re going to chat again. I look forward to it. And actually, we’re just going to have a uh a press up competition. So, stay on the line. Okay.