Nvidia vs. AMD and the AI Power Crunch

Transcript generated automatically from the episode audio and lightly cleaned up (filler words removed, misheard names corrected where clear). Speaker labels were added afterward and may contain errors.

Rob May: Hey everyone, I'm Rob May. I'm the co-founder and CEO of Neurometric and one of your co-hosts here today.

Ryan Eppley: And I'm Ryan Eppley, co-founder and CEO of RootAccess, the other co-host. This is the AI in NYC show and we're very excited about our guest today because he runs a blog consulting company. Out and we'll we'll figure out what you guys call it.

Doug O’Laughlin: Something new under the sun is what we like to say, but it's we'll say consulting research. I think it's the first full stack AI semiconductor research firm. That's really what we how we think about it. But yeah, we from the you know from the chips to the models to the data centers, we cover it all. That's SemiAnalysis.

Rob May: Yeah. So Ryan and I both been semi- analysis readers for a long time. And so for for people listening, Doug O'Laughlin is the president of SemiAnalysis. Doug, welcome to the podcast and I want to start by ask you something we were talking about in the beginning, which is, people have asked for a long time supposedly AMD's GPU specs beat Nvidia's, but they have almost no adoption. And roughly about a year ago, I think you guys wrote an expose on how hard it was to use the AMD software to get your so to take your model and get it on the GPU. And it seems like that was from the outside looking in it seems like there were lot lots of things bubbling up but that was the catalyst that got Lisa Su's attention got her to maybe put a tiger team on it and now AMD stocks up at 80% this year or something like that and they they seem to be winning some deals but like tell tell us the inside story about like why did you guys decide to do that and did it actually have the impact that I think it had from the outside.

Doug O’Laughlin: I think it did have the impact. I'm not going to say that we're the only reason, right? You know, there's definitely, you know, internal and external will have to come together. I wasn't on that call, but the apocryphal story I think was like the day after Christmas cuz I think we wrote the article the day before Christmas last year and Dylan and the technical team were on with Lisa and in the beginning she kind of like was really dismissive like hey this is like oh you know whatever like oh is this true and essentially knew she was like head of ROCm and doing like a great job. We meet with him all the time on on X and he's he's like yeah it's actually all true and like we like helped bug fix like these little gem shapes and like all these things that like just to make a working accelerator and like our takeaway cuz we spent a lot of time doing and bug fixing and working with AMD is like dude yeah it's probably impossible to inference on it and it's nothing's ever been trained and so we wrote this big thing and AMD took it seriously like there was a huge sense of urgency to catch up and yeah I think obviously that's the outside.

Now, the inside is I know some of the jokes are like we have some friends, you know, in at AMD and they're like, "Yeah, we hate SemiAnalysis. Now we work 996."

and so they they definitely spent a lot of effort catching up and I think they've done a really good job and part of it is just like highlighting kind of looking at the truth of what was actually happening versus like the story right at the time especially in late December or man was that two years ago like the MI300 story when it first came out was we have better silicon with higher performance and more me memory bandwidth and more memory right and you're like well obviously X and Y you would always choose Y you choose the better product but the reality reality is getting transistors to work is a whole other problem.

And it wasn't the AMD story of old, which is like not copying, but doing a better x86 CPU, which had an entire mature library that both AMD and Intel could use. I think AMD thought if we hit parody at a GPU, we're going to have a better GPU. But the reality is they hit parody for the silicon and the transistors and the package, but they didn't hit parody for the software ecosystem. And so that's why, you know, the NVIDIA or the CUDA mode is real because you got there with the chip, but you didn't actually get there with the developer ecosystem or a usable chip. And so we've done a lot of work since then. Actually, inference max is our new our new benchmarking suite of tools that kind of through that show real world performance, not just specs. Like that's something you see over and over. Hey, this one has more flops, this one has memory, it's better, cheaper, whatever. But the real final rubber to the road proof is how many tokens comes out of this GPU at what rate and at what estimated cost. And that's really what we did with the inference max team and product. So yeah,

Ryan Eppley: I mean it's extremely validating that your publication is getting in the eyes of CEOs of Fortune 500 companies. But like take us through how did you even start this publication? How did it hit the limelight so fast?

Doug O’Laughlin: Okay, so I definitely feel like that story has to start with Dylan Patel. I knew Dylan before he became famous. Actually we were just buddies on Twitter u when it was still Twitter right so I started a newsletter called fabricated knowledge and I got like kind of like merged in joined with Dylan about I feel like it's a little under two years a little under two years feels like 10 years ago at this point and Dylan dude Dylan at the time was like and he's always been this way he's like a crazy poster who loves technology and like will do whatever it takes to get to the truth he will argue to hell but he's very focused he's very truth seeeking and that's like an important part of SemiAnalysis

And so, dude, I like, you know, at the time I was working at a hedge fund. Like my background originally is finance. And I wrote my newsletter. I left to go do I left the first job I had to do this big sbatical kind of, it was during COVID. I didn't really, you know, my sabbatical started in 2021. I went like this big hike. Did like a through hike. It was awesome. But I had some free time on my hands. And so, I just I was really passionate about semi-huckers and semic stocks. I was like, this industry is completely changing. And the person who I found most compelling to talk to on a day-to-day basis was Dylan. And cuz he just like always knew what was up. And so Dylan I definitely it was funny. I think one one point of his time in his life, he wanted to do something else. I was like, "Dude, get your head in the game. Make a Substack. You need to get started on this." He was originally on WordPress. And so yeah, I think I was there early, knew who he was, have always kind of like collaborated, worked with him, talked to him. Our, you know, spheres of influence have always been similar. And that's why I chose to join him because like, dude, competing in Sillin is pretty like he he's he's relentless. He's ruthless.

Ryan Eppley: And so good to have that type of person on your team.

Doug O’Laughlin: Yeah, 100%. But I also think what's like what's important is like you know if you want to if you want to go far you want to go with someone else. And so that's what that's the thing we want to go we wanted to make this a sustainable awesome research business right and I think that SemiAnalysis currently like I think already today we are a worldclass AI semiconductor research firm right but like I think we're just starting this journey we have a lot more to go and the entire space and ecosystem is pretty nent it's pretty hot right now maybe maybe even too hot but who knows but I think the thing that's important is we're both really young motivated people and we want to build something Right. And so that's what SemiAnalysis is and what it will continue to be. We're really truth seeking. We actually do research. That's quite crazy, but like you know, you can't pay to be in our Gartner quadrant, right? That's not how our business works. And so, yeah, that's kind of the blurb of SemiAnalysis and how it came to be.

Ryan Eppley: Yeah. And the timing could not have been better. Right after the Chips Act, you guys were kind of primed and ready for it. And then all these other people that weren't resource focused or or research focused, you kind of were already ready for this mass adoption of SemiAnalysis being the go-to source, you know, can you just talk me through like how quickly it went from a niche industry to something that was kind of mainstream on New York Times and other things?

Doug O’Laughlin: Yeah, it's a great question. I feel like it's one of those things where it fuels overnight, but at least when I first was DMing Dylan on, you know, on Twitter in like 2018, I was like, "Oh my god, everything's going to change." Like the long-term bet was Moore's law was dead and that because of that, the new technologies of the future would cost a lot more. And so, we had to do like like something had to change. And so, a myopic obsession with semiconductors is where we both started.

We're like, "Dude, this is the most important industry in the world." And so, watching 2020 happen, right, and having the chip shortage, that was just a confirmation of what I already knew was true. That chips and hardware are the most important techn or like the most important technology that has like ever been invented. We essentially made the information in our brains into the physical around us, right? By we're able to program and use physics to manipulate information at will. That is magic as far as I'm concerned. So we always knew it would be so valuable and pretty much compute continued to chug along and then the you know obviously the magical new thing that came in is AI. A lot more demand for compute and now all of a sudden now all of a sudden chips goes from something that is you know kind of just like a boring plumbing behind the scenes that no one even thinks about because of the CPU world to critical front row most important thing. And I mean it it feels like it's overnight. But part of it is it's been a longer story to get here and I think by having conviction and always knowing that that was true and positioning ourselves to be there, it it it felt inevitable. So yeah.

Ryan Eppley: Do you feel like the conversation around Taiwan has like increased this a little bit over the years? And you know, how do you see that being a geopolitical kind of play?

Doug O’Laughlin: Yeah, so Taiwan's crazy. Because for better, for worse, all the good chips are made there, man. It's always it's kind of like one of those interesting facets of the entire semiconductor industry and the entire semiconductor like conversation, right? Is that we outsourced chip manufacturing overseas and we literally lost the keys of the kingdom and no company has ever come back from falling off

Rob May: which Andy Grove apparently warned about right from Intel.

Doug O’Laughlin: Oh dude, you don't even Okay, so my favorite I got this in the Wall Street Journal I thought was really funny. Dude, Andy Grove wrote a book called Only the Paranoid Survive, right? And the entire book is about strategic inflection points that essentially these big things come in your industry that changes everything and it's your job like screw hell or high water. You cannot miss them. Okay. And then Intel proceeds to miss every single strategic point from that point onward, right? They missed mobile and then they missed AI and then they they you know they totally like whatever and so there's like this deep irony, right? But yeah, 100% the these are like really critical technologies and I think the reality is Intel was not an a young company when when it came when these new inflections came and it just wasn't nimble enough to capitalize on it.

So, yeah, I think that that's a huge part of the conversation because let's just say maybe not 100% 85 90% with some of like TSMC Arizona, but 95% of the chips that really matter for AI are made in Taiwan. And 85% or 75% by value, I think it's 80% by value is made in Taiwan. Like, these are the chips that are not even part of AI. It's part of our our mics, our computers, the, you know, the phone in my pocket. Every little bit of technology that is thinking or or working hard it's made in Taiwan and we've kind of lost that key technology.

So it's really important because what happens is that little that island is the you know if we're talking about like compute being the most magical thing we've ever done, right? That is the source of the magic and obviously it is 100 miles off the coast of China and China really does want that territory back for like a long list of reasons, right? But we're not going to I don't want to go into that. And then obviously for America, for better, for worse, is if if Chinese ship production was or Taiwanese chip production was to go down, it would be it would be catastrophic like like truly our lives would be really impacted and I just don't think people appreciate the concentration. So yeah, the magic of making it.

Rob May: And do you think people understand like you just talked about how much happens in Taiwan, you know, we've TSMC's built these places in Arizona, but as I understand it, everything's not there. Like you they can the chips there, but they still have to send them back to Taiwan for packaging, right? And stuff like that. Like there's still a whole bunch of pieces of the ecosystem that even as we are moving things over here like like like walk us through what maybe the way to phrase it is this. What are some of the common misunderstandings people have about the chip industry and how easy this is to solve? Because it I think it's much harder than people think.

Doug O’Laughlin: Man, I could go about I could go on this one over and over and over. Look, my favorite analogy is apparently, and this might I you're going to have to fact check me on it, but I think 2200 blood cells fit inside like a single transistor at 2 nmter. So, you can't even fathom how small of the scale. And this is the precision that we 99.99%. Because if if an if there's like a mis contact of the metal, like your chip won't work. So, so the precision that we're talking about is just insane. It's we've never made a more precise smaller technology at scale and it's at it's at massive scale billions of devices right yeah and we making that whole thing is thousands of steps it takes months for it for it to be done and each one of these steps are like the cutting edge of science fiction like or science but it almost is like science fiction it really is and so each one of these steps have a special manufacturing process a special supply chain a special a specialty in knowledge, right? And so you have all of these thousands of steps and everyone who is the best at it lives in Taiwan.

And so all of the chemical suppliers who are best at supplying chemicals to making semiconductors live in Taiwan. All the everything the entire supply chain, all the like the best builders of clean rooms in the world live in Taiwan, right? There there's this ginormous talent density there that we are trying to recreate, but we're we're having a hard time. And yeah, part of it I mean like part of it is like we just do not build things as quickly here. There's there was like some of during the chips act there's some there was some conversations about how much cheaper it is to build abroad, how much quicker, how much subsidies, whatever. But we have lost that magic for whatever reason. And bringing it back is like you know when you lose the fire I feel like it and bringing it back from embers it's a lot more effort than people appreciate.

Ryan Eppley: Has the chipped act like helped at all? Have you seen that at least make a dent in the problem?

Doug O’Laughlin: Well, the chips act has been completely scrapped as you know and it's mostly gone to Intel. I am I am cautiously optimistic on Intel for the first time in a long time. I can kind of talk about that if you guys would like to go into that. Look, Intel was a story where there was a true belief that if they built it customers would come but the reality is they built let's say 80% of the stadium. Okay. And the last 20% is the PDK which is a pain. And they've only ever it's like building their entire process was built for a customer of one. A oneofone design process. It is like a oh you know if you want to start the car you have to bang the dashboard. All these special little design tricks that only Intel knew because they built and grew it all the way till now.

Rob May: And you're talking about for the fabrication.

Doug O’Laughlin: Yes. For the fabrication.

Rob May: Yeah. Now for our audience like walk them through like there's the design side of Intel then there's the actual manufacturing side.

Doug O’Laughlin: There's two broken sides. Actually, the story is actually more complicated than you thought. So, so on the So, yeah, on the process side was a whole problem too. 10 nanometer, which I guess is seven because it got renamed, was so complicated and had so many little tricks shoved into it, it just got delayed four times, right? And so, what happened was all of these delays allowed the competition to catch up. And also the competition was building an open ecosystem when Intel was building you know a product suited for one. It's like a chef with only one customer and then someone else orders a different meal and they're like I don't really know how to make that.

I've only been making spaghetti meatballs and you're asking me to make a pizza like sure but I just can't do this. And so that's kind of the problem is that they they lost process density meaning the density of the transistors at the same time as another competitor came in with a much broader ecosystem. And then also on top of it on the process side right this is like Intel's place in the world not just in a manufacturing way right because they also lost it on the design side the entire world went the way of accelerated compute GPUs were the next big thing and CPUs a generalized compute was not getting better at a at a quick enough speed and so because of that they lost their strategic relevance and they also lost their process technology in one go that's a terrible setup and so That's kind of that's kind of like the two sides of it.

And so I think the most strategically valuable part of the business is the manufacturing portion, right? They like I was saying, manufacturing these things is like science fiction, but they're still kind of within a viewing distance of the leading edge. And so there's a belief that they could probably get their things together with 18A and then 14A, their next generation process. And I think it's at the point where the process is mostly fixed. But the issue now is how do you make it that how do you make it so that like this process that you are selling to your customers is exactly what they're getting right you know it's like a it's a it's a quality control thing where it's like you don't have those idiosyncratic bugs that Intel had designing in this process in their chips for like 20 20 years or 40 years right and that's kind of their challenge and so they got there they're if they built it they would come they mostly built it people weren't really going to come because why would you risk your most important product on an unproven process technology with your former competitor who's been underexecuting for the last 10 years I wouldn't

and so what happened was the chips act got completely rolled into what it was a ginormous investment into Intel okay and so the carrot was if if they build it will they come now welcome the stick okay you got to invest in America the largest shareholder in Intel is actually the United States 10% Yeah, 10% is the United States government. And I think the need and the force via tariffs, whatever, is a really good counterweight to push customers to from the top down to force them to choose Intel. And I think that that's the unfortunately the right way to do it, right? These relationships get built from the top up. And someone making a big bet in a an organization that is not performing doesn't make sense to me. But someone making an a big bet with knowing the counterparties with I guess the human capital element where like Trump is saying like you got to do this or else is a big deal and I think that that's I'm the most hopeful I've been on Intel for a long time.

Ryan Eppley: All right. So we're cautiously optimistic. You know Rob, you've been talking so much about these you know I guess newcomers to the party. Like what walk through like your thoughts on this and let's get an expert take on.

Rob May: Yeah. I mean it's interesting. We were talking about before the show some of the investments I've made. So former ASIC designer and I think a lot of people that watch the show that aren't technical don't realize that there's lots of ways to do compute right analog neuromorphic data flow architectures all this kind of stuff 95 like pretty much everything outside of like DSPs and some specialized chips like we use what's called a von Neumann architecture and it's pretty standard and that's how CPUs have been GPUs are kind of that way whatever but anyway the point is up until AI workloads came along there was no need to try all this other stuff CPUs got faster programming languages got higher and Now these AI workloads like they stress the hell out of even like the cutting edge GPUs and so we've seen this explosion since maybe 2013 2014 in new chip architectures etched senova cerebras mythic rain positron we talked about a bunch gro

Ryan Eppley: yeah yeah grock with a Q not the K it's kind of confusing

Rob May: so tell us you know why don't we why don't you talk about at a high level like what do you guys see is there an opportunity here for one thing, you know, so I guess three-part question. Is there an opportunity for alternative accelerators? If so, what segments of the market? Number two, why doesn't it seem like Nvidia has some skunk works project? Like I dude, I scour Reddit. I looked, but like I cannot find any evidence that they're working on something crazy new beyond their next version GPUs. Seems like they should be. And then number three, let's talk about Cerebras, which are seems to be doing the best of them, but you guys aren't a fan of. Just lobbing up a simple question for you.

Doug O’Laughlin: Yeah. Okay. So, let me show let me let me like let me like think about this. So I think one of the biggest scops if you guys are familiar with the term is that the GPUs are general purpose anymore, right? Or like there's they do graphics, right? In a long in a lot of ways they've been like since at least the Hopper generation a little bit, but definitely the Blackwell generation. You can argue that Blackwell is a transformer processing unit. It's been very specialized and f focused on making the most of transformer throughput. And so there's this belief because if you look at like Bitcoin mining for example, right, that's the old analogy is that the ASIC will have another 10x speed up. I don't know if they quite have that much architectural speed up over and then also the problem is too GPGPU which is general purpose GP GPUs has flexibility so that if an architecture changes you can still recycle some of the processors that the what are they called the t the tensor cores right yeah the tensor cores to still be productive but it doesn't have to be like a very defined matrix multiplication shape

And so like that might be the problem for example with etched right they're like hey we're going to bake a hard bake a transformer in right but what happens if like some kind of clever algorithm trick state space models yeah state space models there's so many there's like many many many ways this can go that it it is not like just the biggest systolic array possible that is like the correct way forward and then all of a sudden their bet ends up becoming brittle and so that's kind of the thing you can argue that this is you can look at history of compute and C kind of a folded in fractal of the same problem. Why did CPUs win? There was actually a lot of ASICs in the beginning in the very beginning. Why did CPUs win? Because they got so much better every single year and there was a and they supported a broad use case, right? And so as long as Nvidia continues to have a killer pace, what happens is you don't need to have these like orders of magnitude increase. You just hold a steady pace until your competitors run out of money and fuel. And that's what's been happening and what we've seen.

So a lot of the early ASIC companies, they they like tape out a first chip. As all first chip tapeouts go, it's delayed. They have to do respin. And their original thing that was going to be 10 times better than the A100 is now kind of competing with the H100. And then because you know when when it came out the H100 is like you know just you know just 6 months away. And then they also don't have a software stack and they don't have like all of the like these nice to have developer tools. And so by the time you get it and you purchase it and you have like a hundred of them H100's already exist and you have your entire ecosystem completely read written out and ready to go. And so that creates a lot of workload inertia. And so that's why like Jensen internally is obsessed with making sure it's backward compatible so that as long as they continue to push it forward every all the work they've done to this point doesn't need to be rewritten and then that can keep the inertia into the into the GP GPU ecosystem. And so that's kind of the bet.

I think there isn't a special skunk works that I'm aware of, but they are pushing the limits of what is possible at a transistor package data center level, right? They are dragging forward other people. And I think a lot of the innovation that we've been seeing from Nvidia, ironically, is not at the nanometer transistor magic level that I've been like I love and like truly appreciate, enjoy, but it's actually at the package level, at the at the rack scale level, right? Being able to move some of the power. If you think about it, this is just a ginormous distributed problem. And they say it's like it's not just a single chip, it's over a whole data center. One of the things you don't want to do is you want it to be as dense as possible. It's the same problem as Moore’s law. The reason why Moore’s law works so well for so long is that if you move two transistors closer together, it takes less energy to move between it and it's faster because it's like, you know, it's less distance.

That same problem is what they're doing with liquid cooling. By moving and densifying these racks by having more more MSI like like millions of square inch of scale out on on the rack level and then having more liquid cooling, you're able to do what you're effective by moving it all together into a denser and denser rack. And what they're doing right now that they're talking about is we're talking like I mean mega I've heard me megawatt racks. That's crazy. A megawatt like you know a thousand that's a gigawatt 6,000 you know so essentially like so like if you think about it 6,000 racks my favorite analogy I've been using is all of New York City since we're an AI NYC show is like 6 gawatt so that' be 6,000 racks which is not that many yeah not that many compared to what we need for for AI yeah

Ryan Eppley: so are you bullish on nuclear powering these data centers is that something that you've like kind

Doug O’Laughlin: I Okay, so yes and no. It makes a lot of sense in the long run for power generation. It's like the most obvious thing, but making nuclear power in the United States is really hard in this like if you could argue it's the exact same problem we have with chips. We completely like let it go. We haven't made a new like the last nuclear reactor in that was made in the United States, Vogtle one and two, bankrupted two companies. No one wants to put down the dollars because I think they both were like 50 60% over cost and they were like $10 billion projects and so like that's a number that will put down almost anyone. That's I mean honestly the SMR revolution is more about small because you can eat it right you can pay like someone is willing to say I'm I can pay I'm in for a billion I'm not in for 10 you know and manu like modular but also manufacturable. I think that that will definitely happen and I think Trump has made it a high priority and it's very clear the DOE is extremely extremely focused on on nuclear power generation.

That will happen but the timelines for that is probably outside of what is currently happening right now. So I think that that is a 2030 story at best and that's like for what is going on and how quickly things are moving. That's kind of outside of the forecast scope that we are really focused on today. But I definitely have been seeing a lot of new announcements. A lot of new exciting technology companies. For example, are you familiar with HALEU Halu? HALEU. Yeah. They're starting to be they're starting to even downgrade some of the weapons grade nuclear power nuclear fuel in order to make more normal nuclear fuel for reactors. So, we're going to see a lot. But, one nuclear power plant is 1 gawatt of the old one. SMRs are actually often much time a lot of times smaller. And just like all the other problems, building things in America is extremely slow and getting a nuclear power plant from start to finish is 5 years. So we'll see that later.

Ryan Eppley: Well, I'm rooting for Valar Atomics. I know you've probably, you know, seen them come across the headlines recently, but I really want them to win cuz they hit some type of criticality in the last couple months here. So hopefully there's some big news in 2026. That makes you think that that it can happen before 2030s.

Doug O’Laughlin: I mean I'm not aware if they have a reactor approved by the what is it the FRC? Is that the correct one?

Ryan Eppley: Yeah. No, the NRC.

Doug O’Laughlin: Yeah. And so you have to have the reactor like that's actually one of the headwinds, right? It's so long for these things to come into play. But yeah, I'm really optimistic because I mean it it's obvious, right? It's an obvious technology with like that's really wellmastered and we've done it at scale so long. We should do it again. It's not it's not rocket science. In fact, it's like one of our most stable old technologies, right? So yeah, I'm very excited for nuclear power in America. I definitely think it's going to happen in a big way. I feel like we need it to be even an an inch of green for how much power is coming online.

But I think in the near term, the real story that we're focused on at SemiAnalysis is focused on fuel cells as well as on-site power. So specifically, this new this new grid u mandate is kind of saying no to large loads, meaning data centers. And they're specifically saying you if you want to make a data center in this area of power, you have to bring your own generation. So it's BYOG, which is like very like this is like in the meetings of all these grids, BYOG. And so BYOG is going to be a big deal because what happens is if you BYOG you can shave some of the peaks. I don't want to go into like the utilization

Rob May: talk about people buying jet engines for this.

Doug O’Laughlin: Yeah. Yeah. Yeah. No. So the aerodynamatives. Yeah. So like yeah all the aerodynamatives are a perfect example of this. And so yeah, we're seeing a lot of this and we think it's a really obvious play. We're going to continue to see this until I mean I mean we're talking like tens of gigawatts that's in the pipeline. And so you got to do you got to pull if you want to pull that forward you have to do things like that. And we I expect to see a lot more especially in BYOG land. So yeah, I'm excited too.

Rob May: What do you think about let's transition a little bit to you? You also write fabricated knowledge which is a little bit about the stock market and everything else. There's been the big debate of the last three months has been is there an AI bubble? Is there not? You know I would argue AI is a very heterogeneous thing. So like where is there an AI bubble where there's not? But what do you think about like let's put that argument aside. Maybe there is, maybe there isn't. What do you think is going to be the thing that's going to drive the next wave of people getting really excited about AI again in the public markets? Is it just psychologically getting comfortable with something happening? Is it a new technology? Breakthrough use cases?

Doug O’Laughlin: I think I think Okay, so one, we're like still 5% off alltime highs right now. And so like I do think like guys, relax. It's going to be okay. I don't know why the hell my phone is. That's fine.

Ryan Eppley: Do you do you want to check it out? We edit it. You can pause it.

Doug O’Laughlin: No, I want to just silence it, dude. I think I know what it is. Okay. Yeah, I'm in a ginormous group chat. And for some reason, okay. So I think one of my favorite sayings cuz like I'm a stock guy like the for first and foremost like that's my OG background is nothing changes sentiment like price. Nothing changes sentiment like price. Okay, people are calling me like, "Hey, is it all over from the industry side?" And I'm like asking them like, "Do you think it's all over?" They're like, "No, but this stock market stuff is worrying me." And that's what that's the reality. Nothing changes sentiment like price. And so we've seen some of the price come off. Part of it is like year to date, we had a very like kind of a crazy frothy market in certain parts of the market.

Like like Oklo is actually a perfect example for talking about nuclear power. I think they're pretty far from revenue and they had like a $50 billion market cap on zero revenue. That's insane. That's bigger than like real big companies. That's like bigger than CoreWeave who has billions of dollars of controversy whatever but like has billions billions billions plural tens maybe of backlog. Aqua had nothing. So there's examples where you look in the market and you see like massive massive massive like speculative froth and I think that that's like kind of healthy. We I'm not surprised that this happened.

But I think that there is a little bit of a vibe shift some of the TPU things that that a lot of people are focused on the TPU side specifically. We wrote an article about this recently. The TPT TCO is going to be really good for people who know how to use it, which is Google Anthropic. And so we're starting to see a lot of adoption there. And it's very like Gemini 3 is being so I think what what what's been happening a little bit is that OpenAI is the one who's saying I'm going to go spend a lot of money, $1.4 trillion, right? And they're they're really in a broad open ecosystem. They'll rent from whatever. They'll they'll partner with whatever.

And so the multiplicative effect of like GDP up or like spending up is much higher at OpenAI versus Google who has the TPU. It's totally vertically integrated. They do the data center, they do the chip, they do the model, right? And so the spend there is a lot more constrained in a way like obviously big numbers, right? But we're not talking like oh we're going to build a gigawatt a week like crazy things, right? And so I think Gemini winning makes kind of shows that there's a competitive threat to the leader and if the leader all of a sudden is on its back foot, maybe there'll be less spending. And so that that really dampened the AI trade in public markets.

But I think the real I mean part of the story too was like we weren't going to cut there for a little bit, but apparently we are going to cut rates. I think if I'm being super candid, we are probably in the beginning stages of a bubble. I believe that if you look at the past, there's a really good book on this called technological revolutions and financial capital.

Rob May: Yeah, I caught a lot of press.

Doug O’Laughlin: Yep. Goed book. One of my favorite ever. And so what's funny is I actually, you know, spent a lot of time reading about bubbles and capital formation and all this stuff. And I've been reading a lot about railroads actually. And if you go look back, every single technology that happens like this that meaningfully changes like let's say supply distribution cost down, right? Like railroads were a ginormous innovation over canals. Like there was a canal bubble, guys. There's a railroad bubble. The internet was a ginormous innovation over print, phone, fax, all or like all these things where it just wasn't as efficient, right? These giant bubbles, capitalism is extremely margin focused and very focused on like return on invested capital until every 20 to 40 years it loses its mind getting really excited about a new technology and it says, "Actually, I'm going to bet the house." And so we see this in front of every giant technological change. And for better or for worse, it seems to like weirdly work because the United States and capitalism continues to be first in line to these new technological regimes because it says it. If we do not if we do not win this, we you know, we're not going to be relevant. And so all the powers and things that be are pointed in the same direction.

And that's really important part and the powerful thing that happens every 20 to 40 years. So, I'm do think there will be a bubble, but if we're going to look at the earnest and real signs of what makes a true bubble a bubble, I would say leverage is the thing that we're we're just starting at, but historically you're talking about lots of leverage, right? The telecom bubble specifically was known for companies buying each other with billions of dollars of debt and not a lot of cash flow. I think the thing that's different and maybe shocking because the sky, the upper, the upper limit of what could happen here is like kind of staggering is that the most profitable, valuable companies in the history of capitalism are the ones leading the way. They are not even close to being levered. They actually all have net cash even with some of the recent massive reserves.

Rob May: Yeah, massive reserves.

Doug O’Laughlin: And their free cash flow is still extremely positive. So, it's kind of like, you know, it's like pick your own poison. You know, the most crazy version of this is everyone's free cash flow goes to zero. You go on a debt binge and then like, you know, we're talking trillions of dollars in investment and it's like crazy. Okay. We're not there. I think we're not there. Now borrowing debt to do large investments for future thing, we are there like the meta deal for is a good example of that. Oracle is getting very levered to win all the their RPO from from OpenAI. And I think you're starting to see some of the beginning signs and of what looks like a classical capital formation.

Ryan Eppley: Well, I'll be interested to hear your take on this in a couple years. We'll see where the bubble is or where it isn't, but you know, we're very bullish on this, but we're also very skeptical. So, it's great to hear someone like your expertise kind of weigh in. Before we end the show, we always ask our guests something New York-based. Our question of the week for you here in New York is here much recently. What is your favorite New York myth? Think about the lore or something that's a misconception. What's your favorite myth about the city?

Rob May: I'll tell you mine while he's thinking. Yeah. It's the alligators in the sewer, right? Like that that that can't actually

Ryan Eppley: I haven't heard it in a while.

Rob May: Yeah. It's a thing that people say that a bunch of rich people bought alligators and they escaped, whatever. It's like it's so stupid, but there's there's a lot of myths like that that go around, I feel like.

Ryan Eppley: Yeah. What do you got, Doug?

Doug O’Laughlin: I think that like the okay like the one that I think of like dude I'm relatively new to the city to be candid. The one that I always think of is that like people are like oh everyone's like super mean and cold here. I don't think that's quite the case. It's just that there's a lot of people so you can't be like can't be nice to every person dude. There's 100 freaking people and like you have to you have to be considerate. I do think people are like oddly more considerate otherwise you like people it just wouldn't all work. But there's like an impersonal because there's just so many people. But it's not like it's not like this like you know I'm from Texas originally. You're like oh these southern people are really nice but there's like a bless your heart to it. You know there's like a double-sided to it. I think people if they don't like you they'll say it to your face and there's like a it's like a little cleaner. They're kind in their own way and I kind of like that. I think that that's been I guess different for me from what I've seen.

Ryan Eppley: Nice. Yeah. Well with that we're going to end the show. We appreciate you tuning in again. And this is AI in New York City and we'll see you next time.

Rob May: Thanks for being on.

Doug O’Laughlin: Yeah, thank you guys for having me.