Dylan Patel on the infrastructure powering the AI revolution | The Next Big Thing

All right, hello everyone and welcome back to our next episode of The Next Big Thing podcast. Here I'm joined by my colleague Blake Hyman and Dylan Patel, our new partner, founder of SemiAnalysis, the research group. It's an exciting day that we get to sort of go through the current state of the AI infrastructure landscape with Dylan. You've probably heard Dylan on many different podcasts. I follow his content all the time, as well as the newsletter on the SemiAnalysis site. Most recently, one of the ones I saw was about data centers in space. So if anyone is interested in that topic, they have a great piece, a lot of detail out on that. But Dylan, I would love to hear a bit about where the idea for SemiAnalysis sort of came from. I know in the Substack community these days, people are talking a lot about the revenue of the firm and the success the firm has achieved. But I feel like a lot of times you sometimes see the current level of success and forget about the journey and the origins and how much work probably went into all of it.

Yeah, so I think the origin of SemiAnalysis really comes from shitposting, posting online in, let's say, a less serious manner, right? So when I go back to what were my first posts about semiconductors, They were when I was in my tweens, right? And I was just posting on the internet about chips, about smartphones, about smartphone displays and smartphone SoCs and all these other things. Before I ever even had a smartphone, I was just obsessed with them. And same with gaming hardware, right? PC hardware, console hardware. I was just posting about this stuff all the time on the internet on various forums. And by the time I was 12, I was moderating a lot of forums related to Android and Apple and Google and Intel, Nvidia, AMD, like the hardware Hardware topics of the world, various forums on Reddit related to these things. And so that's sort of where the origin of it all comes from is I was just, I've always been a poster. I've always posted my opinion. I've always replied. I've always thought, taken comments. And, you know, I think one of the things the people on my team who, you know, now we're a 90-person organization. So I actually have like, you know, people in marketing, they're like, Dylan, stop replying to random bozos on the internet. Like you're making us look bad. It's like, it's like, I just have that burning desire to respond to anyone on the internet when they try and criticize me. And so that's the, you know, maybe, maybe it's a bad thing, but. Basically, throughout my teenage years, I was moderating these forums. I started investing as soon as I started making money in my later teenage years. I was a quant for 2 years, and then I started my firm. But the whole time I was posting, posting, posting, posting, posting. I had anonymous blogs, I had anonymous posts. But then in 2020, I got fed up with my job, disillusionment of being a quant is not as amazing as it seems. Yes, you make money, but it's not as amazing as it seems. And I kind of quite quit my job and started my company. And I wasn't exactly sure it was going to go. But I was posting on a WordPress website that I made. And I was posting under my real name. And I was posting sort of this mix of technology, business, finance, supply chain, sort of the things I was most interested in, because I grew up in a small business, right? I grew up in a motel, my parents owned a motel in rural Georgia, and we lived in it. And I kind of just know business. We later had gas stations too. And so sort of lived in it, grew up in it. So I always loved business. Supply chain was always interesting from the perspective of investing, from the perspective of how things are made. I think that's always just been a knack, like how are things made? You know, technology aspect, of course, is super exciting. And the finance aspect are super exciting. So, you know, taking the combo of that, you know, my first posts were things like, you know, at the time China was banning or the US was banning Huawei from access to TSMC. And then, so my first post was actually about how MediaTek was the biggest winner because Huawei was number one market share in China for smartphone chips and smartphones in general. Obviously, that was going to tank because they no longer had access to TSMC. And so the US market thought, oh, Qualcomm's going to win. But MediaTek, which is the Taiwanese firm, I thought, you know, they're actually going to win a lot more share because, you know, it turns out geopolitically, China would rather buy from a Taiwanese firm than from a US firm, given we just banned Huawei. And so both firms benefited, but MediaTek benefited a lot more. So it was that like mix of like, technology, supply chain, finance, geopolitics, all these things sort of melded together. And over the coming years, I converted my WordPress into a Substack at one point, started charging for it, and I was writing about all these topics across the entire semiconductor and AI supply chains. Given I followed AI a lot, I did a little bit of AI when I was a quant, as well as following semiconductors out of a passion. And that just grew and grew and grew and grew. For 4 years, I traveled all around the world and I went to every conference in the world. I was going to 40 conferences a year. I didn't live anywhere. I was just going to every conference I could, whether it be an AI conference like NeurIPS, ICML, ICLR. So these are sort of AI researcher conferences, mostly researchers and some companies go there and present their research, all the way downstream to random conferences for, let's say, chemicals that are inputs into the semiconductor supply chain. So I'd go up and down the stack, whether it be servers, networking, fabrication, AI, just all up and down the stack. I'd go to 40 conferences a year. There'd be some that are super niche and there'd be 300 people there and they only spoke Japanese except for like 5 people. And I'd be like, well, shit, whatever, right? That is what it is. And then there'd be others where it'd be 10,000, 20,000 people and it'd be huge, right? And it'd be the whole spectrum and continuum. And so I was able to go across the whole ecosystem. You go to a conference 3 times, you actually know the language now, you know people there, you build all these contacts, you can ask them questions. And sort of, you know, I developed this whole ecosystem and learnings and I was covering, you know, the inflections at each of these, right? I was very curious technologically, but then once something stood out, whether it be technology or supply chain, I learned from a conference, I knew where it would lead to on a supply chain or finance basis. And so I was writing about like the whole mix. Sometimes reports would be centered around technology and no one in finance would care. And other times they'd be like, wow, this is the bottleneck, or this is the inflection that's happening, or this company is going to gain a ton of market share because they're next generation technology. And I would call it before anyone else on the street, before any hedge fund, before anyone. And so sort of that was, that was like the start. And then as that Substack grew and grew in 2022, I started hiring people. My first 2 hires were people I knew from Discord for years. But then after that, my 3rd hire was Myron. And he worked at a hedge fund before, and he was moving to Japan to live with his wife there. And so he was, he was sort of a free agent. I'd put in a post about how, you know, it's interesting. This is a very interesting post at the time. In early '23, memory was the biggest loser from AI. And the reason was because the amount of memory that AI chips used, uh, and AI servers used versus regular servers was a lot less, right? Regular servers, about half the BOM was AI chip— was memory. In AI servers, it was much less. Part of that was NVIDIA's margins are much higher. Part of that was, you know, there's a couple different factors. Um, of course, NVIDIA's next-generation chips have increased the memory content dramatically, um, and now it's like, you know, way more. But at the time, it was saying memory is the biggest loser. And in the paid section, I said, hey, I'm hiring. And Myron reached out, and he was the first person from a hedge fund background, right? The other 2 people were technical backgrounds. And so as soon as he joined the firm, we started making all these models. And we really converted the business from, you know, we still do the newsletter, we still post a lot of amazing content there, more than ever before. But we converted it from that to being one around selling information services, you know, selling these reports, selling these datasets. And that sort of, as that started to happen, the ball started tumbling down the hill. And from 2023 to 2024, I went from, you know, 2 to 7. And then from '24 to end of '25, to beginning of '25, I went from 7 to 20. And from '25 to '26, I went from 20 to 60. And now this year, we're at 90. So we've added 30 people this year. So it's just been a ball tumbling down the hill. And we've just added sectors. You know, I've always been interested in everything. But now I've been able to add people that are experts. So, you know, I think the most exciting thing about SemiAnalysis is, I don't know another firm that has the level of expertise and a concentration we do, right? I have people who have worked at ASML, Applied Materials, Lam Research, the equipment companies that build wafers, all the way upstream to people who have worked at Intel and TSMC and NVIDIA and Microsoft and Amazon, of course, but also people who have worked at OpenAI on model. Someone has worked at OpenAI on model. Someone has worked at Tesla on FSD. Someone has worked at Cohere. We've got people who have worked on the model layer. And then on another vertical, we've had people who have worked on data centers. There's someone at my company who built a power plant in Kazakhstan. It's like we just have this most insane talent density that's awesome. So half the firm is people who have done engineering across the industry. And then the other half of the firm is either ex-hedge fund or ex— just like random people from the internet who are super passionate. You know, I found them on Twitter or Discord. I'm like, you're smart, come work for me. And it works. And so this is sort of like, you know, what's built up. And now Subby Analysis has like got many different lines of businesses, you know, obviously data services, consulting, information services. You know, we have the newsletter, we're doing all these media, we're having a conference soon that's going to be big. And, you know, all sorts of different stuff that we do. And it's just a hell of a ride.

And speaking of the ride, I had a moment, Dylan, where WisdomTree and SemiAnalysis have been on this journey for a number of months working together. And so NVIDIA GTC happens in March, as it does every year, and I'm sitting watching the livestream here in Charlotte, North Carolina, and I think there's 55,000 other people watching the livestream, and suddenly And there's 20,000 people in a stadium.

There's 20,000 people in the stadium, dude.

And he's referring to you directly that you were like sandbagging— you said he was sandbagging a certain number and your charts like right up on the stage. And so admittedly, I had a moment, I guess, probably on your behalf, basically watching the CEO of the biggest company in the world basically refer to your research and how you were criticizing, you know, his take on some of the numbers. I'd love for you to sort of I guess you were in the stadium from what it sounds like there.

Yeah, yeah. So that moment was quite surreal, right? So basically one of the things that SemiAnalysis does is we actually have a number of engineers and we do open source benchmarking of all the AI models that are open source as well as all the hardware. It's a pretty awesome effort. So there's a number of engineers on my side, but we collaborate with the industry heavily. So we get hardware, you know, so we have over $50 million of hardware donated to us from companies like OpenAI, Microsoft, Amazon, Google, CoreWeave, Nebius, Crusoe, all of the major clouds you can think of that have donated us— Oracle have donated us hardware that we run these benchmarks on. And so we have 8 different kinds of GPUs, all H100s, H200s, Blackwell, AMD's various GPUs. In addition, we have TPUs from Google and Trainium from Amazon. And what we do is we run benchmarks on the latest version of software every single day. And so why that is, is because every night, A CUDA version could release, a PyTorch version could release, a driver update could release, inference engine, VLM, SG Lang version could release, etc. And so we run these benchmarks every single night on the entire curve of how fast do you want the tokens versus how cost-effective do you want them on the optimal scenarios. And we run all of this every single day. It's like an automated benchmarking suite that runs. And Jensen, when he originally launched Blackwell, had claimed it would be a 25x improvement. And at the time, no one believed him, right? It's Jensen, right? He's marketing. And even us at the time, we were like, oh, okay. We were more bullish than ever. We were like, we think it could be like a 15 to 20x improvement based on some of the simulations we were running. Because we had a simulator. We have a simulator for performance. But as we built out this inference benchmarking called InferenceX, we got to the point where actually, oh wow, in DeepSeek V3, Blackwell is 30x faster than Hopper somewhere on the continuum. And so then you ended up with this I emailed him as soon as we had the results and they automatically published onto the open source GitHub. It's an open source collaboration. NVIDIA people helped and they knew, but I highlighted it to him. I was like, hey Jensen, I emailed him. I was like, hey Jensen, back in 2024, you said, or back when you launched Blackwell, you said 25x and everyone gave you crap. Here's all the people that gave you crap. Even I gave you crap. I was like, there's no way it's 25x. It's maybe 15 to 20x. And a lot of people were like, no, no, no, it's like 3x. So we were quite bullish, but Jensen, I was wrong. You were sandbagging. It was 30x. And so he took that and, you know, I didn't know that he was doing anything with it. I'd heard from a couple customers. Someone at Meta told me there was a meeting they had and Jensen was using that as proof that he doesn't sandbag numbers. He was like, because he was telling about the next generation chip. And anyways, this all happened. And anyways, I didn't expect it to happen on stage. In addition, in InferenceX, we created this belt. It looks like a WWE belt. It's a WWE belt and it says Inference King. And we sent it to all of our collaborators. We sent it to our collaborators, we sent it to NVIDIA, we sent it to AMD, we sent it to some of the other folks, SG Lang, VLLM, all these different people who helped us with the benchmark, people who donated hardware, whatever. Because it's an open source effort where I'm spending a couple million dollars a year on engineer salaries and then other people are spending millions on hardware and donating it or millions on engineer salaries and donating it towards this open source effort that we run. And I sent them this belt, and he has this belt on the slide, and he had held it up, and then there were our charts, and he's on the slide for 5 minutes talking about how Dylan said I was sandbagging, but I wasn't, and our performance is the best. And it was just, it was such a surreal moment.

Oh, man.

He talked about us longer than anyone else in the entire presentation. The only other person who he even talked about as much was OpenClaw. Which has obviously taken the world by storm. So it was an incredible moment.

Yeah, Dylan, so you mentioned a couple of things there. I think it's quite interesting that you mentioned open source. And now that there's kind of maybe transitioning a bit towards, I'll say, some of the recent developments and maybe a bit of, I'll say, the markets, there's been kind of these discussions around the actual inference efficiency of some of the open source models versus the closed source models. And there's also even, you know, still to this day, a lot of investors that are like questioning, say, the ROI on all of this. I mean, just in the last week or two, we had some, I think it was a Bloomberg economist kind of talk through, you know, how a lot of these AI initiatives may be, you know, I'll say potentially failing at some of the firms out there. And I know you've highlighted how your firm is using AI extensively and really leaning into You know, giving lots of, giving your employees lots of access to tokens. And then obviously you just highlighted that you're hiring. So I'm just curious, kind of what's your take on this kind of like end demand, the, you know, and kind of this, the fact that this end demand really does drive, I'll say this big buildout that we're seeing that is kind of full of all these constraints that have been at least over the last month besides the most recent couple days in the market. that's driving up a lot of the, I'll say, different stocks tied to some of these themes?

Yes, I would say a few things, right? When you look at the overarching question here is sort of ROI and are the companies making enough money from AI? Is this gonna continue? Are the people using AI actually getting value out of it? It's this overarching question that a lot of people have. And when I look at it and when I think about it, right, there's a few ways to dissect it, right? One, first and foremost is Anthropic is free cash flow positive. positive, and they are profitable in Q2. Even in April, they closed April's books, they were profitable, they were free cash flow positive. In May, they were free cash flow positive, they were profitable. June looks like it's going to be the same way. It's not fully closed yet. But at least for 2 of the 3 months, they've been free cash flow positive and profitable. Their recurring revenue has soared past $50 billion ARR, and they're doing fantastic. But I think that's one side of the coin, is Anthropic is printing. Obviously, there are a lot of companies that aren't printing, but they're getting there, right? And OpenAI's revenue has started to inflect as Codex adoption has grown, and others as well. So these companies are all getting much more profitable, and Anthropic's gross margins are really, really high. They're above 70%. But ultimately, That's one side of the coin. The side of the coin that people are starting to, and you were alluding to, is sort of what about the company spending on AI? At least at SemiAnalysis, we went from our annual— I like to call it like, it's not an ARR, annual recurring revenue, it's ARS, annual recurring spend. And so my annual recurring spend in December and in November before Claude Code— Claude Code really started to take off in December for us last year, but our annual recurring spend in November was less than $100K. What we had was that we had a subscription to every model, or we had a subscription to the $200 tier for ChatGPT for every user. And that was about it. So we were spending less than $100,000 on— if people wanted xAI or Claude, we'd give it to them as well. But Standard was giving everyone the $200 OpenAI subscription. And that's what was the state in November, which was— I think we were on the bleeding edge even then, But then Claude Code really started to hit its inflection point with Claude Opus 4.5 and 4.6 and so on and so forth. And by the end of January, our ARS, right, our annual recurring spend had hit $4 million. And that's because people were using Claude Code. Now today, it's about $11 million. And the highest, you know, day on— if you take the week and multiply it by 52, a week of spend multiplied by 52, that we're at $11 million. The highest we've ever had was $14 million. So, you know, it oscillates a lot based on what work people are doing. But right now, the average looks to be about $11 million of spend a year. For a 90-person firm, that's freaking insane, right? I just want to be clear. We're spending, you know, more than a third of the spend, you know, employee spend, a third of it is also on top of that is AI. And we'll probably get to half-half or, you know, by the end of the year, depending on how Mythos and other models start coming out and get better and better and better. That's a huge amount of spend. Now the question is, what's the ROI? I think there's been huge ROI because we've been able to build product, we've been able to sell more, we've been able to increase the efficiency of everyone in the company. And so I see the ROI, but I think a lot of companies are questioning, hey, if I'm spending hundreds of— if I take a really good developer and they make, call it $300,000 a year or more, right? If I take a good developer, they make $300,000 a year, There's a lot of devs who make a lot more, of course. Their spend on AI is starting to approach 1-to-1 for good developers. And then for non-developers, the spend ranges, it can be lower, but even at least at SemiAnalysis, a lot of our biggest spenders are people who don't know how to code. And they just kind of tell the model what they want and then they iterate, iterate, iterate, and they get what they want. And so you see this soaring spend per employee. And a lot of companies are now rightfully asking like, hey, our entire AI budget, we blew through it in Q1, Q2, right? For the year, we blew through it already. Now what do we do? And the question is, well, do we cut spend or do we cut elsewhere? And a lot of companies are saying, oh, maybe we need to slow down on spend on AI. But a lot of companies I've seen are starting to cut elsewhere, right? They're cutting other SaaS products that they have used historically. They're saying, hey, we can grow faster, so we'll just do it. They're saying, hey, it's okay to spend on AI, we'll take the hit temporarily. AI keeps getting cheaper even as adoption soars. For what I used to do 6 months ago, AI is much cheaper today. Of course, what I'm doing with AI today is much more extensive than what I did 6 months ago. There's a variety of different approaches people are taking. Some people are even cutting on employees instead of cutting on AI, and some people are clamping down on AI, but those companies are going to get left in the dust in terms of productivity gains and what they're able to build.

Gotcha. And then one of the, I'll say, ways to kind of mitigate some of the incremental cost is choosing cheaper, maybe sometimes less intelligent models, maybe not always being at the leading edge there. And now, like I've said, it's very early, I think, in terms of some of the rumblings there, but I'm curious, is there a point where firms like yours kind of decide, you know, there's some use cases that are more optimal for using, you know, a DeepSeek V4 type of model for a certain type of work. And then obviously you might need to rely on things that cost a bit more for, you know, I'll say things that require a lot more intelligence. Is that part of the calculus here?

I think that's absolutely part of the calculus for some folks. And you have to sort of break out AI workloads into 2 types, right? One is, hey, this is AI integrated into a process that I have. And in those cases, right, it's like, oh, when a customer sends me a document, I check it for XYZ and I put it into the model and the model checks it and it's done. And there I just need to hit some level of quality. And then from there I can stop improving the model and I can start decreasing on cost by waiting for newer models or cheaper models or the cost efficiency. Right. And we've seen, we've seen AI models improve at a rate of about 60x per year in cost. You take a quality level and then a year later it's 60x cheaper. People freaked out about DeepSeek because it was 600 times cheaper than GPT-4. It was about 2 years after GPT-4, so 360x. 60x times 60x is 360x. It actually ended up being 600x cheaper. So somewhere on the curve is whether it's 60x a year cheaper, or 90x cheaper a year. Somewhere on the curve is how much cheaper it's getting a year. But DeepSeek V3 versus GPT-4 was 600x in 2 years. And as you step forward, it's somewhere in that range. And so if you have a workflow and you integrate AI into that workflow, then you get to a quality level and then you go cheaper. The other range of work is AI assistant, right? And that's where I think there's actually a bit of a misnomer. If I am doing my day-to-day work, And I'm asking the model to help me with this, help me find this, help me figure out that. Cost optimization is not actually going to a cheaper model. Cost optimization is oftentimes taking the newest model because the newest model, whereas, uh, you know, Claude 4.6 Opus would take 100,000 tokens to do a task. And it might, it might take a couple turns, right? Me talking to it back and forth. So it might take 100,000 tokens and, you know, 10 minutes of my time. Claude 4.8 Opus can do it in a quarter of the tokens, 25,000 tokens, and it might only take one back and forth. And so the cost is actually less because the number of tokens being generated is less and the amount of time I'm using is less. And so when I look at a developer or I look at someone who is doing intelligence work, how do I reduce the cost? It's actually not by using a cheaper model, It's by taking an existing task that could be done sometimes with the model after fighting with it and going to newer and newer models. And now it's able to either just one iteration or one shot the entire workflow. It's able to do it in fewer tokens. And so what we've seen from 4.6 Opus to when 4.7 Opus, my cost actually fell for a week before it soared back up because people were using it more and more. And now why did it soar back up is because people, you know, they were like, okay, they have to adjust to the new workflow, right? Which is, okay, the work I was doing is done. Let me do more, right? And so it goes back up. And likewise, when 4.7 came out to 4.8, the cost fell for about a week, a week and a half, and then it soared back up because people were like, oh yeah, now I can do more work. And so, you know, you have to measure the productivity alongside the cost. And when it's an AI assistant, actually token efficiency is really important. And so this is sort of why Anthropic has been beating OpenAI is because their models are more token efficient than OpenAI. Actually, OpenAI's models, On the edge cases, in terms of leading science, leading math, leading code, oftentimes can do a task that Anthropic models cannot. But they take 3x as long and 4x as many tokens, and therefore it costs a lot more. And the feedback loop of human and AI is not as rapid, and therefore it actually ends up being worse on a customer perception basis. Because it's one thing to say, hey, model, do this task. And then you come back and you check if the task is done. So another thing to say, hey, I have 4 hours to do this task. And whether it's 1 call to the model and it does work for 4 hours, or if it's 4 calls to the model and it goes back and forth, which one does it better? It turns out Anthropic, when you have this human-in-the-loop feedback loop, is actually way faster and better because it's more token efficient. And so you end up with— that's the main reason why we still remain a majority Anthropic shop. Some tasks people do use OpenAI, and oftentimes the tasks that they let running overnight are the ones that they give to OpenAI Codex. But most tasks they keep with Claude Code. This is one of the interesting factors of what's going on with the models and token efficiency. Cost is a bit of a— it's hard to pare it out, but some tasks freeze the model quality and you wait for the models to get cheaper. Others, Actually, I just want the smartest model because it is cheaper.

Dylan, I was curious on your thoughts, shifting it a bit to the hardware side. And I know earlier this year, one of the newsletters— I've been a big newsletter fan for multiple years, if the audience hasn't realized it. But there was an article talking about memory. Memory has usually been a cycle, meaning maybe it's 18 to 24 months, you go up, 18 to 24 months, you go down. And we know that it feels like almost everything is in shortage. Like if you're involved in a component that goes into a data center, it feels like it's not a question of can you even get the component? It's more, okay, how long are you going to have to wait? Because it feels like in the world today, you can barely get any components. So With your experience, having looked across the hardware side, what do you expect is going to be changing with something like memory, which used to always be this just commoditized product? You ride the upswing, you go through the downswing, and it just repeats going back the last 40 years.

Yeah, so I'm not saying there's not going to be cycles anymore. I think cycles will happen. Obviously, we're in a supercycle where the upswing is crazy, and there will be some downswing. it'll be brutal as well. But the downswing, obviously, you know, trough to trough, there's still a lot of growth, right? You know, even if, you know, you're in an upcycle. And so I think what's relevant now about memory and other components is the shifting faces of what's happening, right? Historically, we had, you know, upcycles would be, you know, up 50%, you know, for the end market, and therefore, you know, for commodity markets like memory, where pricing power, you know, pricing is more elastic and things like that, you'd end up with, oh, those stocks 2-3x. But what we've gotten today is instead of up 50%, we've gotten over just the last few years, we've gotten spend has, you know, already doubled and it's going to double again, right? And so what we've got is total spend has doubled. And then when you look at the elasticity of different end markets, memory, we've had the pricing, you know, go up like 4x and it's going to go up another 2x, 3x again. in addition to capacity growth. And so you've got the stocks just ripping like crazy before going back down. And so I think what's really, really exciting about memory is it's not just an end-to-market thing. It's not just, oh, the market's ripping, it's a very elastic good, and memory is a commodity, and therefore the pricing of it is very elastic with the end-market demand. What's actually interesting, and this is something we wrote in 2021 when O1 came out, OpenAI released O1, which was the first reasoning model. And it created a new boom of reasoning models that OpenAI, Anthropic, DeepSeek, and many others have been exploiting to get the models to go long horizon agent tasks. So when we look at that, what's interesting is when O1 came out, the immediate thing we noticed is the workload changed dramatically. See, when we were doing chat, when you're talking to ChatGPT, you may send a prompt that is, it might be 50 words, it may be 500 words, but you're going to send a prompt and it's going to give you a response back. And that ratio, the context length, is a few thousand. So you might have a context length of, let's call it, 2,000. And what that means is, when you're running inference, every time you generate a token, you read all the weights into the chip, you read all the context into the chip, you process a token, and then you iterate again. You read all the tokens, the context, and weights. And the context is called the KV cache, right? It creates this relationship between all these tokens. What's interesting is when you're running a model inference on the weights side, if the context length is 1,000 or if the context length is 100,000, you still have to read all the weights. So memory intensity on inference is the same on the side of the weights, but on the side of the KV cache, the context, your memory, when you have 1,000 tokens that you're reading in versus 100,000 tokens, is a humongous difference, even though the compute amount is roughly the same. The compute amount is, because of KVCache caching on memory and things like that, you can sort of get away with your compute costs don't soar, but your memory costs soar. And so what we highlighted in our O note was, in December of 2024, we talked about the scaling laws and how pre-training scaling laws were giving way to reasoning scaling laws, and O was a big step function change. We talked about how KV cache was going to explode because of reasoning, and therefore memory was going to be the biggest winner. And so we did that in December '24, and multiple times in '25, we were really excited about memory. But really in, in, in, I think January of '26 is when we wrote the note, which was talking about, you know, at the time people were like, okay, memory has gone up 50%. You know, is it the top of the cycle? Do we need to keep going? And we wrote a note that was basically like, no, no, no. Uh, I don't think you guys get it right. Memory capacity is only growing 20, 30% a year for the next 3 years. And yet demand is doubling, is doubling. And so what's going to end up happening is memory prices are going to keep soaring. Users of memory who are less elastic or less capable of adapting to the elasticity of pricing will drop out of the market. Smartphones, laptops, because the costs are going to soar so much, they're going to drop out of the market. And that's going to all give way to AI. And what that means is the price is just going to have to soar and soar and soar until that happens. Because capacity is not going up enough. And so ultimately our point there was memory's in a shortage, and this is not a short-term shortage, it's a shortage that's going to last years. And, and, and so what we've seen so far over the last, over the, you know, rest of Q1 and, and now Q2 is, is memory's just been gangbusters. It's been soaring. Uh, there have been, you know, days where it's gone down 7, 8% for some random reason, uh, but ultimately the chart has been up and to the right. And And where we— not investment advice, but where we see it continuing to go is it's going to continue to soar because pricing continues to go up. Because we still haven't had— we've had some of the Chinese smartphone makers in the mid-range and low end, like Xiaomi, say their shipments are down 40%, but we haven't seen the high-end market get impacted yet. And next year, iPhone prices have to go up. Next year, MacBook prices have to go up. And right now, If MacBook prices or iPhone prices go up $100, that market's not going to adjust too much, but that memory is going to keep getting more and more expensive until AI gets its fill. Which means smartphone prices aren't just going to go up $100, they're going to have to go up a few hundred bucks. And at some point, there's going to be an equilibrium where AI gets the demand it needs and mobile and consumer hardware gets pushed down enough. But obviously at some point, people still need new phones and new laptops, so they'll still buy. So we're going to have to reach a new equilibrium because this end market capacity of memory doesn't grow fast enough. And as we extend across the ecosystem, what really matters is, hey, a lot of different components are in shortage. Who is taking elastic, who has an elastic price and who doesn't, right? An example is TSMC is not elastic on pricing, right? They're pretty, they're a pretty good company, pretty fair with their customers, partner long-term. They're like, we'll take up price 5, 10%. Memory companies, they're in a commodity market. They're, they let the spot market and contract market, you know, sort of supply demand balance really adjust pricing. And so you see these 2, 3x's in pricing. And someday you'll see pricing half, right? You know, because memory necessarily doesn't deserve a margin of 85%, right? Which is where it's headed to though, right? We're still not at 85%, 90% gross margins for memory, but we'll get there. And then at some point from there, it'll also half down back to like 70s or maybe even lower. And so we'll see this oscillation in memory. In TSMC, you don't see so much oscillation. In other areas like ASML, we don't see much oscillation on the pricing, right? They may make equipment, but different parts of the ecosystem will oscillate differently based on one, how much of the end AI demand flows through to them, right? Different parts of the supply chain are going to have, you know, for every dollar spent on AI, it might be 1 cent on this product, but it might be 5 cents on this product. And so obviously there's, uh, in this end market, right, memory versus something else. Um, and so ultimately different end markets in the infrastructure supply chain will benefit differently, but then in, in terms of demand, but in addition, what are the market dynamics there? Is it one where there's a monopoly or an oligopoly? Is there one where there's a very competitive large market? Is there one where the pricing is pretty stable and it's a lot of long-term agreements? Or is there one where it's, it's quite a commodity market and pricing is based on supply demand? And so all of these factors determine whether a specific end market, whether it be memory, or now people are talking about shortages of MLCCs, or, you know, PCB drill bits, or PCB foil, copper foil, or all these random components, you'll go online and see, This is the next shortage, this is the next shortage. Um, you'll see all these different things, but what matters is how much flow-through of demand is there actually? Is this end market doubling? Is it going up 50%? Or is it quadrupling, right? Um, is— how much is the pricing going to go up based on the market structure? These are what really determines what happens in the infrastructure supply chain.

And if you take that framework, because it feels like year by year the market wakes up to exactly what you said, a new quote-unquote potential Earlier this year, we have the OpenCLAW virality on various sites, awakens people to the world of AI agents and all the possibilities. Take the framework that you just described, and I'd be curious to hear your take on the CPU market, which for the first 3 years of AI, I don't think I heard the word CPU. And this year I'm hearing CPU everywhere.

Yeah. Yeah, yeah. So on the side of CPUs, what's interesting is in some of our institutional research for our clients, in November last year, we started talking a lot about it. And that's because OpenAI and Anthropic had started striking deals with Amazon, Google, Microsoft, etc., on buying all the CPUs they had in their fleets, renting them out. And then over the course of late last year and now this year, CPU demand has just been inflecting. And the reason why, right, let's talk about the reason first. Sure. So AI initially, when it was training and inference, and inference was mostly a short context thing, it was mostly just predicated on compute and networking, right? But as training, as pre-training shifted to reinforcement learning, right? And as chat-style inference turned into agentic, we had this big inflection of now CPUs are demanded more. Now, why is that? In the case of pre-training to reinforcement learning, pre-training is all about training the entire web dataset on Whereas reinforcement learning is the model generates some synthetic data or generates a reasoning trace, and then it checks it against an environment. That environment may be running unit tests on code. That environment may be a sandbox that looks like a website. That environment may be a sandbox that looks like an engineering system or some other platform that you would use, whether it be like a website or a shopping site or what have you. It would be something that you utilize on the internet. It might be compiling the code, and those environments require a lot of CPUs. Whereas before pre-training, the actual processing of tokens didn't require much CPU. It was all the environment checking, right? Okay, I've generated these tokens. Now, are they valid? What do they look like inside, whether it be a Python or a C compiler or within a website if I'm trying to buy something, e-commerce, whatever it is as an agentic workflow? I'm testing these things constantly. That requires a lot of CPU. And then the flip side is when you're doing live inference now, when you're doing chat, it's like, okay, I tell it something, it gives me an answer back, done. Okay, I might ask it a few more questions, but that's it. But now when I talk about agentic workflows where the model is making tool calls, right? So the model's like, okay, I'm going to go search for this. It's like, oh, I'm going to look this up in a database. Oh, I'm going to go ask the Python interpreter and I'm going to write a little bit of code to check my work. Oh, I'm going to write some code and I'm going to compile it. and deploy it, these agentic flows end up requiring more and more and more CPU because they have to actually interact with the regular world. It was one thing if the human is interacting with the model, where I'm telling the model something, the model gives me a response, I read it, I'm like, okay, copy and paste into whatever it is. It's a different thing when it's the model reacting with the internet world. And there ends up being a lot more compute in the loop, a lot more AI in the loop, or a lot more CPUs in the loop, sorry, that are bouncing back and forth the answers. And so both of these, whether it be reinforcement learning or agentic workflows need a lot of CPU. Now, what has ended up happening now? Because, okay, we need a lot of CPU, but let's evaluate the prior things in our framework, right? What is the market structure? Well, okay, so there's a few people in the market. There's Intel and AMD. There's Arm now releasing a CPU. So Arm stock has gone gangbusters because of that, because they're a new entrant into the market that looks pretty competitive. And then you've got Amazon, who's the leader of this, but also Microsoft and Google releasing their own CPUs that they've developed internally. And then you've got Nvidia who's releasing their own CPU. And so you've got a lot of different competitors in the market, but really up until 2 years ago, all of the market was Intel and AMD. And now Amazon's gotten a good amount of share. Now Nvidia and Arm are starting to get more share. But ultimately what happens in the end market is, well, Intel's actually able to increase their price AMD is actually able to increase their price. And so they've both increased their pricing. They've obviously gotten demand to go up a lot. Amazon is able to extract incredible margins out of CPUs because they don't make them and sell them, they make them and rent them. And so their Graviton CPUs are renting like crazy and they've increased the orders massively. Nvidia, who was previously only selling CPUs attached to their GPUs, is now selling CPUs standalone with Vera. And so they've given the guidance of $20 billion of CPU revenue. Now for Nvidia, that doesn't really scratch the surface of, you know, it's like, okay, that's like, you know, a few percentage growth. No, I'm just kidding. Um, you know, I think, I think, you know, when you look at other companies like Intel, AMD, Arm, um, Amazon, you know, who gets the, the IaaS revenue instead of the just, uh, you know, sales revenue, there's, there's huge things happening there.

Dylan, maybe on the back of that, With CPUs now, some of the discussion, at least that I've heard, has been that the CPUs for agents are different than, say, the historical CPUs in some regards. So the cores are more optimized for agentic activity is what I remember hearing Jensen kind of saying or implying around the Vera CPU. And then there's also a lot of discussion around this GPU to CPU ratio, which obviously, you know, highlights maybe the direction of the demand and need for CPUs. Can you maybe give us a little bit more color on each of those topics? Because I think the concept, I think, makes a lot of sense at the high level to people, but then there is also some of the technical things that are, I think, are probably maybe kind of pushed under the rug, if there are any. I'm not sure if it's just marketing or if there's a reality to this.

When it comes to agentic workflows, use of CPUs varies a lot. You have some workflows in agentic, it's like, okay, if my model's running and then I send a response to send all the tokens to some CPU workflow, and then I'm waiting on the CPU to do something, and then I send it back to the model and the model works some more. The question is, A, did the compute that the model's running on stall while you're waiting for the CPU? In some cases it does, in some cases it doesn't. And in the cases where it does stall, right, the compute that's running the model just stalls waiting for the CPU's response, then the CPU needs to be architected very differently, right? You know, the basic concept is, do I want more cores or do I want faster cores? And, you know, there's sort of like a law within CPU architecture, which is basically if you make the CPU core twice as big, right, which means I have half as many CPU cores on the chip, my performance doesn't go up 2x per CPU core. My per CPU core performance may only go up 50%. Obviously, there's a lot of engineering and the trade-off is not that simple, but to overly simplify it, that's a simple way to think about it. So if I look at an NVIDIA CPU for Vera, right? It has less than 100 cores, but those cores are faster than the AMD cores. AMD's leading CPU has 256 cores. So you've got this big delta number of CPU cores. But the NVIDIA core is faster, but it's not twice as fast, right? It's not twice as fast as AMD CPU core. And so there's this sort of trade-off that people are making on the design space. And for some workloads where the AI compute has to stall while waiting for the CPU, then you need, you know what, who cares if I spend, you know, if I have half as many cores and they're only 50% faster. And so in total, the performance of the CPU is is less, but the per-core performance is higher. And therefore I'm never— I'm not waiting on the CPU cores as often, right? I don't need a super parallel workload. What I really need is this one workload done now, right? And so, you know, in that case where the AI compute is stalling, I want to have the fastest core possible and I'm willing to sacrifice multi-core performance. And so that's some types of agentic workflows. Other types of agentic workflows, right? If I talk about how I use Claude, Day to day, or how the team uses Claude, how we spend $11 million a year on Claude on an ARS basis. What is that? Well, I'm calling Claude, Claude is processing a bunch of tokens, but they're not just using only me. They're batching hundreds of thousands of users together, right, across all of their compute. And so if I get the response back and now it's waiting on me to implement it, whether it's waiting on me or a CPU core to implement it somewhere, you end up with this That's okay because the compute is still running, just not for me. It's running for other people, right? So if the CPU is slower, but I get way more of them, it's a different sort of task. And another one is sort of like, is it the active use of AI or is it the use of— is it what's AI generated and then you're taking what AI generated and deploying it, right? The beauty is if we look at GitHub commits globally, they're up multiple fold versus last year. year, right? Not just— it's not just like a, oh, total GitHub commits are 10% up or 50% up. No, no, they're up like multiple X. And what that means is all this code is being generated into the world and people are deploying a lot of the code. A lot of the code is slop, but a lot of code's being deployed. And when it gets deployed, it's being put on CPUs. And it's standard code, right? It might just be like a web scraper, it might be like some analytical engine, it may be some business process automation. That doesn't necessarily need to be on a super fast CPU core. It can be on a cost-effective CPU And so when you look at the continuum, right, NVIDIA has built the highest performance CPU core, but it's not necessarily giving you, you know, if I have a chip, what is the maximum number of CPU cores times the performance per core, right? They're actually not so great at that. Whereas if I look at AMD and Amazon, right, they have a lot more cores, right? Hundreds, but they have less per-core performance. And so you end up with this like trade-off of, you know, and ARM is on that end too, right? And so where in that continuum do you want to go? And for some workloads, you do want Veron, for some workloads you want the Graviton or the AMD CPU. And I wouldn't say it's as simple. Now, as far as the other question you mentioned, which is ratio, it is indisputable that CPU demand is going up. We were the first to call it out late last year in our institutional research and in January this year in our newsletter. And since we published that, some of these CPU stocks have ripped. ARM has gone up multiple X, right? Intel, Nvidia— Intel's gone up multiple X, AMD's ripped, These stocks have ripped, but now the sell side, who doesn't really understand technology at all, is just making up stuff. And it's getting to the point where the ratio of CPUs to GPUs, or ratio of CPUs to AI compute, is getting lopsided to the point where it's more in the favor of CPUs than it is AI compute. And that's false, right? Just to reiterate, if you look at a Blackwell, full all-out Blackwell, it's like $50-something thousand per chip. If you had a 1-to-1 ratio, CPUs cost like $5,000, right? And so you end up with, okay, well then for $300 billion of Blackwell to sell or $500 billion of Blackwell to sell, you would only get, you know, you know, you'd get $30 billion or $50 billion of CPU sales, right? And so that's sort of like another thing that people are missing. Like, yes, this end market is ripping. Ultimately, the majority of the dollars are still going to AI compute and memory. This market was underpriced. And it's more fairly priced now, right? And I think that's something that people need to recognize is it's not like CPUs, CPUs are going to like keep growing and growing in demand beyond that of AI ASICs. It's, it's a bit of a, like a rightsizing, right? So basically in '23, '24, '25, there were years of selling millions of AI chips and very few CPUs. Now all of a sudden the CPU demand is inflected to now the ratio shouldn't be here, it should be here. And people are in catch-up And so now I need to buy a bunch of CPUs to catch up all the compute I've historically bought, as well as in addition to the compute I'm currently buying. Well, once I catch up that backlog of all these AI chips that I had bought previously and I catch them up for CPUs, well, that demand isn't there anymore, right? I've already caught it up and now it's only the incremental. And so if you think about it, right, if there's a ratio of, let's say, 1 to 2, 1 CPU to 2 GPUs, And each of those GPUs cost, let's say, $50,000, and each of those CPUs cost $5,000. Well, then for every $100,000 I'm spending on GPUs, I'm spending $5,000 on CPUs. And that's actually not that great of a market dynamic in terms of CPU growth. It's still great for CPUs. It's way better than it used to be. But if you flip back to, oh, well, what happens if I have 10 million GPUs and AI ASICs that I shipped over the last 3 years that don't have any CPU attached really? Okay, well, now that 5,000 has a huge catch-up, right? And so that's what we're experiencing right now is a huge catch-up as well as the ratio has shifted up and then that huge backlog is getting caught up. And so you're seeing like demand be ridiculous, but it will sort of subsume and then it'll be at a steady state, right? We're in sort of a mini cycle of CPU.

No, that's excellent context. Really, really helpful. And then maybe just moving to networking to kind of, you know, move into another area of the stack. I think this is one that's come to a lot of investors' attention, particularly as they dive down, I'll say, the optics supply chain and some of the constraints there. And we're seeing some estimates that co-packaged optics is maybe something that's talked about a lot, but is really probably going to be deployed 2028 or so, 2027, 2028. As you think about the, I'll say, there's obviously this You know, the concept of, you know, using copper when you can, optics when you must, in this kind of transition from optics to copper. And then we also had Jensen talking a lot about it at Computex as well. And, you know, a lot of other discussions in propping up, or at least bringing a lot of attention to firms like Marvell. Is there, I mean, is there any, I guess, additional thoughts you have around Optics and how you see the architecture of the data center within the networking domain evolving over the next 2 years?

Yeah, I mean, it's pretty clear as models get bigger, how do we run them across models? How do we train models? There's a lot of different domains within the optical stack, right? There's telecom optics, companies like Ciena have been ripping and many of the constituent supply chains around them. You look at datacom, chip-to-chip communications, that's got a copper domain, that's got optics domain today, and those all ripping because the growth of networking content is faster than the growth of any other content in terms of percentage, right? Networking's going from sub-10% to above 10% of spend associated with AI chips. And when we get to CPO, networking grows even further. It's like 20-30%, right? So we've got this huge uplift in networking content. But on the flip side is CPO is such a huge step function change in the industry and everyone sort of recognizes CPO now, But it's, you know, I think people are getting a little exuberant today. Currently, there's a little bit of a low. I would say currently people are a little bit too excited on CPO. It's not coming in '27, in my view, really in the tail end of '28. But '29 is the real ramp for scale-up co-package optics. There's been a lot of problems. It's a manufacturing thing, right? If we could deploy it today at good cost, amazing, everyone would do it. But it's really hard. The manufacturing volumes are not there. The yields aren't there. The chips aren't really designed there yet. It's like, it's a very complex, difficult thing to ramp. And so people are going to stay in copper as long as they can. And that means, you know, uh, Rubin is all copper. Uh, Feynman on the GPU is still copper, right? Which is next generation NVIDIA GPU after Rubin, Rubin Ultra, then Feynman. Uh, and we're not even at Rubin shipping yet. Um, so sort of Rubin just starting to ship. So sort of we've got a few generations of chips before we get to co-package optics on the GPU. There's co-package optics on switches, which is coming earlier than on the GPU. or the AI ASIC. But ultimately, that's sort of, you know, I would say even without that though, right, as the cluster size gets bigger, you need more optics per GPU, or more active electrical cables and things like that. So what we've seen is this big dynamic and shift. And so like, actually on Monday, we released a note at SemiAnalysis for our institutional research subscribers, which was on a localized time, you know, not saying end market, right? Obviously the technology, our agreement at CPO is going to happen. We've been pushing that for a long time. Our agreement You know, copper will subsume over time, but on a medium-term basis, we're actually very bullish copper, and we're very bullish optics that aren't CPO, and we're actually kind of bearish on CPO because of certain delays on chips that we see downstream. Right? Fine men not being full of CPO, you know, and other things like that that we've seen where copper names like Amphenol, right, who make all the backplane connectors and cables, are actually going to do way better over the next few years than previously expected. expected because we previously thought CPO would ramp sooner, but now it's delayed out. And so things like this happen in the supply chain, but ultimately optics is one which, if you close your eyes today and open it 5 years from now, it's going to be way bigger. A lot of that is priced into stocks, a lot of that isn't, and I'd say there's some local disjunctions, but this is sort of like part of the research that we kind of do and work that we've been doing with you folks as well is how do we weight that, right? How do we You know, what amount is is you know CPO favored optics versus non CPO optics, right? Typical optical transceivers versus copper, right? Because copper has actually got a long ways to go. There's a lot of things happening in the copper industry that are innovating and pushing back CPO, right? It's like why would I do CPO? Because at the end of the day, integrating optics is so much more expensive than sending something electrically. Except if I have to send something electrically, I can't go that far unless I add repeaters or optics. And so there's like this this trade off and continuum and. And CPO will happen, but it looks like it's getting pushed out a little bit.

And Dylan, as we sort of go into our probably— we probably have time for our last overall topic. We've done models, GPUs, CPUs, memory, networking. We'd probably be remiss to not mention the elephant in the room at any data center, which is how are you getting the electricity and how are you sort of getting that electricity into the form. And I know you've written through the newsletter at least certain things about direct current versus alternating current and certain elements that in theory, when you see the hyperscalers spending all this money building these data centers, potentially even putting the power plants on site behind the meter, how are we to think about the electrical demand, sort of the grid versus non-grid? And I know it's a big topic, but I feel like we'd be remiss to not at least mention it here.

Yeah, so I would say data center growth is massive. This year we're deploying 20 gigawatts of data centers. Next year that number goes up 50 gigawatts, or 50%, sorry, 50%, sorry. Oh, sorry, 30. 30 gigawatts. And then it'll be 50 gigawatts the year after that. The growth in data center capacity is massive. There's a lot of local disjunctions that people are having to deal with, right? And so sort of, I would say energy is one of the biggest ones. The other one is sort of political, and third is like sort of construction. Building data centers and getting permits and filings is kind of tough politically. People are trying to stop it, but the primary factor gating it is really the energy at the end of the day, right? And so what's happening there is energy you can break down into a few things, right? There's generation, right? Where do I generate the electrons from? There's transmission. How do I transmit the electrons from where it was generated to the data center? And then there's conversion, right? Because the power that it gets transmitted at is at a form factor that the chips cannot consume. And so the chips need to consume it at a different form factor. What does that conversion pipeline look like? And I think in all 3 of these, there are very bullish aspects, right? In the transmission side, that's the one that's the hardest to be bullish on because the regulatory and political difficulties with building more transmission capacity and the way local monopolies of utilities work and how if they build a utility line, they have to amortize it across all users, not just individual users. There's all these various weird locations with transmitting power. And so building more grid capacity is kind of difficult on a transmission basis. But generation-wise and convergence-wise, there are 2 interesting things. So generation-wise, obviously there's more generation happening on the grid. There's also this big shift to generating for the data center. So we would predict in, in a couple years, half of the power for data centers, for incremental new power for data centers, will be generated onsite, not offsite. Um, and so behind the meter is sort And we see this with the behind-the-meter tracker we have in our data center and energy models. I mentioned someone on our team, she built a power plant in Kazakhstan. She's leading our energy model, Ellie. And so she's sort of been tracking and we've been building this model of the entire grid, every generation asset, every transmission asset, all the load assets, as well as all the behind-the-meter work. And what's interesting is we've seen this huge boom in behind-the-meter. There's been a a lot to fight the permitting and regulatory side there, you know, whether it be people not wanting to allow air permits or people wanting to have, you know, not allowing the gas pipeline to be built to the side or things like that. We've seen that with an Oracle data center. There's a lot of different aspects of this that are happening, but ultimately the end state of it is that behind the meter is soaring, right? And a lot of that is gas, you know, a lot of it was, you know, dual combined cycle gas reactors from GE Vernova or Mitsubishi. Or Siemens, but beyond that, there's also been a lot of different types of energy sources, right? There's reciprocating engines, there's industrial gas turbines, there's various types of diesel engines, there's train engines. People have taken train engines, people have taken boat engines, people have taken truck engines and converted them into power generation for data centers. So we see a sea of innovation happening there. It's not like we don't have the industrial capacity, right? You know, the U.S. has makes millions, can make millions of industrial I'm sorry, reciprocating engines a year, right? There's just engines that burn fuel and spin. And it's pretty trivial to retool those to be for gas rather than diesel. But even if it's diesel, that's fine. And then you stick an electrical engine on it and basically backdrive it, and then that generates electricity. And so then you can do this in like large, large volumes to generate power. And so we see, you know, 10 gigawatts plus of data centers that are going to be built with technologies like this, right? Taking diesel truck engines, converting them to gas, which can be done at the time of production, very simple, back-driving an electrical motor, and then sticking them on site on a data center. And you have hundreds of these backing a data center. And then you hire a bunch of people from car mechanic shops. And if, you know, these things need to be serviced, right? And so they just run around servicing these diesel engines all day. You know, and so you have some buffer so that you don't like, you know, when they go down, you can service them and keep them going and have the max power. And obviously you need some batteries in between because you don't want the up and down of the data center to mess with, to blow up the engines that you've got. And so you've got like this entire supply chain of behind the meter, which is exciting. But in addition, you've got, you know, over, you know, in about 2 years, solar plus battery will be cheaper than gas. Now, supply chains for solar plus battery are difficult, and it depends on what level of reliability you want. If you have just enough battery to get through the night, it's cheaper. But what if you need enough batteries to get through night, 3 days, right? Because it might rain for 2 days. You know, things like this, you know, how many nines of reliability do you want? But solar plus battery are getting cheaper and cheaper at an incredible pace because of China's manufacturing excellence. Um, and, and some of the subsidies too, um, that, you know, it's going to get cheaper at one point to, uh, at some point to do solar plus battery. Um, and then you've got space data centers, right? Where you don't even need a battery. You just stick it in space and you've got a solar panel and that's it.

Right.

Um, and so you've got this whole continuum of ways to generate power. whether it be all the way, you know, taking diesel engines and making them gas engines, or, you know, using dual combined cycle engines, all the way downstream to, you know, let's just ship the chips into space instead, right? So there's the whole continuum, and there's a lot of money to be made there. There's a lot of interesting dynamic things to do there. That's why, you know, actually the largest dataset and research vertical for SemiAnalysis, which you'd think is semiconductors, is actually data centers and It's data centers, industrials, and energy. We call it the DEI team, the data center energy industrial. It's a pun internally. You know, the tag is @DEIteam. So Jeremy, he leads that team. He came up with the name. But anyway, so data centers, energy, and industrials, that's actually our biggest research vertical because we're tracking every data center and every power plant. And so the market, you know, when we identify like a delay or this is happening or there's going to be this build or this company is going to have this many data centers go online in this quarter, it's something that no one else in the industry can do. And that's of why it's one of our biggest verticals. And everyone's interested in that, right? Google's interested in what Meta's able to deploy. Meta's interested in what OpenAI is able to deploy. But also all of these guys are looking at what the supply chain is able and who has capacity. And then also all the investors are looking. So that's our biggest dataset. But I would say it's one where it's a market where it's very decentralized. In the case of memory, there's 3 names. It's pretty simple. In the case of accelerators, there's just a few names. In the case of Semiconductor wafer fabrication equipment, there's just a few names. In the case of this, there are hundreds of names in the supply chain, making all these random little widgets. There's dozens of companies building data centers, and there's dozens of companies trying to do, whether you're being an IPP as power producer, or you're doing it behind the meter, or you're offering some sort of battery service, or all these different things. It's a very complex supply chain, but one that has a lot of dynamism And ultimately, I think there's a lot of innovation happening. And so while data centers will continue to be a constraint of sorts, they will also not be a constraint because, you know, it depends on how crazy you're willing to go. But like I said, you can just take truck engines and convert them and hire a bunch of mechanics and run a site like that. It's not going to be the best. You know, a lot of people say, oh, that's disgusting. How reliable is that going to be? You know, oh, that's going to be really annoying to do. do, but people are doing it and it will work. And it's a pain in the ass, but it will work. All the way to like, I'm going to shoot it into space. It's going to be a pain in the ass. It's going to be really hard to make it work, but it will work. And so you've got solutions to the data center problem, whether it be going full dirty or going fully into space, whereas other parts of the supply chain you don't. And so I think that's what makes this market so dynamic is you're going to see people go up and down a Yeah. And then I guess the other part, so that's on the generation and transmission side, but on the conversion side is the other thing, right? How do you get the power from where it is, what it gets generated at or it gets transmitted at, to what the chips want? And there's an entire supply chain of stuff going on there, whether it be IGBT, silicon carbide, various types of MOSFETs, GaN, gallium nitride MOSFETs, all the names there. What happens when we go from 12-volt to 54-volt to 800-volt DC? And the conversion supply chain there. What happens with solid-state transformers as those get innovated? All these things are happening in the space. What happens with UPSs, you know, universal power supply? Is it universal power supply? Battery backups and supercapacitors and all these other different ways to smooth out the power, make it from, you know, the dirty variable power that gets created on the left side to the super clean power, but also variable usage of it on the right side. How do you match that? And that entire conversion pipeline is one that's super, super exciting. And so we're We had a blog on that in 800-volt just last week, and we've talked about more recently to our institutional subscribers some delays that are happening there on the side of Nvidia as they delay it out of Kyber. Rubin Ultra Kyber doesn't have 800 volts anymore, and so what does that mean for the supply chain? Well, it gets pushed out a little bit.

So Dylan, I want to thank you profusely on our side. This is this is the first. This will be the first episode, if we think in terms of chapters. This will be the first time we've had Dylan on the podcast, but certainly not the last, because there's a lot more information. And like he said directly numerous times across the entire stack, everything's changing all the time. It's a bear to keep track.

But as he stated, he has 90 people. The other thing I would say is this supply chain is so freaking crazy. And a lot of times we talk about the big ones, right? Memory, CPUs, data centers. But actually when you drill down to the supply chain, the local bumps are very small and many, right? For a couple of months we were talking about PCB drill bits, right? The drill bits that drill into PCBs for the holes for, it's like random copper foil that goes on PCBs. There's all these random small things in the supply chain that also have these dislocations and also the companies that exist in them them are all over the world. They could be trading in Taiwan, they could be trading in Japan, they could be trading in Korea, they could be trading in, you know, um, in all parts of the world. It's not just easily accessible to investors. And so I think that's what's really exciting about, you know, our partnership and the way we're working together is, you know, we, we get a— we get to influence sort of, you know, what's going on. We get to talk a lot about these, these supply chain disruptions, but also what's, what's really interesting in the framework that I laid out earlier and, and the entire landscape that we're trying And so, you know, looking forward to coming back on the show more and our other collaborations.

Absolutely. And with that, I do have to state one thing here for our compliance team to clarify. The views and opinions expressed in this podcast are those of WisdomTree and are subject to change. Anything we present in this podcast is not intended to be relied upon as forecast research, nor as investment or tax advice. The information and opinions expressed in this podcast are not a recommendation, offer, solicitation to buy or sell any securities, and reliance upon them is at the sole discretion of the listener. Please remember, past performance is no indication of future results. Thank you everyone for taking some time with us today, and we look forward to coming back in the future.

Take care.