La oferta y la demanda de tokens de IA

[00:00] What used to matter a lot was execution was very very [ __ ] difficult and ideas were cheap. Now ideas are cheap and plentiful but execution is very easy. So really only the good ideas are the ones that can justify the spend on super cheap implementation. >> [music]
[00:30] >> You told me this incredible story about how your own team's use of tokens has [music] changed dramatically this year. Yeah. Retell that story and what it is teaching you about what's going on in the world. >> Last year we thought we were heavy users of AI. Everyone's using chat GPT. Everyone's using cloud. Everyone's got you know I'm providing whatever subscriptions anyone wants on the order of spend of like tens of thousands of dollars for our firm. This year the spend has just skyrocketed and and it really started in late December with Opus that included Doug who's president
[01:01] uh Doug Olaflin. He's very much like leading the charge in the sense of like non-technical people using uh AI for coding. Um and so he's basically pled the whole firm slowly over time. I think he's been the the leader in doing that. Obviously the engineers were using it anyways but spend in January just started to inflect and rocket and rocket and rocket and rocket. Um, we signed, you know, an enterprise contract with Anthropic and it's gone to the point where now, um, I think when I last talked to you it was 5 million spend rate. It's actually 7 million spend
[01:31] right now. >> That was last week, by the way. [laughter] >> A lot of that is just the usage, right? What's what's really, you know, people people who are have never coded before are using cloud code and spending thousands of dollars sometimes a day, but across a firm, we're spending $7 million a year now on cloud code at the current rate. um versus our salary expense being in the neighborhood of $25 million. So, you know, we're north of 25% of spend on cloud code as a percentage of salary. And if this trajectory continues, then, you know,
[02:02] we'll spend more than 100% by the end of the year. Uh which is a bit terrifying. Thankfully, I don't have to decide between people and AI because our company's growing so fast. It's, you know, more so like, okay, well, I don't have to hire nearly as fast and I can spend a lot more on AI and it works and we just grow faster. But I think other folks will start to reckon with the fact that huh if this person can do the work of five to 10 to 15 people uh using cloud code then all of a sudden I should probably cut people. But right now I think the use cases are so broad for
[02:33] example one thing is we have a reverse engineering lab in Oregon that we've been building for a year and a half. We have a bunch of you know fancy microscopes scanning electron microscopes. The whole purpose of this is you reverse engineer chips. You get uh the architecture out of it. you get the materials that they're using to manufacture and this is some of the data we sell. This is a very slow process of analyzing that data. Instead, um, one person on the team, they've been able to spend with a couple thousand dollars of cloud tokens. They've been able to create this application that is GPU accelerated, runs on a server that we have at Coreweave and anytime we send it
[03:04] an image, it's able to take the picture of the chip and overlay where every single material is. Oh, this part is copper. Oh, this part of the gate is uh tantelum. This part of the gate is germanium. This part of the gate is cobalt. And so you can do a finite element analysis of the entire stackup of the chip very very quickly visual with a dashboard guey. It's everything few thousand dollars of to claude. The person previously worked at Intel and he said that was an entire team's job to build that and maintain that. Now rack that up across you know the entire firm. It's it's insane. Another example that I think is super fun is Malcolm. He's an
[03:36] economist at a major bank before um their economist department was like 100 or 200 people. What he built was the most incredible thing ever. He piped all of this different data, you know, FRED data and all these other data, right? Employment reports and all these other things from various APIs. We signed a couple contracts with folks to get API access to data. Pulled it all in, started running regression, started looking at the impact of various economic revolutions on the economy um from a deflationary inflationary perspective. The BLS has this entire um
[04:08] Bureau of Labor Statistics has this entire like set of like 2,000 tasks. And so he did that with AI, which ones can be done by AI, which ones cannot, and grading them across a rubric. You know, about 3% are doable now with AI. Um, and so he's created this like metric so that you can measure things that can be done by AI, what what the massive deflationary uh, you know, what the cost of being able to do those with AI and therefore the deflationary aspect of it. You know, output can go up. It's called phantom GDP is what he called it. Phantom GDP. Output can go up, but because cost falls so much, actually GDP theoretically shrinks. So he created
[04:39] this whole analysis and a brand new benchmark of uh language models um a set of evals across 2,000 different evals. Right. >> He did this all by himself. >> This is all by himself. Yeah. And he's like, "Dude, this would have taken the team of 200 economists a year." And he's just like he's like completely cracked out on claude. He's like everything has changed. >> How do you think about as a business owner going from close to zero to 25% accelerating towards whatever percent of total spend? Like at what point are you like, whoa, I need to put the brakes on this and be careful how much we're spending. Maybe we don't need to spend
[05:10] on the most cutting it on Opus 4.7 which came out today. Maybe I can throttle it back to something that's a little bit cheaper. >> Ultimately, like I'm in the information business, right? That is, you know, we sell analysis, we sell, we do consulting, we create data sets. I don't see why this wouldn't be completely commoditized on a pretty rapid basis if I'm not constantly improving. my first product that I was selling as a data set actually it is you know like there's more people trying to do it now we've made it constantly better and better and better and more detailed and so therefore it has a market but the way we were doing it in 2023 is not terribly
[05:42] different than you know is it's it's basically what everyone else is doing now if I don't move up the bar then I will be commoditized if I don't move fast enough I will also lose my edge so the question is yes AI commoditizes things just like it commoditizes software those who can move fast and keep control of their customers and keep providing them an awesome service and keep improving the service won't shrink. They'll grow. They'll grow faster. Those who are incumbent and not doing anything, they're going to lose. And so, it's a bit of an existential like if I
[06:13] don't adopt AI, someone else will and they will beat me. Uh, another easy example is the energy space. So, we've had a few energy analysts for a couple for like a year now. We've been trying to build out this energy model. It's very complex. Energy's data services market is something like $900 million. So obviously a huge market for me to try and break into but it has you know we really hadn't broken into the energy data services business despite a year of having multiple people on the team. Um then cloud code psychosis hits one of the people who leads the data center and energy and industrial sort of business at semi analysis uh Jeremy hits him and
[06:44] now all of a sudden in 3 weeks um he spent a lot he was spending like $6,000 a day. It was an insane amount but he scraped every single power plant in the US every single transmission line above a certain voltage. um and created this entire mapping of the entire US grid as well as a lot of demand sources all from various public sources of data. Um and we've shown it to and and we built and it's got like this dashboard where you can view and check you can see all the micro regions of the US where there's power deficits and surpluses. Um all of
[07:14] these details built in a handful of weeks we started showing some of our customers who buy our data center data set but are energy like traders. We showed some of them and they're like wow how long did this take you? This is really good. this is better than XYZ company. We're and then we like get dig deeper. XYZ company has 100 people and have been working on this for a decade. Now obviously our thing is not fully robust as robust but in some ways it is better. I'm going to commoditize these energy services companies, data services company. Who's going to come commoditize me if I don't move faster? And so the question from a business owner's
[07:44] perspective is yeah I'm spending a lot but what is that spend getting me? Is it getting more revenue? Yeah. >> Most software companies try to maximize your time on their app to juice engagement. RAMP does the exact opposite. RAMP understands that no one [music] wants to spend hours filing expense reports, reviewing expense reports, and checking for policy violations. So, they built their tools to give that time [music] back, using AI to automate 85% of expense reviews with 99% accuracy. And since RAMP saves companies 5%, [music] it's no wonder that Shopify runs on RAM, Stripe runs on
[08:14] RAM, and my business does, too. To see what happens when you eliminate the busy work, check out [music] ramp.com/invest. OpenAI, Cursor, Enthropic, Perplexity, and Verscell all have something in common. They all use Work OS. And here's why. To achieve enterprise adoption at scale, you have to deliver on core capabilities like SSO, SKIM, [music] Arbback, and audit logs. That's where Work OS comes in. Instead of spending months building these missionritical capabilities yourself, you can just use Work OS APIs [music] to gain all of them on day zero. That's why so many of the
[08:46] top AI teams you hear about already run on Work OS. Work OS is the fastest way to become enterprise ready and stay focused on what matters most, your product. Visit works.com to get started. Felix by Rogo is a personal finance agent that turns a single prompt into finished client ready work using your firm's own templates, context, and standards. Send Felix an email like, "Take these comments and turn them for me, or update my tracker with the context of these [music] emails, or run the ability to pay math on this buyer." And Felix sends back finished PowerPoint
[09:16] decks, Excel models, [music] and sourced research. Felix works the way your team already does, delivering work quickly and accurately around the clock. Learn more at rogo.ai/felix. Are you worried that in the limit the people that control capital and investing capital, who are often hiring you for for what you do, will just say, "Well, we have analysts, too, who are really smart about this. Like, we'll just build this ourselves." Like if it's getting that easy, at what point does it just all pull into the investment firms that stand to gain the most because they have the most leverage on top of the
[09:47] data or the insights that that they glean? >> First of all, any information services business, obviously I don't generate as much value as my customer does from such information. Uh because if I sell you information for a dollar, you're only buying it for a dollar because you know that information helps you make a decision that lets you make more than $1. And so therefore, you have you have arbit you you have made more money off of me than I did from the information myself. An investment fund, these investment funds all have their own information services, you know, especially like the super like the Jane Streets of the world and the Citadels.
[10:17] They're they're really detailed on their data. And yet, um, these sort of folks also purchase data from us and continue to do so and continue to grow with us because I think there's just some some it factor, right? We move faster, we're more nimble, we're a smaller team that's focused on just one specific thing. uh AI infrastructure and and the huge revolution that causes in AI um and tokconomics and all these things and and we sort of really see where it's headed and so we're moving faster and building faster. Um I think investment
[10:48] professionals just would you know yes they'll try and build some of the stuff we do and um more likely they'll just buy the data from us and it's cheaper for them to buy the data from us and then to build and then build on top of it than it is to build it themselves. But ultimately some may try. I feel like every conversation I have with you, what I'm always getting at is just supply and demand of tokens like that's the thing that's interesting to me in the world right now. What has this experience taught you about the demand? Has it changed your view on the demand side of that equation? Just feeling it viscerally yourself. >> If we take a step back and look at the
[11:18] macro lens, right? Enthropic has gone from 9 billion revenue to what they're at 3540 billion now. Probably by the time this airs 40 45 billion, who does [laughter] ARR? Their compute has not grown to the same degree. Um, and if you do the calculations and you assume they didn't decrease their research and development compute, they clearly didn't. Their release, they have methos. They have up is 4.7. So they clearly didn't decrease their research compute spend. Um, so ultimately what they've done, even if you assume all incremental compute they've gotten has gone towards inference, their margins are at a floor
[11:49] of 72%. In reality, some of that incremental compute they've got probably went to research and development. It may be higher than 72% gross margins. To be clear, at the start of the year, they started uh there was um there was a leak by someone from their funding some some of their funding round docs. Someone leaked it 30 something% gross margins. Where on earth does a business like this grow margins like that? And it's in principle, right? Their demand is so high. They're able to cut back on usage limits, rate limits, all these things. Um, what really matters is having an anthropic rep and having an enterprise
[12:19] contract with them and getting the rate limit increases that you need because otherwise tokens are ultimately super super in demand. Whoever whoever can pay for them anthropic has the same problem, right? Like I mean not problem, it's it's just the reality of how capitalism works. Yes, people are spending sending them $40 billion ARR in tokens and but those tokens are generating way more than $40 billion in value. Various businesses will have different value generation per token. But as we get more and more intelligent, what really matters is access to these most
[12:49] intelligent tokens and leveraging them at things. You as a person deciding what is the best way to leverage these tokens to grow business and generate value because a lot of folks will want tokens and generate tokens. Uh but the shitty SAS startup and and and and SF who is using Claude to generate, you know, their software product is not necessarily actually creating a ton of value and therefore they're going to get priced out of tokens uh soon enough. >> Are you at all surprised that I I had this experience just today where on the flight here I got really limited out on
[13:21] something. I saw 4.7 came out and what I immediately wanted was like to be on 4.7 that second and I was it just I couldn't think about using 4.6 anymore. or not this 47 is out. I was perfectly happy with 4.6 for the last many weeks. It's amazing. Are you surprised that people are so insistent on going to the most expensive leading edge thing to the degree they are? >> Without a doubt. One of my funniest memories in the past month and a half is myself and a buddy of mine, Leopold, being on our knees in front of an
[13:53] anthropic co-founder begging him for access to Mythos and then pretending it doesn't exist [laughter] cuz we knew it existed. were like, "Please give us access." And he's like, "I don't know what you're talking about." >> What was your reaction to that rate card or that eval card coming out? >> It was rumored in the Bay Area. Everyone, you know, we sort of like knew it was supposed to be really good, but um if you just look at the benchmarks and obviously benchmarks change over time, Mythos is potentially the biggest step up in model capabilities in like 2 years. I think that's really really an
[14:23] an important detail that you know it it's so good that they're like don't want to release it even though they're they they already announced the price to their people that they did a selective release for cyber for and it's like five or 10x the token cost. They just don't want to release it um because they're worried about the like impact on the world and they're releasing a [ __ ] worse version Opus 47 to us and they explicitly said in the model card hey we actually preferentially made it worse at cyber. I don't know if you read that. Whoever you are, if you have enough capital, you should get a freaking
[14:54] enterprise cloud uh enterprise enthropic subscription where you pay per token, not with these like subscriptions because then you won't get rate limited much. And then you must you need to figure out how to leverage those tokens to the highest value task um and make money off of it because ultimately what you're doing maybe maybe like a year from now or two years from now the business is actually just arbitrageing tokens, right? the tokens are amazing, but let's figure out what direction to point them in. And then three or four years from now, the model will know, you know, what to do with the tokens and how to make the most value. You know, you can look at this retroactively. Pick any benchmark. The cost to hit a certain
[15:24] capability tier used to cost X and now it cost 1/100th or 1/ 1,000th of that. Deepseek, for example, on GPD4 was 1/600th the cost. And since then, the costs have fallen further for GPD4 class models. Of course, no one gives a crap about GPD4 class models. They want the Frontier because the Frontier lets them create the economically valuable things. But GP4 class models can still be used in like stuff and so people are using them in some like tiny use cases. It's just the cost have fallen so fast. It's
[15:54] it's not really what's driving the demand. What's driving the demand is is all these new use cases. Yeah. Current 4.6 6 opus or 4.7 opus tier models a year from now. My spend for the same exact quality of the model would probably be like 70k. I bet you it'll be 100 times cheaper. Irrelevant because I'm going to be using a way way way better model which can do way way better things. Enthropic mythos is more expensive as a model, but it spends a lot less tokens to do the thing and therefore it is actually cheaper in most
[16:25] tasks than 46 opus because it's just way more efficient even though each individual token is smarter. >> When I last saw you, Methos had just come out maybe the day before or something or the the card had just come out and you said something like uh it actually made you feel like a little scared it was so good. What did you mean by that? Anthropic's whole like goal in 2025 was and and even a lot of 2024 they're like hey by the end of 2025 we need an L4 software engineer uh in our model and and they by and large achieved
[16:55] that with 46 opus what they didn't say is that you know and if you look at mythos and if you compare like the benchmarks it's like an L6 engineer so L4 is like pretty new L6 is like quite well experienced I think anthropic said that the model internally was available in February So in 2 months they've gone from L4 engineer to L6 engineer. Uh what's next? Um you know when when you think about the model progress it's only accelerated. Enthropic release cadence has compressed. Open's release cadence
[17:25] has compressed. Why? Because these models generally to make a better model you need a few things right. You need amazing compute. Comput is very expensive and it has a time scale that we you know we track and it's like you know it's growing but like you know it's it's sort of set in stone for the next you know short short term. it's like kind of set in stone what you've already signed. Um there will be delays and shifts and some somehow you can find a little more but it's generally pretty set in stone. There's amazing researchers that people are paying tens of millions of dollars for. And then lastly there's implementation. Historically has been very difficult. If I have an idea now I have to implement it. Implementing is hard. Now ideas are
[17:57] there. Implementation is very easy. It's expensive but it's very easy. So how do you how does one decide what ideas to implement? And it turns out if your implementation is just so much easier now you can just implement more ideas and move on the treadmill faster and faster and faster. Whether that is AI model research and so now your model release cadence is shrunk to down to 2 months from where it was 6 months before or hey I want to I want to take every power plant in the US and every transmission line and model it and run regressions and see the micro supply and demand. I can also do that. The idea is
[18:28] cheap. You know which idea makes sense? which idea is worth the capital that you have to spend on the tokens because the implementation is there. It's it's that's the I think the key learning and if implementation costs continue to tank which they are um we don't even have mythos yet. It's only been you know a handful of hours since Opus 47 launched but you know my team is pretty excited about it internally. What now comes to the world uh it's a complete reordering of how like economies work. What used to matter a lot was execution was very very
[18:59] [ __ ] difficult and ideas were cheap. Now ideas are cheap and plentiful but execution is very easy. So really only the good ideas are worth are the ones that can justify the spend on super cheap implementation. >> So are you actually scared or are you just is it just does it just introduce an uncertainty that's hard to grapple with? >> Uncertainty is there. Um but I do I do think that causes some fear in terms of how does society reform itself? How does
[19:29] one exist in a world where actually any you know your ability to implement something is not actually that important. Your ability to choose the correct idea for AI to implement and then your ability to sell that idea or sell what the AI has implemented is what matters. Your ability to garner capital towards that is what matters. And going back to the point of like it's very important to have the newest model always. Who's going to have access to the newest model? Anthropics project. I know it's not called earwig, but I troll anthropic people by calling it earwig. Um
[19:59] glasswig. Anthropic earwig you know where they only release mythos to certain companies for cyber. That's just going to be something that continues. Models will have less broad and less broad deployment. I know I know open AI and anthropic and all these people are like we want to have great AI for everyone. AI is very [ __ ] expensive. Who's going to pay for the trillion dollars of infrastructure? People who have money and can can build useful things with AI. And then you don't want people to distill your models. So you don't release them broadly. Uh you release them to a fewer and fewer set of
[20:29] customers. Those customers are also now wrestling over the tokens unless anthropic jacks them. You know, they could double their pricing on Opus and I would continue to pay and I bet most users would continue to pay. I bet that wouldn't solve their humongous capacity problem that they have. So then the question becomes where does this cycle end where you know token usage and therefore the benefits of those tokens the additional value generated on top of those tokens aggregates among fewer and fewer and fewer companies. I don't have mythos. You know who has mythos? Top freaking banks. Um now they're only
[21:00] using it for cyber security. At some point I can envision a world where hey maybe I because I have an enterprise enthropic contract and because enthropic people kind of like me they're willing to give us like slightly earlier access or slightly higher rate limits or something for a model. I hope that's what happens. And then my competitor whoever that is doesn't have that and I'm able to [ __ ] crush them. There are people who are like Ken Griffin of Citadel is like super wellconnected and super rich and he's like he he just signs, you know, who knows? he goes and signs a deal with open aranthropic that's like, "Yeah, I'm going to get
[21:30] access to your models. Um, and I'll buy the first $10 billion worth of tokens each year. So, whenever you release the model, you know, I'll spend the first 10 billion tokens and then everyone else can get the model after that." And it's like, okay, well, now what does that do? Well, now he's going to crush everyone in the markets. That's just an example. Could be cyber like Anthropic is worried about, oh, now I can hack people. Could be information services business like myself where I crush someone else. I think, you know, it it's it's such a broad base. We don't know what these models can do. Anthropic doesn't know what these models can do. No one knows what these models can do. It's up to the end user to figure out where they can leverage the tokens to see what they can
[22:01] build and imagine which is tremendously productive and uplifting for humanity. But then what happens to the concentration of resources and usage of it? >> Presumably right now robotics or robots consume relatively zero tokens versus everything else. Do you see what's your view of that? If that's like a second demand curve that could start to ratchet there's a new startup every single day, you know, within a mile of here trying to build something interesting in robotics. So there's this concept of software only singularity which is that the world has you know AI singularity but only in software and now what about
[22:33] the rest of the world vast majority of the world is physical you can see the world orient around hardware not software that's actually why I think software only singularity is like just a blip and not like a you know we we do get everything else because once software is super easy what makes robots really hard it's like programming microcontrollers and actuators and controlling all this stuff is very difficult right now The interesting thing about models, AI models, is they're actually really inefficient in learning. It's just we're able to give them so much data that they're able to
[23:03] learn and surpass us in certain ways. Robots currently, the robot models, um, VAS, uh, vision language action models, which is very popular right now, is probably not going to be the thing that ultimately scales beyond. They are inefficient in data. Um, and we can't scale the data for them fast enough. there is going to be some way to large scale pre-train robot models where just like humans see all this data throughout their lives and what's interesting is humans the reason why we're so good is we're sample efficient one example two example we're good and so applying that
[23:33] to robotics so once you once you have this software only singularity implementation is super cheap anyone can start to build these mo people can start to build models that now robots are actually useful and so I think in the next 6 to 18 months we'll start seeing real breakthroughs in robotics that enable able few shot learning i.e. there's a pre-trained robot model and now there's a robot that you have hired or bought or whatever. You show it a few examples and it's able to do it. You tell it to stack these two things or you tell it, hey, this can can actually like
[24:03] balance perfectly, you know, and and it starts doing these things. [laughter] >> One shot. >> No, trust me, I've spilled many of times. >> So, I think I think robots will you get fot learning right now. You know, there's a lot of companies doing robots for like, you know, advertisement or robots for like simple stuff like that, but it'll be like, oh, folding clothes, but it's going to get really niche like robots just for cleaning chalkboards. Um, and it's a rental service or, you know, it'll be it'll be a model package that you download onto your standard robot that then does that, right? And
[24:33] and you pay for that use. And anyways, there will be a huge explosion in physical good acceleration and and deflationary effects there. But and and so that's that's ultimately going to keep token demand going crazy. I I don't think token demand slows down personally. >> Did you learn anything else about the world based on Mythos's results and how it was built? My way of asking like the you know if you break down the the components of the scaling laws like the >> so mythos is a materially larger model than prior models and so yes it is a much larger model. Now whether or not
[25:03] it's it's what chip it's trained on is not really relevant. It's the scale and obviously you know to a 100,000 black wells is equivalent to hundreds of thousands of prior generation chips. TPUs and tranium have their different release cadence. So it's not exactly like mirrored one to one. Um but ultimately yes mythos is a significantly larger model. It's proof that the scaling laws still work. Um everything about it shows that the trend line continues of models more compute into model makes model better. And along the whole way it's not just more compute into model makes model better. Along the whole way we're also getting these compute efficiency wins which are you
[25:34] know as as all this research compute that the labs are spending is actually turning into if I want X capability tier model every 6 months that cost or every two months that cost is dramatically decreasing but then if I scale it up massively I get a humongous capability jump as well and so yes it's it's proof that this is still happening. Google and Enthropic are not heavy heavy users of GPUs on the training side, but OpenAI, they'll they'll start having their new class of models. I think they're taking a more sensible principled approach to scaling uh in small steps. Enthropic
[26:05] really went for a huge jump. We'll see better and better models throughout the year and the release cadence is only going to get faster. >> We've gone a long way in the conversation with saying almost nothing about OpenAI, which would have been so strange. >> So, so this is this is the interesting thing. Everyone's like, "So, okay, so Anthropics just won, right?" You know, they had Methos in February. They never even released it because they didn't feel the need to. They're already sold out. Their revenue is already adding $10 billion a month. Um and then you've got Opus 47 today all before open eyes, you know, um alleged Spud release, which you know, media such as the information and
[26:36] others have have posted about. So clearly Anthropic is in the lead, right? And OpenAI is cooked. What's interesting is because Anthropic has such bounds on compute and they can only grow it so fast and sort of to the point of you know you know Daria Daria used to gloat about how OpenAI was being too aggressive on compute and Anthropic was more sensible in their scaling and now Anthropic is like [ __ ] we should have I wish we had a lot more compute. OpenAI is able to pay the bills perfectly fine. In fact, they've raised a ton of money to get incremental compute in addition
[27:08] to the irresponsible levels of compute that they were buying from Oracle and Core and SoftBank and all these people and Microsoft uh you know such as Tranium. Now they're getting tranium as well from Amazon. Um so so they've done this like insane thing on compute and they need know they also know they need more. But what's interesting is if you were to say Opus 46, you know, let's ignore models getting better over time. Let's just take diffusion of this technology. You and I may get jump on the model immediately day one, but other businesses take time and it takes time for people to learn and the spark of oh
[27:40] [ __ ] claude psychosis moment doesn't hit everyone at the same time. And so by the end of the year, let's say a 46 opus tier model the economy would spend $und00 billion on. I don't think that's unreasonable. It's spending $40 billion right now. >> That's like a linear extrapolation. >> It's a linear extrapolation, not a not an exponential. To get the exponential, you need the better models. Enthropic won't have enough compute to do that. And so and and presumably OpenAI and Google will hit that tier soon enough. Whoever hits that tier next, sure, Enthropic may get to charge 70 plus percent gross margins, but if OpenAI
[28:11] hits it next, they charge 50% gross margins. They still get all of this incremental demand. And probably they also won't have enough compute to serve all the users. And so sure, maybe Mythos is a model where if the world had enough compute, it'd be $500 billion of revenue or something crazy. There is such demand for these tokens and such limitations on compute, you know, and we see this with H100 prices skyrocketing and the useful life of these GPUs continue to extend. It's pretty clear even the tier 2 lab is going to be sold out of tokens, let alone the tier one lab. The tier one lab
[28:41] will have better margins, but the tier two lab will be sold out and probably the tier three lab will also be close to sold out. Economic value that the best model can deliver is growing faster than our ability to actually serve those tokens to people via the infrastructure. And so this gap will continue to grow and the model labs will continue to have expanding margins until people in the hardware supply chain infrastructure supply chain are like wait no why don't I just jack up my margins. So suffice to say I think the assessment today or your assessment of the demand side is completely explosive in your own particular example here at semi analysis but just more broadly that as people
[29:12] fall in you call it AI psychosis as people fall into this experience of what they can do the implementation difficulty going completely away I I've certainly felt that you know my own token spend is just through the absolute roof just in the matter of weeks so that that feels like a pretty good assessment anything we're missing on the demand side >> if you don't use more tokens you'll never escape the permanent underclass just expand on that. >> So either either you use more tokens and you generate economic value outsized economic value for the use of those tokens. Um a lot of people are doing it the boring lazy way. Oh, I guess I'll
[29:42] just work one hour a day instead of eight hours a day and I'll have AI do most of my job. That's the boring way. The cool way is I'll still work eight hours a day and I'll I'll do 8x the work and maybe I'll make 5x the money. Um maybe not you can't do this with a job obviously. There's people who have multiple jobs. Um there's people who like start companies and start selling stuff. get that economic value on on this AI before everyone is using it and it's table stakes. Uh because it's still not table stakes if you don't use more tokens and generate the value from them and capture that value. These there's three different problems here. Using
[30:12] more tokens, generating value from those tokens and capturing value from those tok uh from the value that you created from the tokens. Uh if you don't do these three things, you'll never escape the permanent underclass i.e. as models continue to skyrocket in capability and the concentration of resources potentially happens. >> Okay, let's talk about supply. what is going on like how would you describe the frontier of what's changing or what is changing at the frontier of supplying the the entire stack that's required to serve all these tokens as the demand curve explodes >> as demand skyrockets prices are going up
[30:42] for everything on the supply side um whether it be the NGPUs uh their prices are going up in addition their useful life is extending >> H100 prices look like this >> yeah exactly there's people who have argued GPU's full lives are less than 5 years complete nonsense Um, there are clusters now resigning, three or foury old hopper clusters resigning for three or four more years. Um, there's A100 clusters that are resigning for another couple years. So, the useful life is clearly not 5 years. It's maybe even seven or eight years. Um, arguably, we we don't know yet.
[31:13] We'll see. We'll see when Hopper gets there, but it it's clearly not 5 years. So, the useful life is extending and the prices are going up on that renewal. So, in effect, the gross margin was not 35% on a cluster, it's beyond that. Um so margins are expanding in the in the cloud layer. Margins are um extremely healthy on the hardware layer with you know Nvidia still charging 75 or whatever percent gross margin as we move down the stack. Memory obviously margins have skyrocketed there. Places like optics and logic there are large
[31:44] prepayments um and margins are growing slowly um more so the companies that are making chips like Nvidia are paying huge prepayments. So in effect, the cast of capital or timing of cash flow, the return on invested capital is going up even if the gross margin isn't. And you see this across the whole supply chain. You see ASML is completely sold out and they need Carl Zeiss to expand faster. Everywhere along the chain, everyone's either sold out and margins are going up or they're getting prepayments increases the return on invested capital because the invested capital is lower. And so
[32:14] this is like a consistent trend across any part. It's it's even like you know a PCB to make a PCB requires copper foil and that copper foil is sold out and people are making prepayments for it. It's like anything and everything that like has a pulse and is like sold out people are like jumping to get more incremental supply and fighting over the supply for the years after. >> As your business scales up, everything gets more complex, especially your compliance and security needs. With so many tools offering [music] band-aids and patches, it's unfortunately far too easy for something to slip through the cracks. Fortunately, Vanta is a powerful
[32:45] tool designed to simplify and automate your security work and deliver a single source of truth for compliance and risk. There's a reason that Ramp, Cursor, and Snowflake all use Vanta. It frees them to focus on building amazing differentiated products, knowing that compliance and security are under control. Learn more at vanta.com/invest. I know firsthand how complex the tech stack is for asset management firms. And seemingly every new tool and data source makes the problem even worse. Adding more complexity, more headcount, and more risk. Ridgeline offers a better way
[33:15] forward. One unified platform that automates away that that automates away that complexity across portfolio accounting, reconciliation, reporting, trading, compliance, and more. All at scale. Ridgeline is revolutionizing investment management, helping ambitious firms scale faster, operate smarter, and stay ahead of the curve. See what Ridgeline can unlock for your firm. Schedule a demo at ridgeline.ai. What do you think are the most important bottlenecks? Like typically in economic history when there's this kind of demand, supply reorients and rises very very quickly to meet the demand. It
[33:47] seems like it's almost impossible for supply right now in this moment to keep up. You know, famous last words, every every shortage is followed by a glut historically. But what are the most interesting bottlenecks to you on across the supply side? >> Supply chains are usually very fast to react. Um, one unique thing is that our supply chains now are more complex than ever. and the things we're building are more complex than ever and therefore the lead times are longer. Um, and it's not like we haven't seen 18-monthl long lead times in other industries. It's just building incremental supply didn't take
[34:17] years. Um, and this is the case with memory, right? Memory can only grow capacity, you know, low double digit percentages a year, right? 20s 30% a year. Um, even less for NAND, a little bit higher for DRM. Even though the demand signal was very strong at the end of 2025, the memory companies immediately sort of started reacting. None of that incremental capacity really gets here until the second that they've decided to do in addition to the typical 20 to 30%. You know, they can stretch a little bit, but really the true incremental supply doesn't come till 28, which is a very unique thing. Even if
[34:48] they wanted to build as fast as possible, it doesn't come till 28 uh early late 27 at best. And so the result is memory prices have, you know, gone through the roof. And guess what? they're going to double or triple again. Um, at least on DRAM especially, people are like, "Oh, the memory storage is overplayed. Everyone gets it." And it's like, "No, no, no. You don't get it." DRAM will double or triple from here still because that's that's how much capacity is required and they have to steal capacity from somewhere else. And the only way to steal capacity from somewhere else in a in a capitalist
[35:18] economy is demand destruction via higher pricing. We're not like rationing stuff here. And so ultimately, that's what's going to happen. And so margins continue to go up. Um, I think Logic also has humongous uh capacity problems. TSMC just had their earnings. Uh, they keep upping capex. Ultimately, you know, it takes them quite some time to build fabs. Um, they're trying to do everything they can to squeeze every little output out of every fab that they have. But ultimately, they're not raising prices fast because they're good people. It seems like, um, you know, singledigit price increases instead of, you know, tripledigit price increases
[35:49] like the memory guys have had. And so you ultimately have like this like market where yeah TSMC is a great company but are they are they actually going to extract all the value? I mentioned things like copper foil, glass fibers for PCBs, lasers. These are things that are like well understood and niche supply chains but they're very very tight. Um and ultimately upstream the semiconductor wafer fabrication equipment supply chain is one that like I still think is it's gone up a lot but it's still very underappreciated. TSMC capex this year they say 56. Uh we've had 57.4 4 billion since January. Um,
[36:21] and we may up it slightly more just because we see some some ways that they can get incremental capex. But what people aren't focusing on is what does that mean next year and what does that mean the year after? And it turns out 3 years from now TSMC is going to spend hundred billion on capex. U maybe two years from now, right? Might be 28. Sincerely, they may spend hundred billion on capex in 2028. And people like just can't fathom that. But what does that mean for their downstream supply chains? um you know companies like Lamb Research or Applied Materials or ASML or their further downstream supply chains like MKSI and and all
[36:52] these other companies the tail whip just gets whipped harder and harder and harder and ultimately that's a shortage if you know TSMC wants to spend $100 billion in 2028 which is a real possibility I think people would think that's insane but that's a real real possibility >> what about other parts of the chip ecosystem where GPUs have been completely dominant what about like CPUs or AS6 or things that start to pop out as both opportunities and bottlenecks beyond just like Nvidia's GPU dominance. >> Yeah, I mean AS6 are obviously taking off, but I'll sort of pivot away from AI chips to talk about these other things.
[37:22] There's a project we did on FPGAAS and there turns out there's 120 FPGAs per per um next generation rack um AI rack and then like what about all the FPGA names CPU wise all these reinforcement learning environments plus all the slop code you and I are generating that is now running on some you know Versell instance or whatever it is um or some AWS instant or some bucket that we've spun up all of that requires CPU and so CPUs are completely sold out and demand is skyrocketing there >> help people understand the role that CPU plays and everything.
[37:52] >> Yeah. So there's two there's two main reasons why you need tons of CPUs. One is when you're doing reinforcement learning um the CPU is very critical to that. So so before you would throw all the internet's data into the model, train it, spit it spits and it it spits some stuff out. Now you train all the world's internet you put all the internet data into the model. Then you put it in this environment. This environment is like hey model try this out and it tries stuff out. tries a bunch of different things and in the end there is an environment which scores whether or not what it tried out is successful and it grades it. And these
[38:23] environments can be anything. It can be, hey, check if the text was outputed in the right way, structured outputs. It can be very simple stuff. It can be very complex stuff. Um, and people are starting to get into very complex things, right? Like, hey, I want you to open this file, change it, edit it, update it, submit it to this website. I want you to open up this physics simulation from Seammens and edit this CAD model. So the environment is going to get more and more complex and those environments run on CPUs. They don't run on GPUs. They don't run on AS6. The AS6 run the model that takes the input data
[38:53] from the environment, runs it through the model. The model creates outputs of various different trajectories, right? Ways that it think it could solve it um in different instances. Those trajectories are graded/scored and the ones that are successful you train on and you update and you reiterate and you iterate iterate iterate. And so CPUs are very useful for that one. And then once you have these great models and you're deploying them, those models are generating code. They're generating useful output. That useful output, it doesn't go from a GPU straight to the
[39:23] human brain. Um, it goes from a GPU or an ASIC through to, you know, a deployed app that you're deploying somewhere that actually just runs on CPUs. So that's another area where there's a lot of demand and things are sold out um in a large large way. As you continue to assess and try to be the world's best informed person on both the trajectory of supply and demand, what are things that you wish you knew to make that understanding that you don't know? >> I think the hardest area for us um and for everyone is understanding
[39:53] tokconomics, economics of tokens. Um, I think we have a really tremendously like good insight into how much it costs to run infrastructure, what the cost of tokens are, what the cost of models are, what the margins of these labs are, but the usage and adoption is what's really difficult to model, you know, continuously, right? We we have these like we had like crazy in January, we had crazy estimates for February, anthropic smashed them. How do we calibrate this model? What are the data sources for this? February, uh, we had crazy assumptions for March and then they smashed them. And everyone sees the
[40:23] number of 10 billion and they're like what the how do they add 10 billion in revenue? Who is using all these tokens? Why are they using them? What are they building with them? And then more importantly with what they're building with these tokens, how is that actually diffusing into the economy? And what value is that generating? Because it's not really something that you can capture in any any GDP statistic, right? all of the value of the tokens that I use get transformed into better information which I then sell at a discount to what people used to sell information for relatively because um and therefore that information is now
[40:53] making its way throughout the economy and and people are making better investment decisions or better competitive decisions if they're a semi company or data center company or hyperscaler and now how how much what what is the value of this and what has that what has that done to the economy it's clearly by every subjective metric Amazing. But where is the phantom GDP? What is the phantom GDP? How do we track the real economic? Because because the GDP metrics are not, you know, accurate. If you were to say what is the GDP that
[41:23] Dylan Patel is making, it's tiny compared to what the value that I think is being created. And so ultimately, what is the value being created by these tokens? Not on a basis of, you know, just simple, you know, what is the knock-on effect, right? What is the knock-on effect of all the things that these things are doing? And I think that's the real uh question. and challenge uh that's hard to measure. I think we've got an tremendous, you know, reading on the supply side of things. I think we've got a tremendous reading on even a lot of the demand side signals, but it's it's what is the value these tokens are generating that's hard to quantify and measure.
[41:54] >> I hope we get a chance to do this like every 3 months because this changes so quickly. What do you think is going to happen next? Like when I when I come back 3 months from now and we're in San Francisco together again, what do you expect? >> Large scale protests. >> Really? >> Yeah. I think there will be a large scale protest against anthropic >> end AI. >> Expand on that a little more. >> People hate AI. Um AI is less popular than ICE, less popular than politicians. Confused how Pew surveyed this, but apparently AI is less popular than politicians. You know, with Enthropic adding so much revenue, that's going to
[42:25] start causing business changes downstream. People are going to get more and more scared of AI. they'll start blaming more and more of their own problems and things that are, you know, global, you know, have been deep-seated problems for a long time. Those will bubble up and be blamed on AI. Um, probably some politician or some social media people will start to be able to take uh influencer will be able to start taking and weaponizing AI against people. You look at the comments of news articles where Sam Alman had a Molotov cocktail thrown in his house twice in
[42:55] like two weeks. They're like, people are cheering it on. Uh, and this is just the beginning. So, I think I think we'll see large scale protests against AI in three months. >> What is the counterweight to that? Like, how should the AI industry head that off? >> First of all, Sam Alman and Daario have to stop getting on interviews. They're so uncarismatic. I don't know what they're doing. Every interview they do is like, wow, normal people are going to hate you even more. Like, Sam being on Tucker Carlson probably made all Republicans hate OpenAI. And same with Dario. They just have no charisma. I think that's first.
[43:26] Two, they need to start showing uplifting things that can be done with AI. Um, three, they need to stop talking about how the capabilities are going to change the whole world constantly because then people are going to get fear of that capability because they have no connection. >> They don't know how to use it. Yeah. >> There's no connection to it either. It's like the average person doesn't know an anthropic employee. The average person doesn't know an open employee. average person doesn't know who these people are, what their goals are, and they just view them as like this like sneaky cobball of like 5,000 people at this company that are going to change the world and automate all the jobs and and
[43:57] destroy society. That's what they view it as. And and as people who are funding the building of all these data centers and and power plants that are going to pollute the world, right? They don't quite understand what's happening. You know, they have to stop talking about the future thing that's going to happen and only talk about present, how uplifting AI is. I think it's a huge reorg and rebranding that needs to be [music] done. >> I love doing this with you. Thanks for your time. >> Awesome. Thanks. >> Your finance team isn't losing money on big mistakes. [music] It's leaking through a thousand tiny decisions
[44:27] nobody's watching. Ramp puts guard rails on spending before it happens. Real-time limits, automatic rules, zero firefighting. Try it at ramp.com/invest. [music] As your business grows, Vant scales with you, automating compliance, and giving you a single source of truth for security and risk. Learn more at vanta.com/invest. Ridgeline is redefining asset management technology as a true partner, not just a software vendor. They've helped [music] firms 5x and scale, enabling faster growth, smarter operations, and a competitive edge. Visit
[44:58] ridgelineapps.com [music] to see what they can unlock for your firm. Every investment firm is unique, and generic AI doesn't understand your process. Rogo does. It's an AI platform built specifically for Wall Street, connected to your data, understanding your process, and producing real outputs. Check them out at rogo.ai/invest. The best AI and software companies, from OpenAI to Cursor to Perplexity, use Work OS to become enterprise ready overnight, not in months. Visit works.com to skip the unglamorous infrastructure work and
[45:28] focus on your product.