Worldbuilders: The Largest Infrastructure Project in History with Evan Conrad (SF Compute)

14:48Evan Conrad:Because if you would raise the big round, you would raise the big round to buy a long-term block of compute. And then if your product didn't work, you were in the same position that we were, just on a bigger scale. And you need to be able to get out of that contract because if you don't, your business blows up.

15:04Sumeet Singh:Back on the hook.

15:05Evan Conrad:Yeah, because at the time that we were doing this, businesses blew up. Like, there is no maker of, like, I don't know if I want to say the name. But there were big AI labs that sort of imploded because they had bought too many GBUs because they had like a$100 million contract. And if the revenue doesn't come in to support that, the company doesn't exist anymore. Why did the suppliers on that side take that credit risk essentially then? I think they were assuming that it would work like CPUs. In CPUs, people take credit risk all the time, just at a much smaller scale, typically, relative to their balance sheet.

15:44Evan Conrad:I think everyone who took credit risk at that scale basically lost money or blew up or made less money than they should have. So coming soon, there's not really going to be on demand at any serious scale for the next chips. So like in A100s and H100s, yeah, Yeah, people built speculative clusters, and then were trying to sell them on demand, or were taking on a lot of credit risk. And so it meant that you could kind of get this. In Blackwell, you just can't. The customer has to sign a long-term agreement before the cluster is built. MARK MANDELAVYSCHENKOVICHERSON WONG - Why is that? What's the sort of key difference, you think?

16:22Sumeet Singh:MARK MANDELAVYSCHENKOVICHERSON WONG - Because everyone got wiped out. MARK MANDELAVYSCHENKOVICHERSON WONG - They just learned from the last cycle, basically. There was a cycle that happened, and insiders know. Outsiders don't necessarily see that.

16:30Evan Conrad:MARK MANDELAVYSCHENKOVICHERSON WONG - Correct.

16:31Sumeet Singh:Yeah, we saw it because we were the brokers.

16:33Evan Conrad:And we had an order book. And we could buy H100s for$0.40 an hour on us. Because the people that owned the chips actually

16:42Sumeet Singh:had to get rid of them as well.

16:43Evan Conrad:Yes, because they had built speculative clusters and needed to get out of their speculative position. And capacity on that time was idle, and so prices went down. Nobody in their right mind would do that today. People will build Blackwell clusters, the new chip, in small scales and then offer that on demand at really high prices. But nobody's going to give you a really big Blackwell deployment unless you sign up for a contract up front. Small scale, totally. That's reasonable and it's a good decision to do. But big scale, you have to get offtake. That's the key word for anyone outside of the industry.

17:26Sumeet Singh:So what can you explain, like, what do you mean by an offtake engine, right? You recently raised a Series A to build this offtake engine. Yeah. Like, what does that mean? What are the different parts of it? What are the different components? Like, what does it actually mean to build an offtake engine for the world?

17:40Evan Conrad:Offtake is like the most important word in all of AI. Offtake is the contract that you get, presumably before you build the cluster, that you take to your financier, the person who's providing the money of some sort, either internal capital or external capital, and you say, I'm financing against the credit risk of this offtaker. It's a long-term contract. That's all it means. We're building a company right now. So the way that we've designed our order book is specifically to generate off-tick, to make it so it's really easy for someone to get off-tick. That's what we do. How do you make it easy to get off-tick?

18:20Evan Conrad:Everyone who is in the position of being an off-taker, someone who's willing to sign up for a long-term contract, was in the same position that we were as the beginning of SF Compute. And these would be people on the demand side of your equation. So anyone who's an AI lab that is trying to buy capacity right now, when they are positioned with two options. One option is I can sign a three-year contract or I can sign a three-year contract that I have the ability to get out of. To sell back. To sell back at some price. It is so much better to get the one that you can sell back than not. However, none of the GPU clouds want to do this because you will obviously make less money if you do that.

18:54Sumeet Singh:What is going through the head of the demand side customer here, the AI lab? Like they're thinking, hey, if the demand for my product changes, I'm still on the hook for this contract that I've signed with the cloud, right?

19:04Evan Conrad:I think if you were a smart AI lab, you were looking at all the AI labs that blew up, like a lot of them did. And we don't talk about them anymore. And they wiped out everyone involved. Including the suppliers? So the suppliers, sometimes, yeah. Sometimes the suppliers got hit pretty bad from it. Sometimes, actually, the people that got wiped out was the venture capitalists. If you were smart, if you were a smart cloud, what you did is you got payments up front. And so what happened when the cluster was idle and you weren't bringing in any money is you just wiped out that initial capital that you paid down.

19:37Evan Conrad:And the person who's paying for that is a venture capitalist. So both, I hope the venture capitalists are being smart and encouraging people not to sign long-term contracts. Like, to be clear, the GPU cloud should always ensure people sign long-term contracts. But the demand side... Right, they're financially incentivized. They're financially incentive. Everyone, if they were operating smart, would be trying to do the opposite. They're negotiating the midpoint between us.

20:03Sumeet Singh:Yeah, I often describe SF Compute's role then actually as helping solve for the Rubik's Cube of problems that the supply side is facing and the Rubik's Cube of problems that the demand side is facing. Yes. That is the offtake engine. Yeah. And so what does that end up ultimately shaping up to be this offtake engine? The different components, different paths that you could take here.

20:22Evan Conrad:Yeah. So are you asking, what are we doing today? Yeah, what are you doing today? Yeah, OK.

20:25Sumeet Singh:MARK MANDELAVYSKI - It's been a journey since the early days. MARK MANDELAVYSKI - It has.

20:28Evan Conrad:MARK MANDELAVYSKI - OK, so the first thing is that we're building our own clusters right now. But the way that we're doing that is we're working with outside financial parties who would like exposure to compute. We build clusters for them. We then list them with us.

20:40Sumeet Singh:MARK MANDELAVYSKI - And they're viewing, these are new participants to the market?

20:43Evan Conrad:Are they viewing this as a new? MARK MANDELAVYSKI - Not always. Sometimes it's people who have already been doing this for a bit. But a lot of the world's money would like to turn into GPUs. We help them do that. We look like a property manager. So if you're used to investing in real estate, you might work with a property manager or a developer who builds a building for you, who then operates some sort of rental agreement on it, a hotel, does something with the building that generates a revenue stream that comes back to you. That's what we're doing. We're taking, we're basically cutting out the GPU cloud in the middle.

21:19Evan Conrad:And the reason why we ended up doing that is because we couldn't get access to the rest of the components of the cluster. So our problem in the beginning was we had this cluster, and it would break all the time, and we didn't have the ability to fix it.

21:35Sumeet Singh:Right, it wasn't yours.

21:36Evan Conrad:It wasn't ours. And we also ended up having to build all our software stack that is built on top of the hardware the worst possible way. So in a big GPU cluster, there is the node you are selling, the bare metal node you're selling. But there's also everything else. There's the switches and the DPU and the BMC and the UFM. It's like all these other components around it. If you don't have access to these other components around it, you end up needing to go down a technical path that is much, much, much harder to do. And you kind of can't use any off the shelf components. and some things you can't do at all.

22:17Evan Conrad:That's very painful. So to solve that problem, we ended up saying, we're going to design the bomb. We're going to build the cluster, and then we're going to own the whole thing.

22:29Sumeet Singh:MARK MANDELAVYSCHENKOVICHER You're going to build the cluster and build a cloud,

22:31Evan Conrad:essentially, as well? MARK MANDELAVYSCHENKOVICHER So we built the cluster. Well, build here, meaning we're working with other vendors to do the physical construction and so on. But we're designing the bomb, the bill of materials. Because we have built and owned the whole cluster, We also operate everything like a cloud does. That's a much better technical position to be in than the one that we were in before. But what we do is we work with other financial parties who would like to own clusters and like to get the revenue stream from them because they're taking a position on the industry of compute, the industry of AI.

23:03Sumeet Singh:They're participating in this model economy is what I call it. Yes. As you said, they want exposure to this asset class, to this economy. It has essentially become its own economy. It's the model economy, as I call it.

23:14Evan Conrad:MARK MANDELSKI - This is an objectively worse way to build a cloud in some ways because we, as a company, end up making less money than other GPU clouds. And that is kind of on purpose, because what we're doing is we're splitting that money with the capital party that's putting down the capital. MARK MANDELSKI - Right.

23:30Sumeet Singh:You don't take your own capital risk. You don't take your own credit risk.

23:34Evan Conrad:MARK MANDELSKI - Yeah. So the trade-off here is, we have less risk as a company. We're very happy to do that, because of our early days as being massively risky. We really don't like that. So we get to be a stable entity. The benefit of that is we're selling that risk to someone else who's going to make a lot of money from it. That's a pretty good tradeoff from, I think, both parties' perspective because it gets the financial party exposure to AI. It gives us stability so we can build a product that's great over time.

24:04Sumeet Singh:And these financial parties are not necessarily AI people or cloud people right there? Sometimes they totally are. Sometimes they totally are. They're funds. Interesting. Yeah. Yeah. That want exposure to this.

24:15Evan Conrad:Yeah. So it's a different way to get exposure. It's a way to diversify what you already have. It's a way to not depend on a specific cloud provider. Like you could invest in a particular cloud provider, but if that cloud provider's contracts go bust, you are not in a great position. So it's a way to diversify your portfolio. And then it's also a way to just get more direct access instead of having this like intermediary between.

24:35Sumeet Singh:If SF Compute didn't exist, how do you think the story ends for both sort of, like neoclouds, these financiers, even startups.

24:44Evan Conrad:MARK MANDELAVYSKI - One of the things I think that SF Compute does is it reduces prices for Compute broadly. The way almost all the clouds have set up their capital structure and the way their company works is we're going to make great software, we're going to charge a high markup for it, and then we're going to get really great margins from it. The problem of this is this is divorced from what customers want. It's what the public markets would like them to be. Like, this is why they will present this. But it is not what any buyer of GPUs wants. What every buyer of GPUs wants is they want reliability, they want security, and then they want better price, and then they want shorter contracts.

25:25Evan Conrad:And as long as you are reliable and stable and you are secure, you are mainly focused on price.

25:34Sumeet Singh:They view it as a commodity, but the reality is, like, it's not as simple as a commodity, I guess.

25:39Evan Conrad:Yeah. So the caveat of this is there are lots of little details in GPU clouds that matter tremendously. Reliability, security, price, contract duration are the key things. On the micro scale, it's like little tiny details about the way that your system is implemented.

25:59Sumeet Singh:So now we know if SF Compute doesn't exist, how the story ends. But if SF Compute does exist and we do exist, Like, what do the other clouds do? Like, what should everyone else in the market be doing? If you're a neocloud, you have two paths that you could go down.

26:16Evan Conrad:And many clouds today are doing both. One path is you go straight to the top of the market and you deploy large clusters without a lot of services on them for reasonable prices to open AI and Anthropic. You get offtake from the biggest buyers. and those biggest buyers are cost sensitive and they have internal teams that can do most of the managed services that you offer. And you should totally try and sell your managed services to them, but you should be okay if they don't buy them because you're mainly trying to scale volume for your clusters. The other thing you should do is you should make managed services.

26:55Evan Conrad:The other path. The other path. So this is things like offering a great Slurm product or great storage products, Or, you know, you can push really high up and offer something that looks closer to like modal. Modal is an excellent managed service. But I think these things are hard to do in the same company because you end up focusing your resources on both at once. And they kind of don't necessarily overlap with each other. An amazing example of this is modal's product is amazing and really good. They are an excellent managed service. they do not build clusters.

27:32Sumeet Singh:They focus exactly on that one thing.

27:34Evan Conrad:Yes. And as they scale and they raise VC and they make profit, their money goes into making their thing even better. And their managed service is leagues ahead of any GPU cloud, any normal cloud. It's way better than anything anyone has ever produced because they have focused directly on that. Meanwhile, modal cannot sell like a big GPU cloud or GPU cluster to OpenAI. Like, I don't think that's going to happen. Because the amount of expertise and experience you need in order to land a cluster that is signed off by OpenAI is a lot. And I would suspect that you will see pretty much every cloud split into doing one of the two.

28:18Evan Conrad:Probably a lot of them will keep going on one side or the other. But the winners will be the ones who specialize in one area, is my guess.

28:26Sumeet Singh:Where do you think amongst that sort of two paths does SF Compute ultimately end up sitting in them?

28:31Evan Conrad:Our goal is to sell to people who look like managed services. Part of what we're doing is trying to make the best cloud provider that would sell to managed services. It is a place in the market that's almost underappreciated because if you're a big cloud, the big clouds have competing incentives right now. One is the public markets want them to have managed services because they want them to be a high margin business. But then OpenAI comes along and says, here is a massive contract. We're going to make AGI. Would you like this money? And obviously, you should do the OpenAI thing. And so they try to do both.

29:15Evan Conrad:And we think the thing we're trying to do is split those and make the greatest thing we can for people who want to do managed services. And the things those folks want are the same things that we cared about in the beginning, which is short-term contracts or contracts you can get out of. It's price. It's specifically price. And SF Compute is not trying to be a thing that captures a massive margin on top of it. We're specifically trying to make the prices go down.

29:41Sumeet Singh:Right. And the price is so important because of that sort of difference between CPUs and GPUs that we spoke about earlier. Yes. Can you double-click on that?

29:49Evan Conrad:So in CPUs, almost by definition, like your customers are like SaaS companies. It's Gusto and Rippling or something. And Gusto and Rippling does not need more CPUs to make more money. Gusto and Rippling can make great software. And then they can sell it like crazy. And their CPU build doesn't really go up, not significantly. And they end up with top of line margins because they end up with SAS margins. They end up with 70 % plus. MARK MANDELAVYSCHENKOVICIENCY

30:16Sumeet Singh:You can build it once and sell it in infinite times.

30:18Evan Conrad:Correct. That was the beauty of SaaS. That was the idea. That doesn't exist in AI. It's just like never going to exist because of the bitter lesson.

30:25Sumeet Singh:Here at WorldBuild, our model economy thesis, which I mentioned, really was based on the bitter lesson, right? That the general models with scale, with more compute, with more data will beat the clever architecture. Or even for me, the sort of workflows and lightweight applications sort of built around those models and with those models. Can you walk through why sort of GPU buyers, people that even believe the bitter lesson, like those GPU buyers, why are they different from CPU buyers? Can you explain the economic reality?

30:58Evan Conrad:In CPUs, the customers of the cloud are SaaS companies. And SaaS companies, almost by definition, have really fat margins. Build something once, you could sell it infinite times. Exactly. And then it gets better over time and so on. The problem of GPUs is that is not how AI works. In AI, you have the bitter lesson. And the way to conceptually think about this is if you're doing some sort of generation, like you're generating a video or something, your customer is paying you for the video, and then you turn around and you use compute for that video. And you do that every time. And so as your volume scales, your margins stay roughly the same, and they don't look like SAS margins.

31:38Evan Conrad:They look really thin. You as the customer need to be price sensitive. It is a requirement of your business. And that means that you push down the prices for your GPU cloud too. And so then your GPU cloud also has lower margins than the CPU clouds do. And you care less about fancy managed services. And you care more about like price as long as reliability and security and so on are there.

32:05Sumeet Singh:Is the why for you here ultimately then? like the why around SF Compute to lower the cost of compute?

32:11Evan Conrad:Yeah, totally. If price is everything. Totally. San Francisco Compute is completely distinct in GPU clouds in that we are trying to give you the best possible price. Almost every GPU cloud in existence and every cloud is positioning themselves to the public markets as we are the premium experience, we offer these great managed services, and therefore you should reward us with a high markup for our GPUs. They tell the public markets that. Meanwhile, customers don't agree. Customers instead say, if I'm buying a three-year contract with you, I better get a discount. And the GPU clouds behind the scenes do that when they can.

32:49Evan Conrad:We are explicitly trying to make the cheapest possible GPU cloud. That doesn't mean we're sacrificing our reliability. That doesn't mean we're sacrificing on the product. But if you wanted to design a business model that produced the best possible prices, you you would end up with SF Compute, basically. MARK MANDELOVICIUSSKI - Lower prices, the more Compute can proliferate. FRANCESC CAMPOYOVICIUSSKI - The world gets better when you make prices lower. The world gets better when things are more affordable. It's almost like a rule of reality. And at the moment, I think a lot of what SF Compute has been doing for a while is implicitly betting on people like lower prices.

33:28Evan Conrad:And so that's what we're going to do.

33:30Sumeet Singh:MARK MANDELOVICIUSSKI - And I guess, through the various twists and turns, And now it's, let's build supercomputers, essentially, to lower the cost of compute.

33:39Evan Conrad:To be clear about why our model makes cheaper prices is that if you're a fund and you got a 20 % return on your fund over the year, that's amazing. That's really good. If you're a startup and you make 20 % margins, it's terrible. You don't want 20 % margins. That's really bad. And so the expectation differences between what a fund is doing and what a company is doing are different. And so the more your company is actually a fund, the more you should be a fund. And the less you should try and be a big sort of company with high margins. The problem is all the clouds have set themselves up and positioned themselves to investors as companies.

34:27Evan Conrad:Meanwhile, their assets under their balance sheet are growing and growing and growing, and actually they look like funds. And they're not. And there's actually better people who are out there in the world who have cheaper cost of capital even than the hyperscalers. There's sovereign wealth funds and massive players right now who are entering the space. They have cheaper cost of capital than most GBP funds.

34:48Sumeet Singh:That is the economic reality, and you are playing for the economic reality.

34:51Evan Conrad:Yes. We are the company that takes no bullshit. We do what we say on the tin. Our entire culture is we never had the opportunity to be fake. We never had the opportunity to pretend to be anything other than the reality of the world. So in the branding of S.F. Compute, you will never see us say, we're going to democratize compute or powering the AI revolution or something. It's like, no. We sell GPUs. We rent them to you. The reason why you use us? Price. Reliability. Like super simple shit.

35:25Sumeet Singh:It is straight up. It is for an economic reality. Yes. Versus some grand vision. And ultimately, the buyers of the folks that you work with, they definitely should have those missions. Correct. But I think it's just you make it possible. Part of my model economy thesis is that there will be bouts of volatility. We're in a GPU glut one day to all of a sudden we're in a GPU shortage. There is this constant chatter of a bubble. Are we in a bubble? Yes, this is a bubble. But where does SF Compute sit within that sort of bubble spectrum?

35:55Evan Conrad:If there is a bubble, here is where the bubble is. It is that in order to afford a large GPU contract, you go out and you raise the big round. And you then spend a bunch of that money up front to pay for your GPU cluster because you're buying it for a long period of time. Often longer than most companies fully have a plan for. Like password projections. Yes. You do that right at the beginning of the company. And you pay, and you're not paying over time and you don't have any way out. From the GPU cloud's perspective, they're relatively safe. So the bubble isn't in the GPU cloud? I don't think so.

36:33Evan Conrad:Unless I'm wrong about this and lots of clouds are just like way over leveraged on accounts payable risk. I don't think that's actually the case. I think a lot of people are being smarter than they were before. Instead, I think the bubble is in whoever owns the equity of those companies that just raise a whole big round. Because if they don't succeed, then they can't get that capital back. And there's also nothing else you can do with that.

37:02Sumeet Singh:So how do you sort of frame SF Compute as like, you said this thing before, it's like the anti-bubble company. What does that mean?

37:10Evan Conrad:If you think the bubble is in companies that are buying GPU clusters and it's their ownership right over those GPUs, that's where the coyote that's walking off the cliff is, then what you need to do is you need to give those people liquidity. You need to get those people an out that they don't have to blow up their company just because they bought too big of a contract.

37:35Sumeet Singh:Just because maybe their demand changed or it was different than their projections.

37:39Evan Conrad:And if a bunch of companies sell all at once, totally. The compute price will collapse and so on. But it's not going to be zero. There's something you can recoup. Right now, it's zero. Right now, there's nothing you can recoup if you have signed a long-term contract. You're just dead. So that's the only way out as far as we see. So SF Compute's goal to some extent is to reduce the risk of an AI bubble, but where the risk actually is. Not where everyone thinks it is. You have to be right. You have to know where it is. It's a very complicated subject. And lots of people on the internet say all sorts of things.

38:15Evan Conrad:And most of them are wrong.

38:17Sumeet Singh:Yeah. So as this anti-bubble sort of company, I remember when we first met, the idea was that SF Compute would be this exchange for GPUs. Is that still the right way to position yourself to be the anti-bubble company? Or is there something else?

38:31Evan Conrad:So I think that was the way that we were framing it at the time because it seemed like the easiest way to get the point across. I've currently stumbled upon just calling it like Marriott, which I think is the better analogy. If you don't know Marriott, Marriott doesn't own a lot of their hotels. Instead, the hotels are like owned by somebody else, but Marriott manages them. And then, you know, they have their own genre of market for their hotels that they're managing. That's kind of what we are doing. The clusters, financially not owned by us, but run and operated by us. And to do that, what we had to do was we made a finance grade ledger and an order book that had all the weird intricacies of compute built into it.

39:17Evan Conrad:And we built a KYC program and an AML program and the whole thing that you would otherwise do to make an exchange. But typically when we say the word exchange, most people kind of get the wrong idea or they go so far into the financialization of compute thing. They go to derivatives. Yeah. And that is totally a thing we may do in the future. There's lots of people who are trying to build that right now. We think there's lots more infrastructure you have to do before you even get close to that.

39:45Sumeet Singh:You have to settle the compute, as I call it.

39:47Evan Conrad:Yeah. So, for example, lots of people are trying to make indexes. The indexes, as far as we can tell, are built on people's wishful prices. It's like, what do you want your prices to be? Back to the economic reality. Yes. And I would be very terrified of betting against an index that doesn't settle to a real price or is settled to the price that someone wants the thing to be.

40:10Sumeet Singh:By being Marriott, you are actually at the forefront of that. You have information.

40:16Evan Conrad:Yes.

40:17Sumeet Singh:We have real-deal pricing information.

40:19Evan Conrad:And we have the actual settlement price that other folks don't have.

40:24Sumeet Singh:So to finish the conversation, the current landscape, let's say, in AI, it feels kind of like a cyberpunk dystopia, right? Everyone going crazy about Malt's book recently, let's say. There's anxiety about AGI. It's about scarcity. It's quite frantic. But when I think about the world that SF Compute is building, like the San Francisco Compute company world, it's blue skies, green grass, landscapes. Why is it important for you that the future of supercomputing is calm?

40:59Evan Conrad:One thing you will never see SF Compute do is position ourselves as democratizing AI or revolutionizing compute or something. we like to be very does what it says on the tin right anti-hype by nature almost yeah i think you get like clouded by hype your brain doesn't work well and people's brains collectively doesn't work very well and if you actually think that the current moment is really important and we do you actually shouldn't be very hypey because it makes it hard for you to do anything real The current scale of the GPU build out is multiple times the cost of the Apollo project. For a while, people were saying there was going to be a Manhattan Project for AI.

41:43Evan Conrad:We're like 30 times the cost of Manhattan Project by now. You know, it's the cost of a war. It's more than the GDP of various countries. It's a really big thing. It is an important thing that you should think seriously about. And when you get consumed by the hype, you end up kind of being in this weird mental state that doesn't let you make good decisions about what you're doing. And if you actually think it is important, you should slow down, calm down, and think over a longer period of time. SF Compute right now is trying very hard to think, how are we going to do things over the next 10 years?

42:23Evan Conrad:Not how are we going to do things in the next six months? and that is in part because in the beginning of the company, we had to think in the next month, month by month. And that's so bad for your brain and it's also so bad for the things you were doing for other people like your product and your employees and everything.

42:40Sumeet Singh:Well, thank you for joining us. Of course. And I hope the world learned something new about this model economy and where SF Compute sits in that. Yeah. Thank you for having me. Hey, this is Ben Kaznoka, co-founder of Village Global. Thanks so much for tuning in to the Village Global Podcast, where we go deep on all of the biggest topics in tech. If you enjoyed this conversation, please subscribe to our YouTube channel. You can check us out on Spotify, Apple, wherever you get your podcasts. We'd love to see you for the next one.