Kapitelklicken springt direkt hin
Shownotes
When we spoke to Don Wilson, founder and CEO of DRW, in Chicago last year, he told us he wanted to build a market for GPUs that would eventually be bigger than oil. A year's time in the age of AI is an eternity, and so for the first episode of this year's Chicago live show - recorded at City Winery - Wilson joins us on stage again to talk about what's changed and what hasn't in his quest to help build this market. We get his take on the recent tech selloff, including credit concerns around neocloud CoreWeave. And we also discuss how DRW is developing a “crunch spread” (the GPU version of a crack spread), the latest on his plans to launch GPU futures, and the possibility of token futures too.
Read more: Jump, DRW Bet on New Compute Market as AI Boom Faces Scrutiny (bloomberg.com/news/articles/2026-10-16/jump-dr…)
Only Bloomberg.com subscribers can get the Odd Lots newsletter in their inbox, plus unlimited access to the site and app. bloomberg.com/subscriptions/oddlots?in_source=…
See omnystudio.com/listener for privacy information.
Transkriptvom Podcast bereitgestellt · Zeitmarken springen hin
Bloomberg Audio Studios.
Podcasts.
Radio.
News.
Hello and welcome to another episode of the Odd Lodge Podcast. I'm Joe Weisenthal.
And I'm Tracy Allaway.
Tracy, we're here in Chicago, second year in a row that we've done a live show here in a Great city for us.
That's right.
We're going to be talking to someone who we spoke with last year, but a lot has changed in 12 months. This is someone who is trying to build a market for compute, right?
Right.
So we're going to be speaking with Don Wilson. He is the founder and CEO of DRW. He is really at the forefront of this effort to build futures markets for things like GPUs and so forth. So check out the conversation. We talked to you, Don, basically exactly a year ago, last time we were here. We talked about things like GPU futures. But in like AI time, it might as well have been 10 years ago because so much has changed. We want to talk GPU market structure, all that stuff.
The last couple of days.
Have actually been kind of interesting in the AI trade world. Everyone's watching CoreWeave. Everyone's watching CoreWeave bonds. There was an anthropic investor meeting before the IPO.
A little jitters.
Maybe their revenue only grew up, whatever, 50% versus 51%. I don't know the exact numbers. Give us your read right now. Is there a little slowdown happening in AI? Are we shifting into a lower gear? Is there anything in the trajectory that's changed?
Any new thing that's growing this fast is not going to grow at a constant rate. The first derivative isn't going to be constant. The second derivative isn't going to be constant. So to me, this is just kind of expected volatility. I don't know, maybe people went on vacation in August so they didn't consume as many anthropic tokens.
But you're not reading anything more into it than that.
I'm not. OK.
Wait, so with something like CoreWeave, a big part of what you're doing with Compute Exchange and silicon data is this idea that if you create a market for compute, then cloud companies can hedge their future revenue streams and the cost of capital can come down. In a week like this, when you see the yield on a CoreWeave bond at 15%, what does that tell you? Is the suggestion that, well, if compute futures were up and running, all of CoreWeave's problems would be solved?
I do think that it's highlighting the massive need for risk management products. Think about if you're a neocloud and you sign contracts with the most creditworthy people companies in the world, then you're fine, right? But there are a lot of other consumers of GPUs that are not the most credit-worthy companies in the world. And if you're a neocloud, you want to be able to sell to them. And so I think what the market's saying is, if you're selling on-demand compute or you're selling to less credit-worthy entities, then your cost of capital is just going to be higher. Now, imagine once GPU futures exist, you can, as a neocloud, buy the GPUs, deploy them in your data center, sell a strip of GPU futures, and then just sell on-demand
compute if you want. Or sell six-month contracts to less creditworthy companies that can post a big percentage of the contract value up front.
And the market is going to.
Reward companies that do that that basically just have on a giant basis trade heavily, their cost of capital is going to be much lower.
Terry mentioned, okay, the GPU futures don't exist yet. And there's some message from the White House, like national security, et cetera. And he's like, I don't get it. Do you get it? Can you tell us something about why these products don't exist yet? Maybe what the regulatory hesitancy is about the listing?
Yeah, and just remind us what the state is of regulatory approval at the moment.
So right now, there's a comment period that's open. The CFTC extended it. If they don't do anything, then the futures can get listed November 9th. Now, conveniently, November 9th is after November 3rd.
And so maybe....
Some people in Washington will be more comfortable with the Compute Futures launching. I don't know. I can't explain what the concern is. I've heard all sorts of different theories. One theory is that some of the indices are not very robust.
Yeah, okay, so I'm actually very curious about this, the robustness of the indices, because a lot of GPUs are sold via bilateral offtake.
Agreements.
Right.
So it's like Anthropic has some deal with SpaceX.
Right.
To use their compute. We don't really know the price of that. Why should anyone feel confident then when they look at an index of GPU prices that like it actually represents something resembling the save the market here?
Listen, I think it's buyer beware. learn about the index that you're trading futures on. Silicon Data, that team has done an amazing job of collecting massive amounts of data, and it's focused on on-demand compute from neoclouds. There are a lot of neoclouds. There are very few hyperscalers. So their index is focused on neoclouds, of which there are many.
And there is a lot of transparency about on-demand compute. And, you know, they've done a very good job to create an index that, you know, will be IOSCO compliant.
That's the benchmark. Some of the other.
Providers I don't think have ever heard the phrase IOSCO compliant.
So IOSCO is the.
I don't know, it's the kind of group of international regulators that I think in 2013 came up.
With a best practices approach for indices.
Oh, we should do an IOSCO episode one day.
Now that you've been working on the futures contract for a while, can you maybe tell us what you've learned about, I guess, the pricing trajectory for GPUs over time? Because this is one of the concerns, right? You can say you're bullish on the compute markets, But it doesn't necessarily mean that you're bullish on the value of a single GPU.
It can depreciate over time. Maybe it gets more efficient and people move on to new things.
So that is the trillion dollar question. I'm glad I asked it. And that is why these futures need to exist. Because nobody knows what the value of a B200 will be four years from now, five years from now. And, you know, and so by introducing futures contracts that trade, you know, way out, I think initially the CMEs futures contracts are going to go out three years and, you know, as they become more liquid, they'll.
Go out more.
But, you know, that enables people to hedge that. Back in November of, you know, almost a year ago, the price of H100s, the hourly on-demand price, had just been declining, declining, declining, and everybody was saying, why do you need GPU futures?
These things just go down.
And then, you know, right around Thanksgiving, they turned around and kind of went up and to the right. And what happened, you know, it was Anthropix's newest model came out, which was a massive step up in usefulness. And all of a sudden, there was a ton of demand for inference. And, you know, more recently, we've seen, you know, looking at B200, B300, and H100 prices, another surge upward in price. And, by the way, at the same time that the Silicon Data token index price has plummeted.
Yeah, it's interesting.
And so.
Yeah, I mean, these things are volatile. They can go up, they can go down, and a hedging instrument is badly needed.
Just as a hypothetical, let's say GPUs depreciate over time and there's more demand but in general the value goes down which is something we've seen in a lot of commodity markets right in that case does having a market where you're able to hedge your future gpu revenue streams does that just enable people to i guess borrow more and overbuild versus genuinely reduce risk you're just allowing people to borrow against the same risk, ostensibly with a hedge, but the underlying isn't actually moving in your favor.
Does that make sense?
No. I mean, here, the reason it doesn't make sense is that for a futures contract to exist, there has to be a willing buyer and a willing seller, right? So your hypothesis would work if they could just go and sell futures contracts at some price that was made up maybe too high, and there was kind of infinite demand for that. Okay, well, sure. But there has to be a rational buyer on the other side. And so I think what we'll see is that those forward contracts are going to be, are in fact going to be at lower prices just because there's so much hedging demand to sell in the forward space.
And there will need to be, you know.
Investors who want to generate a return by buying those futures contracts.
What is the– if we cast out a few years, what is the sort of ecosystem of clouds and inference providers that needs to exist for this to be a rich market? Because if the future of AI is just– let's say the future of AI were just anthropic and open AI. And the future of cloud were just meta, Alphabet, and Amazon. There wouldn't need to be a futures market.
It would just be all bilateral agreements, offtake agreements, et cetera.
Yep.
So how much like. How fat of a tale of non-open AI and anthropic inference do we need such that this is a useful vault market that cannot just be solved by bilateral agreements?
Well, I mean, I think we're already there. I mean, look at the growing use of open source models. I mean, it's exploded. And you're seeing companies with many billion dollar valuations that have just been created to put open source models on GPUs and sell the tokens.
Right.
You know, like fireworks and together. And so we already have a really rich ecosystem.
Is it rich enough yet?
For sure. Okay, interesting.
Yeah.
Should we talk about perpetuals?
Yeah.
Can I ask another?
Yeah, go for it.
Can I ask, on GPU futures, So one of the other things that people say, the skeptics, is they say GPU or compute is not homogenous enough for a futures market. And there's a few different things. They say there's different software arrays, and that could significantly change the value. There's also the fact that there's, at least in the inference part of AI, there's a lot of other players who want AMD and and etched and all these other tranium and whatever the Microsoft one was and.
TPUs, et cetera.
And this is just way too diverse of a market to really be captured in an index. How confident are you that either the market is going to be homogenous enough for futures or that there will be enough sophistication about the sort of basis between the H100 index versus something similar such that this market can serve a diverse provider base?
Yeah.
I mean, first of all, the amount of money going into, you know, just H-100s or just B-200s is large enough to support a single futures market. Now, in addition to that, these other things are very correlated. And so, you know, people will be able to trade a basis on these things.
Versus whatever it is. And, you know, maybe... obviously, you know, these are going.
To be off of kind of the, the baseline standard configuration, whatever. But, and so maybe, uh, somebody's planning on doing a training run and they need a giant cluster. Well, that's going to bear a premium, of course. But, but, you know, if the demand for inference goes up massively, then guess what?
That training run is also going to be more expensive. So it's just a basis between the things.
That's one thing.
The other thing is, um, that eventually, and, you know, we may be getting close, you'll be able to trade futures contracts on the output, on the tokens themselves.
Oh, token futures. Token futures.
And so, you know, again, there's, of course, a diverse number of tokens and people will put up the same objections. But I think that over time, that's going to be possible. And By the way, that will introduce a really interesting possibility, which is, now think about if you go.
And rent GPUs, you pay the on-demand price.
And then you load up an open source model, and you start selling tokens through OpenRouter or something. Well, can you do that profitably or not? Well, that profit margin that you can generate.
Is akin to.
What an oil refinery does or what a soybean crush plant does.
So we're going to have a crack spread.
So in oil, we call it a crack spread. In soybeans, we call it a crush spread. So you heard it here first. At DRW, we call this the crunch spread.
That's great.
Wait, is this token futures or something you're actively working on, it sounds like?
Yes. Okay.
Okay.
What sort of stage would you say you're at?
I mean, it's the thought stage, but I mean, here, to be fair, the silicon data token index is providing some pretty interesting data and is widely discussed.
So that's a start.
And so just going back to GPU futures for a second, so every time a new chip comes out, that needs to be a separate contract and a new pool of liquidity. Is that right? Is that how you're thinking of it?
Correct.
Okay.
And how do you, I mean, it seems kind of challenging.
You have to like keep.
Up with the technological cycle, which is very rapid.
It's rapid, but not that rapid. Right.
And, and you're probably not going to launch a futures contract for every single one. You're going to launch futures contracts for the ones that are in the major ones. And, uh, You know, but that's okay. I mean, there are different, you know, think about how many electricity contracts.
There are for every little node. And I mean, you know, thousands of them.
There's a lot of oil contracts too. We just usually talk about two of them typically. Just out of curiosity, and this is actually sort of separate from the futures conversation, but I guess like as a technologist with multiple businesses, et cetera, are you, how assailable is NVIDIA's dominance today? Like, what's your read on that? Because this is a constant question, whether any of these other chip makers can actually eat into NVIDIA's position. What's your take there?
Sure. I think that the space is going to continue to grow. And I can't, it's unlikely that NVIDIA will be able to maintain the market share that they have maintained. But I'm pretty confident that NVIDIA will continue.
To grow as a company.
You know, last year when we spoke, we were talking about AI use in general at DRW. And I think you said people were using agents to sort of debate investment theories or risk allocation and things like that. But you weren't using AI to actually deploy risk at that time. Joe mentioned that things change very, very fast in the world of AI. 12 months later, has that changed? What's AI doing for you now?
Yeah, no, I mean, there's definitely more adoption. There's more frontline.
Adoption, for sure.
Has your token budget gone up?
The token budget goes up. Yes, it goes up.
The token use is like this exponential chart. It's pretty interesting.
Do you have, there's another question, like across your businesses, have you figured out optimal internal token allocation. We've asked this to a lot of executives and at least up until recent, they're like, we're just still trying to figure this out, but you must have everyone, of course, everyone wants more. I assume everyone always wants the frontier token cost.
Have you solved the question of like internal token allocation yet?
Well, here, I mean, I think the most important thing is to be thoughtful about how you route the, the inference requests and, and, And not every inference request needs to go to a frontier model. And so that's the thing that we're trying to optimize for, is minimize token spend by optimally routing inference requests.
Now shall we ask about perpetuals? OK, perps, go.
Why is Terry wrong about perps? I mean, listen, here.
In my view, are... Terry's in the audience, by the way.
I know.
But you're also his client, so.
I know.
And Terry and I enjoy talking about this. My view is that perps serve a useful economic purpose.
They are futures contracts. And I mean, here.
I'll give you a couple examples. The, you know, a lot of people like to buy the crude oil ETF. Yeah, the crude oil ETF generally, you know, at one point, it was always in the front month future, and it broadcast to the world exactly what day it would roll from that futures to the next futures contract. And I don't know, there was probably a little bit of slippage in that, a little bit of inefficiency. Now they've kind of like, now they do it over a window and they hold the first three contracts and whatever.
But there's still a lot of slippage.
Now imagine a PERP that is tied to, let's say, the Brent Index.
And an ETF that just buys that.
PERP. I mean, I think that that would be much more efficient experience for the ETF. And of course, that doesn't mean that that's a useful hedging instrument for an oil producer.
It's not.
An oil producer needs to hedge way out the curve. Great. You can have monthly and quarterly futures and PERPs side by side.
They actually augment.
One another and.
And that's a good thing. Here's another example.
Suppose you are a retail investor and you own a hard-to-borrow stock. And all the hedge funds want to short. They're paying 100% annualized to short this stock. If you have that stock in your retail account, your broker is probably not saying to you, hey, by the way, we're going to give you 300 basis points a day because you're holding this Hard to borrow stock.
Okay, that would be 1,000% annualized, but.
30 basis points a day. But whatever. So you're not getting the benefit of owning that hard-to-borrow stock.
Now, if instead you own the.
PER, then mechanically it's going to pass through. And so it's a real opportunity to democratize the financial markets.
Right, so democratize. This word democratize, I've never really liked it because it just sounds good, but I don't know what it means. But in this case, it makes sense because in this case, what you're saying is there is an economic value that is not being captured currently by retail holders of certain stocks because no one's making the offer.
But the demand would be there if they held that same exposure in.
A different way. Correct.
And the same thing for the ETF, the oil ETF. I mean, that's probably a retail investor that wants to have exposure to oil.
Here's a more efficient way to do it.
I get that it might be more useful and economically beneficial to some, but if the regulation says that a futures contract has to have an expiration date and these don't, why should it not?
Where does it say that?
I haven't read that. Does it not?
Not that I've seen.
I don't know. So Terry said they're swaps.
And you said they're futures.
Why are they not swaps?
I feel like this should be an easily settled debate.
OK. So what's the difference? And I asked this question when Dodd-Frank was being drafted. And what's the difference.
Between a future and a swap? I have no idea. OK.
Well, the answer that I got from the person drafting the legislation for Barney Frank was, everybody.
Knows that swaps are risky. What? I'm sorry.
A swap that is cleared has the same economic risk to society, to the holder, as a futures contract does.
I mean, we've gotten like lost.
We've like lost our minds with nomenclature.
But can you address directly Terry's point about leverage in that case? Like how do you manage the.
Leverage from these?
No, I'm sorry.
Again, like there's this confusion. Well, if it's a perp, it must have massive leverage and auto deleveraging and all these horrible things.
Wrong.
You can, by the way, have a quarterly futures contract with massive leverage and auto deleveraging and all these horrible things. You don't. you could list a perp with reasonable leverage and no auto deleveraging, and it would be a very useful economic instrument.
Joe, next year, we're going to cut ourselves out of this conversation entirely and just have Terry and John to do.
This on stage.
It's crazy to think some of our most animated moments on the podcast will now be, what is the definition of a swap?
But this is why.
These are the conversations that we like to have. Okay, when 24-7 trading inevitably arrives?
It's already here.
What are you talking about?
That's true.
It is here.
You're right. Does that change people's lives? Is that annoying? Is it annoying to have to think about risk at 1 a.m. on a Sunday?
Just from a personal life.
Won't someone think of a financial journalist?
I mean, first of all, we've been trading crypto since 2014. Yeah, so you've been 24-7.
So I'm used to markets moving 24-7. I will tell you that our options market makers....
Are not happy about the prospect of, you know, making SOFR options markets over weekends.
Yeah. And you know what? I'm very sympathetic to that.
But they're happening.
You know what?
It's just going to happen. We're going to have to figure out how to deal with it. All right.
Don Wilson, thank you so much for coming back to Outlaws Live. We're just going to do it again next year, the year after that, and the year after that.
That's right.
Thank you so much.
That was another episode of the Odd Thoughts Podcast recorded live in Chicago. I'm Tracy Alloway. You can follow me at Tracy Alloway.
And I'm Joe Weisenthal. You can follow me at The Stalwart. Follow our producers, Carmen Rodriguez at Carmen Arman, Dashiell Bennett at Dashbot, Cale Brooks at Cale Brooks, and Kevin Lozano at Kevin Lloyd Lozano. For more OddLots content, go to Bloomberg.com slash OddLots for the daily newsletter and all of our episodes. And you can chat about all these topics 24-7 in our Discord, Discord.gg slash OddLots.
And if you enjoyed this conversation, then please leave us a positive review on your favorite podcast platform. If you're watching the video, then please like or leave a comment or better yet, subscribe.
Thanks for watching and listening.
Thank you.
