Oct 19, 2026 · 32 min
Trading firms test markets for the AI compute crunch
Don Wilson on the GPU Market's Trillion Dollar Question
GPU capacity is becoming a financial risk, but building markets around it raises difficult questions about regulation, depreciation, credit, and technology changes.
- 1Compute futures could help neoclouds manage volatile GPU demand, pricing, revenue, and borrowing risks.
- 2GPU contracts depend on credible price indices and long-dated forecasts as chips depreciate and new technologies emerge.
- 3Perpetual contracts can serve economic purposes beyond speculation, but their leverage and regulatory classification remain contested.
Don't miss
Donald Wilson connects GPU markets to commodity-style spreads, proposing a “crunch spread” for the economics of turning compute into AI services.
The brief
At a live Odd Lots recording in Chicago, DRW founder Donald Wilson makes the case for financial markets built around GPU capacity and AI infrastructure.
The core problem is volatility: neoclouds need protection against swings in GPU demand, pricing, and revenue, especially when heavy borrowing magnifies credit risk.
GPU futures have not launched because regulators and market participants must first agree on what the underlying price index measures and whether it can support a contract.
Wilson argues that long-dated futures are essential because GPUs depreciate, chip generations change, and market prices—not arbitrary assumptions—must determine future value.
The conversation’s sharpest turn comes with the proposed “crunch spread,” a commodity-style way to think about the economics of converting compute into AI services.
The discussion closes on perpetual contracts, asking whether flexible, continuously traded instruments can manage risk without simply disguising leverage as innovation.
What was said on this episode
27 statements · 15 positive · 4 negative · 1 mixed · 7 neutral
Rapidly growing AI demand will experience expected volatility rather than constant growth.
“that's growing this fast is not going to grow at a constant rate. It's not going to grow at a, you know, 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.”
Listen at 3:20
Selling compute to less creditworthy customers raises neoclouds’ cost of capital.
“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.”
Listen at 4:48
GPU futures can substantially lower neoclouds’ cost of capital.
“Their cost of capital is going to... It's going to be much lower.”
Listen at 5:34
GPU futures can be listed November 9 if the CFTC takes no action.
“if they don't do anything, then, you know, the futures can get listed November 9th.”
Listen at 6:10
Silicon Data has collected a large amount of GPU pricing data.
“Silicon Data, that team has done an amazing job of collecting massive amounts of data”
Listen at 7:19
The future value of a B200 is unknown over four-to-five years.
“nobody knows what the value of a B200 will be four years from now, five years from now.”
Listen at 8:59
H100 hourly on-demand prices had been declining before Thanksgiving.
“the price of H100s, the hourly on-demand price, had just been declining, declining, declining.”
Listen at 9:39
H100 hourly on-demand prices rose around Thanksgiving.
“right around Thanksgiving, they turned around and kind of went up and to the right.”
Listen at 9:52
GPU prices are volatile and require a hedging instrument.
“these things are volatile. They can go up, they can go down, and a hedging instrument is badly needed.”
Listen at 10:38
GPU forward contracts will trade at lower prices due to strong selling demand.
“those forward contracts are in fact going to be at lower prices just because there's so much hedging demand to sell in the forward space.”
Listen at 13:41
The open-source AI inference ecosystem is already sufficiently broad.
“we already have a really rich ecosystem.”
Listen at 15:14
H100 and B200 spending is sufficient to support futures markets.
“the amount of money going into just H100s or just B200s is large enough to support a single futures market.”
Listen at 16:31
Alternative AI chips are highly correlated with major GPU prices.
“these other things are very correlated.”
Listen at 16:42
Token futures will eventually become tradable.
“eventually, and we may be getting close, you'll be able to trade futures contracts on the output, on the tokens themselves.”
Listen at 17:24
GPU-to-token inference margins resemble refinery or soybean-crush spreads.
“that profit margin that you can generate is akin to what an oil refinery does, right? Or what a soybean crush plant does.”
Listen at 18:14
DRW is actively working on token futures.
“Yes.”
Listen at 18:46
DRW’s token-futures project remains at the conceptual stage.
“it's the thought stage”
Listen at 18:51
Exchanges will launch futures for major chips, not every chip model.
“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 the major ones.”
Listen at 19:34
NVIDIA’s market share will decline while the company continues growing.
“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.”
Listen at 20:28
DRW’s token usage is growing exponentially.
“The token use is like this exponential chart.”
Listen at 22:46
Not every inference request requires a frontier model.
“not every inference request needs to go to a frontier model.”
Listen at 23:24
Optimal routing can minimize token spending.
“minimize token spend by optimally routing inference requests.”
Listen at 23:33
Perpetual contracts serve a useful economic purpose and are futures contracts.
“My view is that perps serve a useful economic purpose. They are futures contracts.”
Listen at 24:06
A Brent-index perpetual could make oil ETF exposure more efficient.
“I think that that would be a much more efficient experience for the ETF.”
Listen at 25:02
Perpetual contracts can pass through stock-borrow benefits to retail investors.
“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.”
Listen at 26:13
Perpetual contracts can operate with reasonable leverage and no auto-deleveraging.
“You could list a perp with reasonable leverage and no auto deleveraging, and it would be a very useful economic instrument.”
Listen at 28:35
Round-the-clock financial trading will occur and markets must adapt.
“You know what? It's just going to happen. We're going to have to figure out how to deal with it.”
Listen at 29:48
Statements are attributed to the speaker as said on the episode and reflect their view at the time, not PodLume's. They are not advice.
Weekly email
State of the Pod, every Monday
The week's most-discussed names and stories from podcasts, with the moments behind them. Free, five minutes, unsubscribe in one click.