The guy trying to make GPUs trade like barrels of crude just got a year closer to making it real.
The Summary
- Don Wilson, founder of trading firm DRW, is building a market for GPUs that he believes will eventually be bigger than oil
- DRW is developing a "crunch spread" for GPUs (analog to oil's crack spread) and planning GPU futures contracts
- The timing matters: tech selloff hitting neocloud providers like CoreWeave while compute demand keeps climbing
The Signal
A year ago, Don Wilson said he wanted to build a GPU market bigger than oil. Most people thought he was doing bit comedy. Now DRW is developing actual derivatives infrastructure for compute, and the timing couldn't be better. Or worse, depending on where you're sitting.
Wilson's bet is that GPUs become a commoditized asset class with standardized contracts, price discovery mechanisms, and hedging instruments. The "crunch spread" he's building measures the margin between raw GPU capacity and processed compute output, the same way crack spreads measure the margin between crude oil and refined gasoline. If that sounds abstract, think about what it enables: compute buyers could lock in prices months ahead. Cloud providers could hedge capacity risk. Investors could trade compute exposure without touching hardware.
"A year's time in the age of AI is an eternity."
The market structure matters because right now, GPU pricing is opaque chaos. You can't price forward compute with any confidence. You can't hedge your AI infrastructure costs. There's no term structure, no standardized contracts, no real price discovery beyond "what will Nvidia let you pay." Wilson is trying to fix that with the same toolkit that made oil, natural gas, and electricity tradable at scale.
But here's where it gets interesting: the tech selloff and credit concerns around CoreWeave show exactly why this infrastructure is needed. CoreWeave raised billions to buy GPUs and lease them to AI companies. Now credit markets are spooked about whether those revenues are stable enough to service the debt. If there were liquid GPU futures, CoreWeave could have hedged that exposure. Lenders could price the risk better. The whole capital stack would be less fragile.
Key developments Wilson is tracking:
- GPU futures contracts that let buyers lock in compute capacity and price
- Token futures, creating a bridge between crypto rails and compute markets
- Price discovery mechanisms that work across cloud providers and hardware generations
The token futures piece is the sleeper. If compute becomes a tradable commodity, and tokens are already tradable commodities, then settling GPU contracts in tokens instead of dollars starts to make sense. Especially for AI companies that might want to pay for compute with revenue from AI services priced in tokens. This is the kind of plumbing that makes Web4 possible: agents that can programmatically buy compute capacity, hedge costs, and settle in native digital assets.
Wilson's timing is deliberate. AI compute demand is still climbing while credit markets are getting pickier about who gets to finance GPU buildouts. That spread creates opportunity for anyone who can bring real market infrastructure to the space.
The Implication
Watch for GPU futures to launch in the next 12 months. If DRW gets the contract specs right and finds liquidity, this becomes the rails for how AI infrastructure gets financed, priced, and traded. For AI companies, it means you'll soon be able to hedge your compute costs the way airlines hedge fuel. For crypto builders, it means compute, the actual resource your agents need to run, might soon trade on the same infrastructure as your tokens. The convergence is coming faster than people think.