Nobody knows what a GPU will be worth in four years—that's exactly why futures markets exist
H100 hourly rental rates bottomed last year, then rebounded after Anthropic released a new model—proving that GPU value swings sharply with AI demand rather than declining monotonically, which is precisely why futures contracts need to exist.
The video won't play here. Listen to the audio instead:
The argument · tap a timestamp to hear it
Any growth this fast has uneven second derivatives—volatility is the feature, not the bug
Don Wilson dismissed market jitters over CoreWeave bonds and pre-earnings volatility with a mathematical observation: anything growing this rapidly must have uneven acceleration and deceleration. The variation is expected, not alarming. He joked it might just be August vacations reducing token consumption. He was explicit: he read no deeper signal from the dips.
— Don WilsonCoreWeave's 15% yields expose what happens without derivatives: no way to hedge
When CoreWeave bonds hit 15% yields, Don Wilson saw a market-clearing signal: desperate unmet demand for risk management. Neocloud providers selling to creditworthy clients face no problem, but they also want to capture lower-credit customers. GPU futures would let them buy and deploy hardware, sell a futures hedge, then offer excess capacity at spot prices to weaker credit counterparties. Markets reward basis trades like this with lower capital costs.
— Don WilsonGPU futures could launch November 9th if regulators don't act—conveniently after the election
GPU futures remain unlaunched; the CFTC's comment period has extended. Absent regulatory action, they could list November 9—right after November 3rd voting. Don Wilson admitted he can't fully parse the White House's national-security concerns. He's heard various explanations, including doubt about whether the index is robust enough for pricing.
— Don WilsonNo one knows what a B200 is worth in four years—so futures contracts must exist
This is the trillion-dollar question: you can be bullish on compute growth overall and still uncertain whether individual GPUs hold their value. They depreciate; faster chips replace them. Because no one can say what a B200 or B300 will fetch in four or five years, futures contracts must exist. CME's initial contracts run three years, extending as liquidity grows. Don Wilson noted that last Thanksgiving, H100 rental rates reversed sharply upward when Anthropic released a new model and inference demand exploded; most recently, B200, B300, and H100 prices surged again—while Silicon Data's token index collapsed.
— Don WilsonLeverage and oversupply fears miss how futures prices form: buyers and sellers must both agree
Tracy posed a sharp objection: if GPU values fall over time, won't hedging tools just let everyone borrow more money and overbuild into the same underlying risk, rather than actually reduce it? Don Wilson's answer: futures require both a willing buyer and seller. If anyone tries to sell at an artificial price ceiling, no rational counterparty appears to buy. His prediction: hedging demand will concentrate on the selling side, pushing distant contract prices downward. It takes return-seeking investors willing to buy futures contracts to provide the other side.
— Don WilsonHeterogeneous hardware doesn't stop futures: H100 alone pools enough capital for a market
Critics argue chip diversity—different software stacks, AMD, Trainium, TPU—prevents building a tradable index. Don Wilson said that H100 or B200 alone has enough capital to support a standalone futures market. Other chips correlate tightly with the standard config; markets will price out the basis spread. He disclosed that DRW is already exploring futures on token output directly, framing the profits from buying spot GPU rental, running open-source models, and selling tokens as analogous to oil crack spreads or soy crush spreads—what they call internally the "crunch spread."
— Don WilsonNvidia's dominance will erode, but the company will still grow—market share shrinks, market expands
When asked how quickly Nvidia's grip could loosen, Don Wilson's take: the compute market is expanding, so Nvidia probably won't hold current market share. Yet he still expects the company to grow absolutely larger. Shrinking share and expanding revenue aren't contradictory when the total market expands.
— Don WilsonPerpetuals are futures, not swaps—and they're how derivatives actually make markets fairer
Terry objected that perpetual contracts are really swaps masquerading as futures. Don Wilson disagreed: perpetuals are futures and work better. An oil ETF tracking a Brent perpetual directly would eliminate rollover slippage versus today's monthly re-ups. Retail holding hard-to-borrow stock can't access the high borrow fees that banks get—perpetuals route that cash mechanically to them, a genuine case where derivatives broaden access. He recalled asking Dodd-Frank drafters what separates futures from swaps and hearing "everyone knows swaps are riskier"—pure semantics, he said. A cleared swap and a futures contract impose identical economic risk.
— Don WilsonIn their own words · checked verbatim
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.
Don Wilson1:03
Now, conveniently, November 9th is after November 3rd.
Don Wilson4:04
Because nobody knows what the value of a B200 will be four years from now, five years from now.
Don Wilson7:07
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?
Don Wilson9:21
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.
Don Wilson14:25
it's unlikely that NVIDIA will be able to maintain the market share that they have maintained.
Don Wilson16:27
A swap that is cleared has the same economic risk to society, to the holder, as a futures contract does.
Don Wilson22:45
Figures
| CoreWeave bond yield | 15% | 2:03 |
| CME GPU futures initial contract | 3 years | 7:07 |
| IOSCO index best practices | 2013 | 6:07 |
Glossary
- crack spread / crush spread / crunch spread
- The profit from processing raw materials into finished goods, measured by price spread; DRW uses 'crunch spread' for GPU-rental-to-token-sales arbitrage.
- perps / perpetual contracts
- Derivatives with no expiration date that mechanically replicate futures leverage.
- neocloud
- Emerging GPU-rental cloud providers, distinct from hyperscale cloud giants.
- IOSCO
- International Organization of Securities Commissions—the global standard-setter for index construction best practices.
- basis trade
- Arbitrage or hedging using the price gap between related but imperfectly correlated assets.
How to listen
Compute traders, investors watching credit risk at neocloud providers like CoreWeave, and tech workers wanting to understand GPU futures regulation.
From 17:41, a discussion of DRW's internal token routing optimization—low information density, skippable.