A multi-trillion-dollar market for hedging data center investments is being built. It could bring stability. It could also drive a crash
Earlier this year Oxford University’s St Hugh’s College needed to update the computer lab used by its math department, but it found the price had nearly doubled. While for the college’s administrators that posed an annoying but manageable IT bill, for one of the university’s professors, it was a symbol of a far bigger challenge facing anyone sourcing computing power.
Rama Cont, an Oxford professor of mathematical finance and adviser to central bankers on financial risks, blames the soaring costs on the data centers powering models such as ChatGPT and Claude. “This is affecting everyone. All good universities, all corporations, are going to be affected this year,” Cont says. But while it’s a mostly manageable budget problem for many institutions and businesses, for the companies building and financing the AI economy, it could be an existential risk.
McKinsey estimates that globally, data centers will require nearly $7t in capital by 2030, including $5.2t for AI. The cost of that infrastructure, roughly the price of 22 Apollo space programs in today’s dollars, must be financed at the same time that the world figures out what its output — so-called compute — will eventually be worth.
A handful of startups and exchanges now want to create markets trading in future compute usage, allowing data center operators to set clearer prices, and lenders to better value the assets they are financing. If they work, they could bring the transparency and price discovery that such futures — effectively a tradable promise to sell or buy something in the future for a set price — have given oil, wheat and electricity. If they don’t, they risk creating a new channel for losses to flow through, connecting an already highly leveraged AI sector more closely to the wider financial system.
BlackRock chief executive Larry Fink, whose investment firm is pouring billions into data center construction, predicted this spring that the demand for compute is so large that traders will soon be able to trade bets on it as its own asset class. In fact, those markets could launch within weeks, pending regulator approval. CME Group, one of the world’s largest trading markets for futures is partnering with tech startup Silicon Data on plans to start a market in compute futures this year. New York Stock Exchange owner Intercontinental Exchange is preparing a rival in partnership with Ornn, another tech startup. Others in Shanghai and elsewhere are working on them too.
Brett Harrison hopes to beat them all. The former president of FTX’s US business — he left shortly before the crypto exchange blew up — runs Architect Financial Technologies. It plans futures tied to the price of renting GPUs on the open market. “I do believe this will be a multiple-trillion-dollar market,” Harrison says.
What you are actually buying
Futures are basically a legal agreement between a buyer and a seller which allows the two parties to lock in the price of something. A farmer, for instance, can offer wheat at a set price months before it’s ready, “hedging” against the possibility of a bountiful harvest causing an abundant crop and prices to plummet.
Marketplace operators have often created new futures markets for niche products. One of the most famous agricultural futures markets on CME was for pork bellies. It traded from 1961 until 2011, but was delisted as bacon became a year-round consumer good with more predictable prices, instead of a seasonal food which needed to be frozen in the winter. There are futures markets for all sorts of goods, from orange juice (the subject of 1983 comedy film Trading Places) to sunflower seeds, all doing the job of evening out volatile prices for farmers and buyers of goods.
Today’s compute “farmers” are the hyperscalers such as Alphabet, Amazon, Oracle and Microsoft, and the numerous “neoclouds,” companies which build data centers and rent out AI chips to others. AI futures markets, in theory, will allow neoclouds to protect themselves if the price of computing power crashes in the future. It could also allow AI companies worried about rising costs to protect against the price soaring higher than expected. No chips actually change hands in this purely financial transaction.
To build their data centers, many neoclouds borrow money against the future revenue they can extract from renting servers, which contain cards like Nvidia’s Blackwell chips. They, and their lenders, have to consider the risks to the value of their assets and their revenue. For instance, if a much better chip arrives sooner than expected making the ones they have installed prematurely obsolete, or if a more efficient AI training or inference algorithm makes the same work much cheaper, that will reduce the price the neocloud can charge.
Without the ability to hedge that risk, Harrison says, neoclouds have become “their own broker, their own insurance company, their own hedger, their own financier and their own pricer,” as well as a data center construction company. A forward price which everyone agrees on, and can trade, could lower their financing costs by giving lenders a way to value future revenue.
That hedge requires someone to take the other side, effectively betting that future value of the commodity, in this case compute, will be higher or lower than what it’s priced at now. “There have to be enough people willing to bet against you,” says Frank Partnoy, a Berkeley law professor and former financial product designer for Morgan Stanley who chronicled the collapse of Enron. That means future markets must welcome speculators who have nothing to do with data centers, and who are simply trying to make money. “That seems to be what all of them are talking about. No one’s talking really about speculation, although if you create the market, that will happen, right? You create the market, inevitably people will use it to bet on these prices,” says Partnoy.
The trouble with pricing an AI cloud
Before you get to predicting supply and demand for computing power, there is the problem of defining what “compute” is and how to compare one unit of it to another. An oil future specifies a grade of crude, a place and a delivery date. Compute has no settled equivalent.
Carl Anthony, the co-founder of Symmetric Research, a startup which aims to create a “standard compute unit” so that banks can better understand their investments in data centers, says you need to look at the overall “computing fabric” of data centers to truly understand their value. That fabric includes memory, networking, cooling, location, software and workload, and whether those thousands of chips can communicate efficiently while training a model or carrying out an AI inference task. Paper specifications like FLOPs, a standard unit of computing power, or even standardized stress test benchmarks of GPUs, don’t give you a true picture of their real-world capabilities, he says. Research by Silicon Data, CME Group’s partner, found that the performance of rented Nvidia GPUs used in AI data centers can vary as much as 38% in real-world conditions.
Architect’s compute futures get around this by setting prices against an index created by its partner Compute Desk, which uses private rental transactions from cloud providers. It adjusts the details for configuration and location, and normalizes them to an anchor such as an hourly rate for a GPU rental in one year’s time. It also looks at the offer prices for computing contracts, and produces what Harrison calls a “smooth, averaged index.”
Harrison admits his index will not be a 100% accurate picture of every GPU rental at any particular time, similar to, he says, futures markets for other commodities. That still leaves traders some risk to manage.
The benchmark might show compute prices rising while a particular data center’s own prices fall, so a hedge tied to the index pays out little even as the neocloud’s income drops, a mismatch traders call “basis risk.”
A bigger gap is what the index cannot see, because it’s built on prices from independent neoclouds, not the private deals struck by the biggest AI companies which increasingly build their own data centers or negotiate one-off contracts. Those deals are secret, so the index may faithfully track a small public market while missing the much larger private one that actually drives the industry. Harrison’s answer is that no futures market is perfect, and that being able to hedge some of the industry’s volatility, even if not all, is still valuable.
This all assumes the prices feeding into it are genuine. A more serious potential problem arises if the companies supplying those prices have an incentive to make them look higher than they really are. David Wender, a former trader and Anthony’s partner at Symmetric Research, says a cloud provider could potentially report a high GPU price while secretly discounting storage or another service in the same package to sweeten a deal. The customer’s total bill remains lower, but the published GPU price supports the value of the provider’s capacity and the financing raised against it.
In a thin futures market, a few transactions like this could distort the index and send money towards projects whose economics look better than they really are.
Have we seen this movie before?
Harrison brings up the shadow of the collapse of Enron, the bankrupt energy trader which manipulated futures markets around the turn of the century. Enron tried, and failed, to turn broadband capacity into a tradable commodity. Like compute, bandwidth must be consumed and produced at the same time and cannot be stored and sold later. Its similarly perishable electricity trading business was far more consequential. Enron’s traders offered private, over-the-counter contracts, and rapidly gained market power and influence over physical supply in order to engineer a shortage of electricity to raise prices, and their profits. That contributed to an energy crisis in California where wholesale electricity costs soared 800% and hundreds of thousands of people were subject to rolling blackouts.
Compute futures are unlikely to recreate those circumstances. They will be traded on regulated exchanges, centrally cleared and backed by collateral. Traders whose positions lose value will face margin calls, limiting the losses that can build up between counterparties.
Those safeguards, however, only reduce the danger that one trader cannot pay another. They do not answer a different question of where larger losses ultimately go. Partnoy says a futures market may transfer risk to investors best able to bear it, or to those who understand it least, such as less sophisticated retail investors who commonly lose money on betting and crypto markets.
Kristin Johnson, until last summer a commissioner at the Commodity Futures Trading Commission, the body which regulates Harrison and others’ futures markets, sees another blind spot. “We do not oversee technology firms,” she says, speaking of her former employer. She says the regulator supervises the financial elements of any exchange, and ensures that it has sufficient capital in place to cover swings, but it can’t regulate the companies producing the prices and infrastructure beneath it, or the statements that technology leaders make trying to hype up the products they are selling. “We have zero capacity [and] we lack jurisdictional authority to supervise the firms that are deploying these products,” she says.
Johnson raised the example of insurance giant AIG, whose $60b reported loss in the Great Financial Crisis resulted in contagion which threatened the world’s biggest banks: “The technology space and financial services are both markets where there’s a significant amount of concentration … We have this concentration risk, we have counterparty credit risk, and the confluence, or the impact of those types of risks emerging simultaneously can be pretty impactful, and in some contexts, catastrophic.”
This is not a fringe worry. In its annual report this summer, the Bank for International Settlements, which advises the world’s central banks, warned that disappointing AI returns could trigger a sudden pullback in financing and turn the capital-spending boom into a “protracted investment bust,” with knock-on effects rippling out into the wider financial system.
In that worst-case scenario, if GPU rental prices suddenly collapse, the neoclouds that rent out those chips suddenly earn less, and the loans they were servicing out of that revenue start to go bad. The data centers and GPUs pledged as collateral are worth less too, because their earning power has fallen. Lenders take losses, the hyperscalers’ vast investments look harder to justify and technology shares slide. Futures traders caught on the wrong side face margin calls, forcing them to sell elsewhere and spreading the damage.
Leverage is the reason that traders like futures. A small deposit of money can control a much larger position, magnifying gains when prices move their way. But that same leverage works in reverse. When prices fall, exchanges demand more collateral and traders who cannot pay must sell other assets to raise cash. Academics call it a margin spiral, and recent episodes in nickel and UK pension markets — and, on a smaller scale, at Leopold Aschenbrenner’s AI hedge fund — show how quickly it can cascade. In each of those examples, rapid price movements of assets forced highly leveraged asset holders to sell, causing chaos in markets. In the case of the UK government bond market crisis, it helped bring down a prime minister and is still affecting the country’s economy.
Debt: the elephant in the room
Cont, the Oxford math professor, is also concerned about the debt that’s piling up to support the data center boom. The $800b worth of annual contracts flowing to data centers, funded by VCs, private equity and underwritten by banks is a big bet on the price of compute. “The big risk,” Cont says, “is really related to the huge debt being contracted by these technology companies right now, for building huge data centers to face future demand.”
“It’s based on this extrapolation of the current trend,” he says, “and if the extrapolation turns out to be incorrect, well, somebody will pay the price. This debt is in the hundreds of billions, and somebody is holding it.” That reckoning, he stresses, would not come from the new futures markets. “It comes from the basic underlying structure of the AI economy.” In theory, futures would not create those losses. But they could cause contagion and amplification as margin calls transmit the repricing more quickly through the financial system.
The trigger, Cont says, might be a successor to DeepSeek, the Chinese AI laboratory whose efficient model caused a stock-market sell-off in January 2025 by raising doubts about how many expensive chips the industry would need. A new model that sharply reduces the compute needed for useful AI, or software that unlocks idle chips, could cause a similar reassessment. More recently, Chinese AI startup Moonshot has created waves in the industry, and markets, after it released benchmarks showing its Kimi K3 closing on Anthropic and OpenAI’s leading models.
That’s the risk of a bear market in compute, but the bullish case is equally important. Jevons’ paradox, a favourite of the Silicon Valley engineers who work in the AI space, is a useful analogy. The paradox is named after the 19th-century British economist William Stanley Jevons, who calculated that greater efficiency can increase total consumption, after England’s consumption of coal soared following the invention of more efficient steam engines. The same is arguably happening with AI. Cheaper AI computing “tokens” lead to more AI products, longer AI generated videos and other tasks. Harrison says that for 30 years, demand for computing has only ever risen, even as hardware improved.
Cont is disillusioned with the state of the AI race, which he says has “resulted in a very wasteful approach, both in terms of compute and in terms of energy costs… and this is very unhealthy.” He compares it to the Apollo project, which resulted in huge technological triumphs and a man on the moon, but “as a technology that is going to be used commercially that’s not how industry has developed tools historically.”
Yet Cont still thinks that the end state of compute — after various booms and busts — may eventually come to resemble today’s highly commodified electricity market. There will just be a lot of unnecessary waste in how we get there. The futures markets being wired up now are a bet on the road to that future. That is Harrison’s big bet too. A successful futures market could help make the currently volatile cost of compute more predictable and boring, giving datacenter builders, lenders and AI companies a predictable price around which to better plan their businesses, with Harrison’s exchange and the others taking their cut in fees. The flip side of that is if its benchmark proves unreliable, and leverage grows too quickly, it could speed the transmission of a future fall in GPU prices from data centers to lenders and investors.
It is also a bet on an AI economy whose costs “are not very clear right now,” says Cont. Wall Street will soon sell you a hedge on the price of compute, before anyone can really define it. The Oxford math department that paid double for its RAM this year will apparently find out what those costs are at the same time as everyone else.
For the AI industry, the arrival of a futures market signals that compute is maturing into a commodity serious and lucrative enough to attract Wall Street’s interest, but it also means that the sector’s inevitable price corrections are unlikely to stay contained within Silicon Valley. Whether compute goes the same way as the pork bellies market and lands with a flop, or demand for compute keeps racing ahead of supply and the future market balloons into the multiple trillions that Brett Harrison dreams of, the exchanges will soon be there to take your bets.