Tuesday, September 29
Skinners’ Hall, London
The Financialization of Compute
Demand for raw computing power, or “compute,” has kept pace with, and in some cases outstripped, the rapid growth of AI. That has created a new source of volatility, and with it, a new source of investment. The financialization of the chips that power AI’s LLMs means that the graphics processing unit (GPU) is on its way to becoming a mainstream trading opportunity. The hardware itself stays physical, but by hedging the price and time of its rental, firms can turn an unpredictable cost into a manageable one: classic financial risk transfer.
From Metal Boxes to Tradable Hours
The scale of this market is enormous and growing fast. The Goldman Sachs Global Institute estimates that building out AI infrastructure could require around $7.6 trillion in cumulative capital between 2026 and 2031, across compute, data centers and power. A market of that size, with this much volatility, is exactly the kind of environment where the ability to price, hedge and trade exposure becomes essential.
It is not surprising, then, that exchanges are racing to partner with index providers to capture this new market, and experienced market makers are turning what were once opaque, one-off compute deals into standardized financial products.
Historically, getting exposure to the compute market meant participating in a physical supply chain, from physical manufacture, through data center infrastructure, logistics and cloud providers, and finally onto the end users and AI labs. But today, a parallel market is forming on top of the physical one. Participants can trade the invisible commodity that the hardware produces: compute hours.
Launch of GPU Futures
According to Reuters, shares in Chinese memory chipmaker CXMT surged 466% on their Shanghai debut on July 27, 2026, Asia’s biggest IPO this year, against the backdrop of a recent selloff in global tech stocks. It’s hard to find a better example of volatility, and that volatility makes it difficult for producers and investors to place long-term bets on capacity. Not surprising then that exchanges are competing to launch futures contracts to help smooth out the pricing curve for hyperscalers, AI labs and investors.
According to their own announcements, both CME and ICE have unveiled plans to launch GPU compute futures this year. CME is partnering with Silicon Data, while ICE has announced two separate contracts, one with Ornn, based on its live spot-price index, and one with NativX, based on its energy-normalized COIL Index. They aren’t alone: Architect Financial Technologies is also entering the space through its AX exchange, with perpetual futures that track daily rental prices for GPUs and DRAM memory, also referencing Ornn’s index.
Prediction market platform Kalshi has also launched its own forward curves for Nvidia B200, H200 and A100 chips. We can expect further announcements over the coming weeks and months as the exchanges finalize their product specifications, regulatory approvals and launch plans.
The Link Between the GPU Market and the Power Markets
The obvious link between compute and energy provides an intriguing cross-product trading opportunity. A GPU doesn’t just cost money to buy or rent; it costs a great deal of electricity to run, and at the scale of a modern data center, that power bill becomes one of its biggest costs. As AI workloads grow, the demand for GPUs and the demand for the power to run them are rising in tandem. Compute is, in effect, becoming a way of expressing a bet on energy, and energy a way of expressing a bet on compute.
That relationship is what makes a “compute spread” possible. Energy markets have traded this kind of relationship for decades through the spark spread, the margin between the price of electricity and the cost of the fuel burned to generate it. A power producer doesn’t only care about the price of gas or the price of electricity in isolation, but about the gap between them, and can hedge that gap directly. A compute spread applies the same logic one layer up: instead of hedging chip costs and power costs separately, a trader can hedge the relationship between them, taking a position on the margin between the cost of building and teaching AI and the cost of the electricity that makes it possible.
For the firms building and financing AI infrastructure, this is a genuinely new tool. It lets them manage the single largest uncertainty in their business, the joint cost of compute and power, as one position rather than two separate ones. And, it places whoever operates at this intersection at the nexus of one of the defining trades of the AI era. This is precisely the vantage point Trading Technologies describes when it frames AI as an accelerant rather than a threat: a platform already connecting energy and financial markets is the natural home for a trade that spans both.
Where TT Sits
Every commodity market that matures follows a similar path. Bilateral, OTC trades give way to standardized contracts; those contracts need somewhere to trade, someone to make markets and a layer of technology to connect it all together. Compute is now walking that path, and the infrastructure layer is where an abstract market becomes a real one.
This is the role trading platforms play. As institutional investors look for exposure to the AI boom without overpaying for tech equities, compute contracts offer direct exposure rather than a proxy, but only if traders can access and risk-manage them through the systems they already operate.
This is exactly where TT sits: our multi-asset platform can absorb a new commodity like compute without forcing clients to rebuild their workflows, which is what lets a market like this reach traders quickly.
Looking Ahead
The financialization of compute is still in its early stages. Product specifications and launch timelines are still taking shape, and the regulatory picture continues to evolve. But the direction of travel is clear. As compute becomes something firms can price and hedge like any other commodity, the tools to trade it will move from novelty to necessity, and the market will increasingly treat a GPU-hour the way it already treats a barrel of oil or a megawatt of power. The firms that understand that shift early will be the ones ready to act on it.
Get Involved in the GPU Debate
TT has always prided itself on being at the forefront of capital market innovation, and our engagement with the emerging GPU market is another example of this. We’d love to hear your views on where this market goes next. To share your perspective or find out more, get in touch through our contact page.