1. What are Jump Trading and DRW doing with compute power contracts?
Jump Trading and DRW, two of the largest proprietary trading firms, are moving into compute power contracts, financial agreements tied to computing capacity. On October 12, Jump announced that it had hired the founding team of Itô, a startup that trades access to computing capacity, to run a new compute trading desk. DRW, meanwhile, has helped build the data and marketplace infrastructure that such a market needs.
Until recently, the idea of treating GPU time as a tradable commodity belonged mostly to conference panels. Graphics processors are the hardware behind the current AI boom, and the two firms are betting that renting them can become something that is priced, hedged and traded in the way oil, power or wheat are.
The developments matter for crypto readers because the same market-making and derivatives firms that built liquidity in digital assets are now applying those skills to a new asset class. Jump and DRW have both been active in crypto trading, and their attention to compute suggests where they expect the next large pools of risk to form.
2. How is Jump Trading building its compute desk?
Jump did not acquire Itô as a company. Instead, it hired the startup's entire founding team, which had focused on trading compute and inference. Inference is the process of running an already trained AI model to produce answers, and it consumes a steady stream of processing power every time a product built on a model is used.
The new desk will begin with mid-market contracts. These are bilateral deals, meaning private agreements between two parties rather than standardized products listed on an exchange. They typically run for three to five years, which gives each side a long window of certainty about price and supply.
Hiring an experienced team is a faster route than building expertise from scratch. A trading firm entering an unfamiliar market normally spends years learning how participants behave, how deals are structured and where the risks sit. Bringing in people who have already worked on those problems shortens that learning period considerably.
Bilateral contracts are also a natural starting point because they do not depend on any exchange infrastructure being approved or built. A trading desk can negotiate directly with data center operators and AI developers, take on risk using its own capital, and gather information about how prices behave before standardized products arrive.
3. What has DRW built around the compute market?
DRW has taken a broader approach, concentrating on the plumbing of the market rather than on a single desk. It co-founded Silicon Data, which produces daily benchmarks for the price of renting GPUs on demand, and The Compute Exchange, a marketplace for spot GPU capacity. Jump and DRW also co-led the seed financing for Silicon Data, and the company later raised a $30.5 million Series A.
Those two pieces serve different purposes. A marketplace lets buyers and sellers of capacity find each other and transact at current prices. A benchmark provides a reference number that contracts can be settled against, which is essential for any product that is meant to be hedged or traded by people who never intend to take delivery of hardware.
This is the same sequence that has played out in other commodity markets. Physical trading comes first, then published price assessments, and then derivatives that reference those assessments. DRW's position across the marketplace and the index layer places it near the start of that chain.
DRW is also recruiting quantitative and fundamental researchers who focus on compute markets. That signals a longer-term commitment than a one-off investment, since it indicates the firm expects to need dedicated analysis of supply, demand and pricing in this area.
4. How could compute futures change the market?
The market can be described in three layers. The first is the bilateral forward and structured contracts that are currently the most active. The second is spot indices that track the going rate to rent GPU capacity. The third is exchange-traded futures, which remain pending.
On the third layer, CME Group and Silicon Data announced on May 12 that they plan to launch what they called the first compute futures market later this year, with the contracts based on Silicon Data's indices. The plan is subject to regulatory review, and no confirmed launch date has been announced.
A future is a standardized contract to buy or sell an asset at a set price on a set date, and for compute it would normally be settled in cash against an index rather than through delivery of chips or server time. Standardization makes such products easier for a wide range of traders to use than private bilateral deals.
Bilateral arrangements tend to favor large and sophisticated counterparties, because each deal is negotiated individually and carries credit risk between the two sides. An exchange-listed product with central clearing could draw in more participants, including hedgers and speculators who would never negotiate a multi-year capacity contract.
5. Why do trading firms want exposure to compute prices now?
The main attraction is risk management. Companies that depend on GPUs face volatile rental prices, and AI developers and cloud providers increasingly treat compute as a critical input to their business, much as an airline treats fuel. A sharp rise in rental rates can erode margins quickly if the cost cannot be fixed in advance.
Hedging offers a remedy on both sides of the market. A buyer of capacity can lock in a cost today and plan a budget around it. An operator of a data center can sell future capacity at a known price, which supports the financing of new hardware. A financial contract allows each party to transfer the price risk to someone willing to hold it.
Trading firms are natural holders of that risk. Market makers earn money by quoting prices, managing inventory and balancing exposures across many positions. Compute is a new instrument for those skills, and an early presence gives a firm data and relationships that later entrants will lack.
DRW's involvement spans the stack, from the marketplace and the benchmark to research hiring. If futures eventually launch on Silicon Data's indices, a company that DRW helped create would be supplying the reference price for a regulated product. Jump's strategy is narrower but quicker, because it relies on an experienced team to enter the market directly.
6. What are the risks facing compute derivatives?
A compute market has features that complicate pricing. GPU hardware turns over quickly, with new generations arriving regularly, so an index tied to one chip can lose relevance as customers move to newer parts. A bilateral contract of three to five years could outlast the chip it was originally priced on, which makes the long-dated deals harder to value than the headline term suggests.
Index integrity is another concern. When a reference price underlies futures contracts, scrutiny tends to follow how it is calculated and who is connected to its producer. Regulators and market participants will want to know that the benchmark cannot easily be influenced by the firms that trade against it, particularly when those firms helped fund the benchmark's creator.
Counterparty and liquidity risks matter as well. Private contracts depend on both sides honoring their obligations over several years, and a thin market can make it hard to exit a position at a fair price. These are the problems that standardized, centrally cleared products are designed to reduce, but those products do not yet exist for compute.
There is also the question of how contracts would be specified. Compute can be measured in many ways, including by chip model, by hours of usage, by region and by service quality, and a derivative needs a precise definition so that every participant understands what is being priced. Disagreements over those definitions can undermine confidence in a new product before it has a chance to build liquidity.
None of these issues is unusual for a young commodity market. They do, however, explain why the first participants are well-capitalized trading firms with research teams, rather than a broad base of retail or smaller institutional traders.
7. What should readers watch next?
The timing of these moves is worth noting. Jump's hiring was announced on October 12, and the CME and Silicon Data plan dates from May. Within five months, a derivatives exchange, a benchmark provider and two of the largest proprietary trading firms have all signaled that compute is becoming a financial asset, which is a fast progression for an idea that was once confined to speculation.
The next major milestone is the regulatory review of the CME and Silicon Data futures. Approval would create the first major standardized, exchange-listed compute product in the United States, and would likely draw additional exchanges and index providers into the same field.
Readers should also watch how Jump's new desk develops. The firm has said it will begin with bilateral mid-market deals, and the volume and structure of those trades will show whether demand exists for long-term compute hedging beyond a small group of large buyers.
Finally, the behavior of the price indices will be telling. If rental rates for GPUs prove volatile, interest in hedging will grow. If they stabilize as supply catches up with demand, the commercial case for derivatives may be weaker than the trading firms expect. Either outcome will become visible in the data over the coming quarters.