Two of the biggest names in proprietary trading are betting that GPU capacity can become a tradable financial market

Jump Trading and DRW are putting real traders and real money behind a market that, until recently, mostly lived in conference panels: financial contracts tied to computing power.

The AI boom runs on GPUs. Now two of the best-known proprietary trading firms want to turn the rental of those chips into something you can price, hedge, and trade like a commodity.

Jump’s move came on October 12-13, 2026, when the firm said it had hired the entire founding team of Itô. The startup specialized in trading compute and inference, the process of running a trained AI model to produce answers.

Rather than acquire the company outright, Jump brought the people in-house to set up and run a dedicated compute trading desk.

The desk will start with mid-market bilateral contracts. These are private, two-party agreements rather than standardized products listed on an exchange.

Those deals typically run 3 to 5 years and are usually valued in the tens to low hundreds of millions of dollars .

DRW has taken a wider and earlier route. Instead of a single desk, it has been building the infrastructure the market would need to function.

The firm incubated Compute Exchange, a marketplace for GPU capacity. It also incubated Silicon Data, which produces pricing benchmarks and indices for GPU compute.

DRW is also a buyer. In August 2026 it signed a GPU-as-a-Service agreement with QumulusAI for a dedicated Nvidia Blackwell B300 cluster.

That contract runs one year, with renewal options that could stretch it to four years.

Daily. Free. Join 34,000+ readers across crypto, finance, and policy.

DRW is additionally recruiting quantitative and fundamental researchers focused specifically on compute markets.

Right now, the action sits in bilateral forward and structured contracts. Spot indices, which track current rental prices, form the second layer. The third layer is still pending: exchange-traded futures.

CME Group plans futures contracts based on Silicon Data’s H100 and B200 rental indices. The H100 and B200 are Nvidia GPU models widely used for AI workloads.

Each planned CME contract would represent 730 GPU-hours . That is the equivalent of a single GPU running nonstop for an average month.

As of mid-October 2026, those futures were still awaiting approval from the CFTC, the US derivatives regulator.

CME is not alone. ICE and Nodal are also developing futures or working with indices, with launches likewise dependent on regulatory approval.

The core appeal is risk management. Companies that depend on GPUs face swings in rental prices, and AI developers and cloud providers increasingly treat compute as a critical input.

A hedge lets a buyer lock in costs. A data center operator, meanwhile, could sell future capacity at a known price instead of hoping demand holds up.

DRW’s strategy touches every part of the stack. It incubated the marketplace, incubated the index provider, buys capacity for itself, and is hiring researchers to understand the market.

If CME’s futures launch on Silicon Data’s indices, a DRW-incubated firm would effectively be supplying the reference price for a regulated product.

Jump’s approach is narrower but faster. Hiring a ready-made founding team lets it skip years of learning how these contracts work and go straight to dealmaking.

The most immediate milestone is the CFTC decision on CME’s H100 and B200 futures. Approval would give the market its first major standardized, exchange-listed compute product.

Exchange listing changes who can participate. Bilateral deals in the tens to hundreds of millions of dollars favor large, sophisticated counterparties, while standardized futures can open access to a broader set of traders and hedgers.

GPU generations turn over quickly, so an index built on one chip model can lose relevance as newer hardware arrives. The planned CME contracts reference both the H100 and the newer B200, which hints at that challenge. A 3-to-5-year bilateral contract could easily outlive the chip it was originally priced around.

Index integrity is another pressure point. When a reference price underpins futures contracts, questions about how that price is calculated, and who is connected to it, tend to follow.