CME Group and Silicon Data Launch First-Ever Compute Futures Market for AI

Silicon Data CEO Carmen Li is building what she hopes will become the world's first regulated futures market for GPU compute capacity. In a recent Bloomberg Odd Lots podcast interview, Li described the project as an effort to bring price discovery and risk management to a market that currently operates through opaque, long-term contracts between cloud providers and their customers. The vision, shared by DRW founder Don Wilson, is to create a financial market bigger than oil. Compute has become the essential input for modern AI development, yet its pricing remains largely inaccessible to anyone outside the largest technology companies.

The partnership with CME Group, announced June 15, 2026, represents the first concrete step toward standardized compute futures contracts. CME brings decades of experience in commodity and financial futures, while Silicon Data contributes proprietary data on GPU pricing and utilization across cloud providers. The collaboration aims to settle a long-standing gap between the explosive growth in AI infrastructure investment and the financial tools available to hedge that exposure.

The core innovation is a GPU compute index that aggregates real-time pricing from major cloud providers and on-premise data centers. This index will serve as the reference price for cash-settled futures contracts, allowing market participants to take positions on the future cost of compute without ever touching a physical server. According to Bloomberg's coverage of the Odd Lots episode, Li emphasized that constructing this index required solving novel data challenges, including normalizing across different chip architectures and contract structures.

Futures contracts will initially focus on NVIDIA GPU hours, the dominant currency of AI training and inference. The contracts will likely trade in monthly expirations with cash settlement against the Silicon Data index. This structure mirrors how natural gas and electricity futures function, where physical delivery is impractical but price risk is substantial. Market makers and proprietary trading firms have already expressed interest, drawn by the volatility and correlation to AI sector sentiment.

Silicon Data's compute exchange ambitions are backed by significant venture funding. Multiple sources report that Ornn, a related entity in the compute futures ecosystem, raised $5.7 million in seed capital to build exchange infrastructure. The Times Magazine and Pulse2 identified this funding as specifically earmarked for launching what they termed the world's first GPU compute futures exchange. The relationship between Ornn and the CME-Silicon Data partnership remains partially opaque, but the capital commitment signals serious investor conviction in the underlying thesis.

The funding environment for financial infrastructure startups has tightened considerably since 2021, making this raise notable in itself. Investors are evidently persuaded that compute price volatility creates genuine hedging demand, not merely speculative interest. The seed round size suggests a capital-efficient approach to building matching engine and clearing technology, rather than the heavy spend typical of consumer-facing platforms.

Cloud hyperscalers face the most complex strategic calculus. Standardized pricing and futures markets threaten the opaque discounting that currently lets AWS, Google Cloud, and Microsoft Azure extract maximum surplus from price-insensitive AI startups. At the same time, these same providers could use futures to hedge their own massive NVIDIA purchase commitments, turning a volatile cost center into a more predictable input. Bloomberg's coverage notes that DRW's involvement signals serious institutional interest in making this market liquid.

For AI startups and enterprise buyers, the benefits are more straightforward. Locked into 12-month cloud contracts, they currently bear full price risk if GPU spot prices collapse. Futures would allow them to lock in costs or at minimum understand fair market value. NVIDIA itself faces mixed implications: greater price transparency could accelerate competition from AMD and custom silicon, but a deeper, more liquid market for compute ultimately validates the centrality of GPU capacity in the global economy.

No commodity futures market has ever attempted to settle against a purely digital, rapidly depreciating asset with such complex quality differentiation. The Commodity Futures Trading Commission will need to satisfy itself that the Silicon Data index is manipulation-resistant and that the reference market is sufficiently deep to support derivative pricing. CME's involvement provides regulatory credibility, but approval timelines remain uncertain.

Technical challenges extend to the index construction itself. GPU performance varies enormously across NVIDIA's product generations, and the emergence of inference-optimized chips from multiple vendors complicates any simple hours-based metric. Li's team must balance comprehensiveness against simplicity, a tension familiar to creators of the VIX or other benchmark indices. The market's success will ultimately depend on whether it attracts sufficient liquidity from both commercial hedgers and financial speculators to produce reliable price signals.