Skip to content
Agentic MEV Day 2026

Compute Markets

Waylon Jepsen, Sebastien La Duca

Session summary

The presenters from San Francisco Compute Company argue that credible GPU futures markets must be built on top of real physical delivery and liquid spot trading, not purely financial products. They contend that contracts need to reflect actual operational constraints: secure multi-tenancy, the ability for buyers to resell excess capacity, and careful handling of networking so multiple tenants can safely share a cluster. They criticize many “Neo cloud” providers for focusing on single-tenant, long-term cluster deals that cannot support resale, and they say current GPU price indices often draw from a single data center routed through several brokers, which GPU sellers regard as inaccurate.

The second half of the talk compares compute to historical commodity markets. The authors note that compute differs from power, crypto, and iron ore because prices depend heavily on contract term, capacity guarantees, contiguity of GPUs, and heterogeneous cluster characteristics. They highlight oil and Brent futures as examples where physically delivered OTC contracts created trusted price discovery before cash-settled indices emerged. They position their own cloud—allowing customers to buy and resell capacity—as an attempt to follow a similar delivery-first path toward eventual institutional-grade GPU markets.

Key takeaways

  • The authors argue that institutional GPU futures require a liquid spot market with real physical delivery and resale of capacity, which in turn demands secure multi-tenant cloud infrastructure.
  • They claim many existing GPU price indices are based on a single data center intermediated by multiple brokers, and that, in their experience, GPU sellers view these indices as inaccurate.
  • They assert that compute’s dependence on contract term, capacity guarantees, GPU contiguity, and non-standardized cluster attributes makes it more like early oil markets, where physically delivered contracts were needed before reliable financial indices emerged.