The Programmable Data Center — Evan Conrad, SF Compute

The compute market requires physical settlement to achieve true liquidity and standardization, addressing the inherent heterogeneity of hardware and the complexities of cluster management. By treating compute as a commodity with a forward curve, market participants can hedge risks and optimize utilization across diverse GPU generations. SF Compute facilitates this by managing the underlying "backplane"—including power, networking, and firmware—without owning the hardware, ensuring the independence necessary for a functional order book. This approach allows for the programmatic subdivision of resources, enabling customers to trade capacity based on specific performance constraints rather than generic units. While NVIDIA and major hyperscalers benefit from this ecosystem, their business models prevent them from acting as neutral market makers, leaving a critical role for independent operators to bridge the gap between physical infrastructure and financialized, tradable compute assets.