San Francisco Compute just put a voltage spike through the AI infra market with a Series A that proves compute is no longer a privilege guarded by cloud monarchs. It is a commodity, and the only people surprised by that are the ones still pretending long term contracts are some sacred ritual. Evan Conrad and Alex Gajewski learned that lesson the hard way back in 2023 when they were trying to train audio models and ran straight into the wall of 1 to 3 year commitments for short term needs. Instead of arguing with the wall, they built a marketplace that turns GPUs into tradable assets, complete with spot pricing, futures, subleasing, and a CLI that feels like you are plugging into a secret GPU exchange under 330 Fell Street. The fact that grad students and major research labs now sit on the same platform tells you the demand is real and the liquidity gap was begging to be fixed.
This week the company announced a $40M Series A at a $300M post money valuation, co-led by DCVC partner Ali Tamaseb and Wing VC founding partner Peter Wagner. That is not quiet money drifting through the ecosystem. That is capital that shows up because the thesis is strong, the execution is clean, and the founders are solving an actual supply chain distortion instead of playing dress up with buzzwords. It also reinforces a simple truth that many founders avoid. Markets reward teams that understand both the physics of their industry and the economics around it. San Francisco Compute is doing both without blinking.
The team now includes CTO Eric Park, COO Ethan W. Anderson, and CRO Eric Menees, a roster that blends low level systems engineering with the operational muscle needed to move real hardware. Their platform manages H100 and H200 clusters with 3.2Tb/s InfiniBand, provisions bare metal Kubernetes, and handles node failures with a repack and refund policy that feels almost rebellious in an industry where customers are usually told to “deal with it.” Subleasing unused capacity has already become a quiet power feature because it turns buyers into sellers and turns stranded supply into revenue. That is how you build a real marketplace instead of a one way street.
The new capital will be used to expand compute supply, deepen the exchange mechanics, and push toward supporting tens of thousands of accelerators at once. The company wants to give smaller labs and early stage AI teams access to supercomputer level performance without forcing them into multi-year commitments that drain agility. They also plan to integrate next gen hardware as soon as it hits the market, which means the exchange grows not just in size but in timing precision. The lesson here is clear. Liquidity changes who gets to compete. Flexibility changes who gets to build. San Francisco Compute is betting on both, and the market just backed that bet with $40M reasons.
Startups Startup Funding Venture Capital Series A AI GPU Marketplace Cloud Cloud Computing Accelerator Technology Innovation Tech Ecosystem Startup Ecosystem
