SF Compute is a vertically integrated supercomputing company based in San Francisco. It builds and operates large-scale GPU clusters, offering compute time on advanced NVIDIA hardware - including H100s, H200s, and B300s - for AI training and inference workloads. By managing the entire technology stack from the firmware level, the company eliminates reseller markups, with prices starting at $1.95 per GPU per hour for H100 compute.
The company operates a real-time compute marketplace where contracts are traded, allowing startups and enterprises to access supercomputing resources on flexible terms with no lock-in. Its technical domains span supercomputing, GPU cluster operations, AI infrastructure, and marketplace platform development. SF Compute has raised $40 million in Series A funding and maintains a team of about 45 engineers. The company provides 24/7 support via dedicated Slack channels.
