The San Francisco Compute Company operates as a vertically integrated supercomputing firm, designing, building, and running large-scale GPU clusters purpose-built for AI training and inference workloads. Its infrastructure includes NVIDIA H100, H200, and B300 GPUs available through flexible contracts - customers retain the ability to exit agreements by reselling unused capacity on a real-time marketplace, a model that eliminates the long-term lock-in typical of hyperscaler offerings.
A Series A round of $40 million underpins the company's expansion. The customer base already spans organisations at the frontier of AI research and deployment, including NVIDIA, MIT, Roboflow, Liquid AI, and DatologyAI. H100 compute is priced starting at $1.95 per GPU per hour, positioning the company as a cost-competitive alternative in a market dominated by well-capitalised incumbents.
The engineering team numbers roughly 45, with technical work spanning GPU cluster operations, AI training and inference infrastructure, and the marketplace platform that enables secondary trading of compute contracts. The firm is headquartered in San Francisco and provides 24/7 support through dedicated Slack channels. Its stated ambition is to build the defining infrastructure marketplace for the AI era.