SF Compute builds data centers and GPU clusters in San Francisco, offering VMs, bare metal, and managed Slurm/Kubernetes via API, CLI, or direct engagement.

Customers reserve long-term capacity (H100, B300, GB300) and resell idle GPUs through a built-in scheduler at up to 4.5x their reserved rate. Pricing averages $1.89/GPU-hour across contracts from one node-hour to multi-year, multi-thousand-node deals.