Requirements
The canonical, versioned requirements SF Compute holds suppliers and facilities to.
sfcompute.comCrawled Nov 2, 2026
The canonical, versioned requirements SF Compute holds suppliers and facilities to.
Vetted GPU clusters with 24/7 support, 1.5TB+ NVMe storage, 1TB+ of RAM, and no ingress or egress fees.
San Francisco Compute designs, builds, and operates clusters under a customer’s brand, with full hardware control and a market for unused capacity.
How Abel, the ledger behind SFC, records who owns every node-hour and settles every trade between credits and compute.
A conversation with Rich Jaycobs about the missing steps between buying compute and trading compute futures.
The page offers isolated RDMA networks for training jobs with scalable, reservable, and resellable compute resources.
Run distributed training across H100 machines with a private InfiniBand network, a few CLI commands, and billing by the minute.
Standard Intelligence uses compute resale to support its goal of larger clusters and higher burst capacity.
Announcements, market notes, and the systems behind on-demand GPUs.
The platform supports autonomous research, scaling from initial GPU experiments to thousands of parallel runs.
A $245 million sale of NVIDIA Blackwell B300 compute to a leading AI company, with the flexibility to resell unused capacity.
DCVC and Wing Venture Capital co-led the round, with Electric Capital and Alt Capital participating.
On-platform customers receive an average 25% realized discount on compute through resale.
The page provides large-scale supercomputers available under contracts that allow subleasing without balance sheet risk.
The company offers GPU clusters, bare metal, managed Slurm, and reserved capacity reselling options.
The San Francisco Compute Company. Buy and resell large-scale, vetted GPU clusters from the terminal.
How SF Compute processes customer personal data.
How SF Compute collects, uses, and shares personal information.
What customers may and may not do with SF Compute services.
SF Compute's uptime commitment and service credits.
SF Compute's terms of service.
Evan Conrad, a Y Combinator founder, addresses the GPU compute shortages plaguing AI labs. His startup develops solutions to optimize GPU usage, easing bottlenecks for AI researchers. The focus centers on making high-performance computing more accessible amid rising demand.