SF Compute

We run large-scale, vetted GPU clusters.

San Francisco, California, United States 0 GPUs available

About SF Compute

SF Compute sells H100 and B300 capacity as virtual machines, bare metal and managed Slurm clusters, priced through a dynamic spot market rather than a fixed rate card. Capacity is bought from the command line for a chosen size and duration, and customers can resell idle capacity back with a minimum price floor, which makes it a two-sided market. Unlike a peer marketplace, the company states it runs its own clusters from UEFI upwards rather than reselling third-party hardware. It advertises virtual machines live in around five minutes, refunds and hardware repacking when nodes fail, no vendor lock-in, and dedicated support with emergency escalation for large clusters. Press coverage puts funding at around $40 million.

Why choose SF Compute

Pricing
Dynamic spot market

Rates float with supply and demand rather than a fixed card; capacity is bought by size and duration.

Two-sided
Resell idle capacity

Customers can sell unused time back with a minimum price floor.

Hardware
Own vetted clusters

Run from UEFI upwards, not resold third-party capacity.

H100, B300
GPUs offered
~5 minutes
Spin-up
~$40M reported
Funding

Services

Spot GPU capacityvirtual machinesbare metalmanaged Slurm clusterscapacity resale

Pricing options

Dynamic spot marketcustom clusterssubleasing

Available GPUs

No GPU pricing data available yet. Check back soon.

Ready to deploy on SF Compute?

We run large-scale, vetted GPU clusters.

Get started with SF Compute