SF Compute provides marketplace-style access to large-scale NVIDIA GPU compute, offering hourly and time-block purchases of vetted GPU clusters, a mechanism to resell unused time, and operational surfaces including a web dashboard, CLI, and API.
Published Sep 26, 2026
Press reporting and vendor materials indicate traction with research labs and startups and the ability to manage large GPU capacity, but exact customer counts and utilization metrics are not independently validated in public sources.
| Aspect | SF Compute | Voltage Park | Lambda GPU Cloud |
|---|---|---|---|
| Starting Price | Market-based per-GPU-hour starting rates (headline, market-driven) | Published on-demand starting rates for H100-class instances (fixed pricing) | Published on-demand on a per-instance or per-cluster basis (fixed pricing) |
| Best For | Very large bursts; flexible time blocks; market model and potential resale | Transparent fixed on-demand pricing; enterprise-style positioning; quick self-serve provisioning | Standardized GPU cloud experience; instances and clusters; enterprise procurement options |
| Key Strength | Market/contract flexibility; large-cluster access narrative; lower headline market-rate positioning at times | Clear pricing; dedicated GPU cloud positioning; published pricing tiers | Mature productization; breadth of offerings; clearer published pricing tables |
| Key Weakness | Product constraints (no public IPs, no persistent local storage, InfiniBand not available in VMs today); limited public security/compliance documentation | Potentially higher cost than lowest-market providers; availability constraints can exist for very large bursts | Often higher per-hour rates than lowest-market providers; very large contiguous clusters may require sales engagement |