The vendors below do not publish per-GPU per-hour rates that we can independently verify against a public pricing page. We do not infer rates for quote-only vendors. The only honest answer is a sales quote.
Fly.io offered L40S and A100 attached to Fly Machines as a managed-edge GPU product. The service was wound down through 2025 and is best treated as deprecated; new GPU workloads should pick from the actively-developed providers in this list. Quote only - contact Fly.io sales.
European hyperscaler with both Public Cloud GPU instances (managed) and bare-metal GPU servers (long-running). Sovereignty proposition; rates require a direct quote. Quote only - contact OVHcloud sales.
GPU procurement decisions move millions of dollars. Inferring a quote-only vendor's rate from an industry average or a competitor comparison would produce a number that looks confident and is wrong. The only honest answer is a quote from the vendor's sales team. For published-rate vendors, see the homepage matrix; for these, click through to the contact pages above.
Last verified June 2026.