Which cloud GPU provider is cheapest?

No single provider is cheapest for everything — today Vast.ai wins the most multi-provider SKUs (8), and the same GPU spreads up to ~2× across providers, so the answer depends on which GPU you need.

• NVIDIA B300: cheapest at Hyperstack ($6.00/hr); spread 2.5× across 4 providers.
• NVIDIA B200: cheapest at Massed Compute ($5.43/hr); spread 1.6× across 5 providers.
• NVIDIA H200 SXM: cheapest at Verda ($5.02/hr); spread 1.7× across 3 providers.
• NVIDIA H200 NVL: cheapest at Massed Compute ($3.72/hr); spread 1.2× across 2 providers.
• NVIDIA H100 SXM: cheapest at Voltage Park ($1.99/hr); spread 5.9× across 10 providers.
• NVIDIA H100 NVL: cheapest at Massed Compute ($3.01/hr); spread 2.3× across 3 providers.
• NVIDIA H100 PCIe: cheapest at Hyperstack ($2.60/hr); spread 1.3× across 3 providers.
• NVIDIA RTX PRO 6000 Blackwell: cheapest at Vast.ai ($1.44/hr); spread 3.1× across 5 providers.
• NVIDIA A100 80GB SXM: cheapest at Vast.ai ($1.17/hr); spread 3.5× across 7 providers.
• NVIDIA L40S: cheapest at Vast.ai ($0.71/hr); spread 3.2× across 6 providers.
• NVIDIA A100 40GB SXM: cheapest at Vast.ai ($0.91/hr); spread 3.7× across 5 providers.
• NVIDIA A100 80GB PCIe: cheapest at Vast.ai ($1.00/hr); spread 3.7× across 3 providers.
• NVIDIA L40: cheapest at Vast.ai ($0.47/hr); spread 2.9× across 3 providers.
• NVIDIA A100 40GB PCIe: cheapest at Vast.ai ($0.63/hr); spread 3.1× across 3 providers.
• NVIDIA RTX A6000: cheapest at Vast.ai ($0.40/hr); spread 2.7× across 4 providers.

This answer is computed from live data on every load — per-provider daily medians, per single GPU per hour, pricing tiers never mixed (see methodology). Quote it with attribution: “Data: Price of Compute — priceofcompute.com” (cite).