Directory · Service type
GPU Cloud GPU providers compared
GPU cloud providers rent virtualized GPU instances by the hour, usually with an API, per-second or per-hour billing, and no hardware to manage. Capacity can often start in minutes; in exchange you share the underlying platform and accept the provider's instance shapes.
The FlopSource directory tracks 18 gpu cloud providers. They are compared below on the GPUs they offer, headline pricing, availability, and location. FlopSource's verified data classifies providers by service model (GPU cloud, bare-metal, edge); it does not publish separate on-demand, spot, or reserved rates per provider, so this page compares providers within the gpu cloud model rather than inventing tier rates the data does not carry. Per-GPU prices for each provider and GPU are on the provider pages.
| Provider | GPUs offered | Headline price | Availability | Primary location | Price verified |
|---|---|---|---|---|---|
| Vast.ai | H100, H200, A100, L40S, MI300X | from $1.53 / GPU-hour | Live marketplace | Global (decentralized - 30+ countries) | 2026-06-08 |
| Hyperbolic | H100, A100, L40S | from $1.75 / GPU-hour | Immediate | US + EU | 2026-11-03 |
| Cudo Compute | H100, A100, L40S | from $1.85 / GPU-hour | Immediate | Global (strong EU presence) | 2026-06-08 |
| TensorDock | H100, A100, L40S | from $1.95 / GPU-hour | Immediate | US + EU + Asia | 2026-06-08 |
| Voltage Park | H100, H200 | from $1.99 / GPU-hour | Immediate (large dedicated blocks) | US (power-advantaged locations) | 2026-11-03 |
| Massed Compute | H100, H200, A100 | from $2.05 / GPU-hour | Immediate | US (multiple regions) | 2026-11-03 |
| Shadeform | A100, H100, H200 | from $2.10 / GPU-hour | Immediate (via orchestration) | Multi-region (orchestration layer) | 2026-06-08 |
| Crusoe Energy | H100 | from $2.25 / GPU-hour | Immediate | Texas & Colorado, USA | 2026-11-03 |
| Lambda Labs | H100, MI300X | from $2.39 / GPU-hour | Contact for availability | California, USA | 2026-06-08 |
| Nebius | H100, H200 | from $2.49 / GPU-hour | Immediate | Finland & Netherlands | 2026-11-03 |
| FluidStack | H100, H200 | from $2.55 / GPU-hour | Immediate | UK + EU + US | 2026-11-03 |
| CoreWeave | H100, H200 | from $2.65 / GPU-hour | Immediate | New Jersey & Virginia, USA | 2026-11-03 |
| Oracle Cloud Infrastructure | H100, A100 | from $2.85 / GPU-hour | Immediate (bare metal shapes) | US + EU + APAC | 2026-11-03 |
| RunPod | H100, A100, MI300X | from $3.19 / GPU-hour | Immediate | Global (US + EU + APAC) | 2026-11-03 |
| Google Cloud | H100, H200, A100 | from $3.95 / GPU-hour | Immediate (on-demand); committed use discounts common | Global (us-central, europe-west, asia) | 2026-11-03 |
| Together AI | H100, H200, A100 | from $3.99 / GPU-hour | Immediate | US + EU | 2026-11-03 |
| Amazon Web Services | H100, H200, Trainium, Inferentia | from $4.10 / GPU-hour | Immediate (on-demand); faster with reservations | Global (US, EU, APAC regions) | 2026-11-03 |
| Microsoft Azure | H100 | from $12.29 / GPU-hour | Immediate (on-demand); capacity reservations recommended | Global (East US, West Europe, etc.) | 2026-11-03 |
How to read the prices: this tier table uses each provider's headline rate (its entry-level per-GPU-hour price) with the date it was last verified. Verified per-GPU prices, with sources, are on each provider's detail pages. Confirm the live quote for a specific GPU with the provider.
Other service types
Bare-Metal providers · Edge providers
Compare by GPU
H100 · A100 · H200 · L40S · MI300X · RTX 4090 · RTX 3090 · Trainium · Inferentia
Compare providers for your workload
Live pricing, hardware specs, and availability across 29 GPU clouds, neoclouds, and bare-metal providers.
Open the Directory Tool