first-party data · refreshed hourly
The live floor price of every GPU on our aggregated fleet of 12+ clouds, normalized to dollars per GPU per hour. Real launchable stock, real multi-GPU nodes, zero markup. Cite it, link it, build on it.
937 GPUs launchable right now across 27 models · snapshot 2026-10-01
Sorted cheapest first. Secure = vetted data center capacity; Community = independent marketplace suppliers. Node $/GPU is the per-GPU rate on the largest real multi-GPU node in stock.
| GPU | VRAM | Secure $/hr | Community $/hr | Largest node | Node $/GPU/hr | In stock | Market range |
|---|---|---|---|---|---|---|---|
| RTX A4000 | 16 GB | $0.120 | $0.087 | 4× | $0.087 | 39 | — |
| RTX 3080 | 10 GB | — | $0.110 | 2× | $0.135 | 15 | — |
| V100 | 16 GB | — | $0.125 | 2× | $0.125 | 1 | — |
| RTX 3090 | 24 GB | $0.500 | $0.150 | 8× | $0.268 | 46 | — |
| RTX 4000 Ada | 20 GB | $0.280 | $0.200 | 1× | — | 8 | — |
| RTX 4080 | 16 GB | — | $0.210 | 2× | $0.255 | 5 | — |
| RTX 2000 Ada | 16 GB | $0.240 | — | 1× | — | 8 | — |
| RTX 5080 | 16 GB | — | $0.300 | 2× | $0.300 | 16 | — |
| L4 | 24 GB | $0.490 | $0.330 | 2× | $0.335 | 20 | — |
| RTX 4090 | 24 GB | $0.740 | $0.340 | 8× | $0.470 | 121 | — |
| RTX PRO 4500 | 32 GB | $0.720 | $0.350 | 1× | — | 21 | — |
| RTX A6000 | 48 GB | $0.398 | $0.405 | 4× | $0.398 | 40 | — |
| A100 40GB | 40 GB | — | $0.400 | 8× | $0.642 | 26 | — |
| RTX 5090 | 32 GB | $0.990 | $0.441 | 16× | $0.867 | 104 | — |
| A40 | 48 GB | $0.490 | — | 1× | — | 8 | — |
| RTX 6000 Ada | 48 GB | $0.650 | $0.740 | 4× | $0.650 | 20 | — |
| L40 | 48 GB | $0.780 | — | 1× | — | 5 | — |
| L40S | 48 GB | $0.800 | $0.790 | 4× | $0.800 | 172 | $1.86 – $2.50 |
| A100 80GB | 80 GB | $1.07 | $0.880 | 8× | $1.27 | 93 | $2.70 – $4.10 |
| RTX PRO 6000 | 96 GB | $2.09 | $1.69 | 1× | — | 56 | $2.50 – $5.50 |
| H100 PCIe | 80 GB | $2.48 | — | 4× | $2.48 | 20 | — |
| H100 NVL | 94 GB | — | $2.67 | 2× | $2.71 | 7 | — |
| H200 NVL | 141 GB | $3.29 | $4.45 | 2× | $3.29 | 6 | — |
| H100 SXM | 80 GB | $3.49 | $3.69 | 2× | $3.69 | 39 | $6.16 – $11.06 |
| H200 SXM | 141 GB | $4.59 | $3.59 | 2× | $5.12 | 20 | $6.30 – $10.60 |
| B200 | 192 GB | $6.79 | $9.38 | 4× | $9.38 | 9 | $8.60 – $8.60 |
| B300 | 288 GB | $7.89 | — | 1× | — | 12 | $17.80 – $17.80 |
| RTX 5000 Ada | 32 GB | — | — | 1× | — | Out of stock | — |
| A30 | 24 GB | — | — | 1× | — | Out of stock | — |
| RTX A5000 | 24 GB | — | — | 1× | — | Out of stock | — |
| RTX A4500 | 20 GB | — | — | 1× | — | Out of stock | — |
| T4 | 16 GB | — | — | 1× | — | Out of stock | — |
market data via Compute Prices
Prices are read live from GPU.ai's aggregated fleet of 12+ GPU clouds and normalized to a per-GPU hourly rate: a multi-GPU node's price is divided by its GPU count before it can set a floor. Every figure is the raw provider rate with zero markup: what you see is what launching costs, billed per second. Stock counts are launchable inventory, not theoretical capacity, and VRAM figures are the minimum actually delivered across the fleet. Market range is the span of public on-demand list prices across major clouds for the closest equivalent SKU (never spot or committed-use discounts). The underlying data is served by our public no-auth API (JSON, linked in this page's Dataset markup), free to cite with attribution. Snapshot 2026-10-01; refreshes hourly.
The L40S floor price across the fleet, one point per day. A dot on the zero line is a day we checked and found nothing launchable; a shaded break is a day we did not sample. We never draw one as the other — every GPU has its own page with the same chart.
Daily trend, updated hourly · Full 15-minute history available via API
Each GPU has a dedicated page with full specs, a cross-provider comparison, VRAM fit estimates, and live availability.