Compute economics · observed rates vs peak silicon
GPU value, carefully measured.
Public hourly rental quotes beside vendor-published dense FP16/BF16 throughput. A narrow theoretical lens on hardware pricing—not an application benchmark.
Observed listings
647
Feed GPU SKUs
35
Cloud providers
42
Comparable model families
5
All observed quotes
GPU rental market
647 of 647 listings · snapshot Oct 2, 2026, 6:20 AM UTC
Data feed: GPUTable
| Accelerator / feed SKU | Provider / source | USD / GPU-hour | GPU count | Dense FP16 / BF16 peak | Theoretical $ / peak TFLOP·h | Quote type / availability | Basis |
|---|---|---|---|---|---|---|---|
| A100 Tensor Core A100 PCIe · Ampere | Vast.ai 80 GB · Spot | $0.1333 | 1× | $0.000427 | Spot Available | Single-accelerator FP16/BF16 vendor peak; no sparsity multiplier applied. | |
| A100 Tensor Core A100 PCIe · Ampere | Vast.ai 80 GB · Spot | $0.1333 | 2× | $0.000427 | Spot Available | Single-accelerator FP16/BF16 vendor peak; no sparsity multiplier applied. | |
| L40S L40S · Ada Lovelace | Vast.ai 48 GB · Spot | $0.20 | 4× | $0.000552 | Spot Available | Single-accelerator FP16/BF16 vendor peak; no sparsity multiplier applied. | |
| L40S L40S · Ada Lovelace | Vast.ai 48 GB · Spot | $0.20 | 8× | $0.000552 | Spot Available | Single-accelerator FP16/BF16 vendor peak; no sparsity multiplier applied. | |
| L40S L40S · Ada Lovelace | Vast.ai 48 GB · Spot | $0.2667 | 1× | $0.000737 | Spot Available | Single-accelerator FP16/BF16 vendor peak; no sparsity multiplier applied. | |
| L40S L40S · Ada Lovelace | Vast.ai 48 GB · Spot | $0.2667 | 2× | $0.000737 | Spot Available | Single-accelerator FP16/BF16 vendor peak; no sparsity multiplier applied. | |
| L40S L40S · Ada Lovelace | Vast.ai 48 GB · Spot | $0.2667 | 7× | $0.000737 | Spot Available | Single-accelerator FP16/BF16 vendor peak; no sparsity multiplier applied. | |
| Instinct MI300X MI300X · CDNA 3 | Azure 192 GB · Spot | $1.1088 | 8× | $0.000853 | Spot Not reported | Single-accelerator FP16/BF16 vendor peak; no sparsity multiplier applied. | |
| A100 Tensor Core A100 SXM · Ampere | Vast.ai 80 GB · Spot | $0.2667 | 1× | $0.000855 | Spot Available | Single-accelerator FP16/BF16 vendor peak; no sparsity multiplier applied. | |
| L40S L40S · Ada Lovelace | AWS 48 GB · Spot | $0.3767 | 4× | $0.001041 | Spot Not reported | Single-accelerator FP16/BF16 vendor peak; no sparsity multiplier applied. | |
| Instinct MI355X MI355X · CDNA 4 | DigitalOcean 288 GB · Spot | $2.97 | 8× | FP16 / BF16 matrix: 2.5 PFLOPS Dense · per accelerator | $0.001188 | Spot Not reported | Single-accelerator FP16/BF16 vendor peak; no sparsity multiplier applied. |
| A100 Tensor Core A100 SXM · Ampere | Google Cloud 80 GB · Spot | $0.3893 | 1× | $0.001248 | Spot Not reported | Single-accelerator FP16/BF16 vendor peak; no sparsity multiplier applied. | |
| A100 Tensor Core A100 SXM · Ampere | Vast.ai 80 GB · Spot | $0.40 | 2× | $0.001282 | Spot Available | Single-accelerator FP16/BF16 vendor peak; no sparsity multiplier applied. | |
| L40S L40S · Ada Lovelace | Vast.ai 48 GB · On demand | $0.4669 | 8× | $0.00129 | On demand Available | Single-accelerator FP16/BF16 vendor peak; no sparsity multiplier applied. | |
| L40S L40S · Ada Lovelace | Vast.ai 48 GB · On demand | $0.467 | 4× | $0.00129 | On demand Available | Single-accelerator FP16/BF16 vendor peak; no sparsity multiplier applied. | |
| Instinct MI300X MI300X · CDNA 3 | DigitalOcean 192 GB · Reserved | $1.91 | 8× | $0.001469 | Reserved Not reported | Single-accelerator FP16/BF16 vendor peak; no sparsity multiplier applied. | |
| L40S L40S · Ada Lovelace | Vast.ai 48 GB · On demand | $0.5344 | 2× | $0.001476 | On demand Available | Single-accelerator FP16/BF16 vendor peak; no sparsity multiplier applied. | |
| L40S L40S · Ada Lovelace | Vast.ai 48 GB · On demand | $0.5356 | 1× | $0.00148 | On demand Available | Single-accelerator FP16/BF16 vendor peak; no sparsity multiplier applied. | |
| L40S L40S · Ada Lovelace | Novita 48 GB · On demand | $0.55 | 1× | $0.001519 | On demand Not reported | Single-accelerator FP16/BF16 vendor peak; no sparsity multiplier applied. | |
| L40S L40S · Ada Lovelace | LeaderGPU 48 GB · Reserved | $0.6361 | 8× | $0.001757 | Reserved Not reported | Single-accelerator FP16/BF16 vendor peak; no sparsity multiplier applied. | |
| L40S L40S · Ada Lovelace | AWS 48 GB · Spot | $0.6919 | 8× | $0.001911 | Spot Not reported | Single-accelerator FP16/BF16 vendor peak; no sparsity multiplier applied. | |
| Instinct MI300X MI300X · CDNA 3 | DigitalOcean 192 GB · On demand | $2.59 | 8× | $0.001992 | On demand Not reported | Single-accelerator FP16/BF16 vendor peak; no sparsity multiplier applied. | |
| Instinct MI300X MI300X · CDNA 3 | Seeweb 192 GB · On demand | $2.655 | 1× | $0.002042 | On demand Not reported | Single-accelerator FP16/BF16 vendor peak; no sparsity multiplier applied. | |
| L40S L40S · Ada Lovelace | Verda 48 GB · Spot | $0.779 | 1× | $0.002152 | Spot Not reported | Single-accelerator FP16/BF16 vendor peak; no sparsity multiplier applied. | |
| L40S L40S · Ada Lovelace | Verda 48 GB · Spot | $0.779 | 2× | $0.002152 | Spot Not reported | Single-accelerator FP16/BF16 vendor peak; no sparsity multiplier applied. | |
| L40S L40S · Ada Lovelace | Verda 48 GB · Spot | $0.779 | 4× | $0.002152 | Spot Not reported | Single-accelerator FP16/BF16 vendor peak; no sparsity multiplier applied. | |
| L40S L40S · Ada Lovelace | Verda 48 GB · Spot | $0.779 | 8× | $0.002152 | Spot Not reported | Single-accelerator FP16/BF16 vendor peak; no sparsity multiplier applied. | |
| A100 Tensor Core A100 SXM · Ampere | Vast.ai 80 GB · On demand | $0.6741 | 1× | $0.002161 | On demand Available | Single-accelerator FP16/BF16 vendor peak; no sparsity multiplier applied. | |
| A100 Tensor Core A100 PCIe · Ampere | Azure 80 GB · Spot | $0.6788 | 2× | $0.002176 | Spot Not reported | Single-accelerator FP16/BF16 vendor peak; no sparsity multiplier applied. | |
| A100 Tensor Core A100 PCIe · Ampere | Azure 80 GB · Spot | $0.6788 | 1× | $0.002176 | Spot Not reported | Single-accelerator FP16/BF16 vendor peak; no sparsity multiplier applied. | |
| A100 Tensor Core A100 PCIe · Ampere | Azure 80 GB · Spot | $0.6788 | 4× | $0.002176 | Spot Not reported | Single-accelerator FP16/BF16 vendor peak; no sparsity multiplier applied. | |
| L40S L40S · Ada Lovelace | Runpod Community 48 GB · On demand | $0.79 | 1× | $0.002182 | On demand Available | Single-accelerator FP16/BF16 vendor peak; no sparsity multiplier applied. | |
| L40S L40S · Ada Lovelace | Runpod Community 48 GB · Spot | $0.79 | 1× | $0.002182 | Spot Available | Single-accelerator FP16/BF16 vendor peak; no sparsity multiplier applied. | |
| L40S L40S · Ada Lovelace | AWS 48 GB · Spot | $0.7925 | 1× | $0.002189 | Spot Not reported | Single-accelerator FP16/BF16 vendor peak; no sparsity multiplier applied. | |
| Instinct MI325X MI325X · CDNA 3 | DigitalOcean 256 GB · Reserved | $2.88 | 8× | FP16 / BF16 matrix: 1.307 PFLOPS Dense · per accelerator | $0.002204 | Reserved Not reported | Single-accelerator FP16/BF16 vendor peak; no sparsity multiplier applied. |
| Instinct MI300X MI300X · CDNA 3 | Hot Aisle 192 GB · On demand | $2.99 | 1× | $0.0023 | On demand Available | Single-accelerator FP16/BF16 vendor peak; no sparsity multiplier applied. |
1–36 of 647
1 / 18