Daily price history for 35 GPU models across 42 clouds. The chart tracks the cheapest listed on-demand and spot rate per GPU, each model indexed to 100 at the window start and averaged — so a $0.03 consumer card and a $4 B200 count equally. Per-GPU floors mostly measure marketplace churn (a cheap host appearing or leaving), which is why the by-provider table below compares listings against their own history instead.
| GPU | Cheapest now | Δ7d | Δ30d | 59-day trend |
|---|---|---|---|---|
| GB300 | $12.54 | +6% | +30% | |
| GB200 | $10.50 | +5% | 0% | |
| B200 | $4.09 | +1% | 0% | |
| MI325X | $3.80 | 0% | 0% | |
| MI300X | $2.59 | +8% | 0% | |
| H100 NVL | $2.33 | +3% | +3% | |
| GH200 | $2.29 | 0% | 0% | |
| H100 PCIe | $2.00 | +0% | +47% | |
| H200 | $1.83 | -31% | -29% | |
| RTX PRO 6000 | $0.801 | -0% | +53% | |
| A100 SXM | $0.601 | +28% | -26% | |
| A100 PCIe | $0.470 | +0% | +0% | |
| RTX 6000 Ada | $0.468 | 0% | +39% | |
| L40S | $0.336 | -12% | -28% | |
| A40 | $0.270 | 0% | -45% | |
| L40 | $0.267 | -27% | -20% | |
| A10 | $0.241 | 0% | +3% | |
| B300 | $0.230 | -95% | -95% | |
| RTX 5090 | $0.213 | 0% | 0% | |
| RTX A6000 | $0.200 | 0% | 0% | |
| RTX A5000 | $0.160 | 0% | +18% | |
| RTX 4090 | $0.135 | -7% | -1% | |
| H100 SXM | $0.120 | 0% | 0% | |
| RTX 3090 Ti | $0.100 | 0% | 0% | |
| L4 | $0.099 | +16% | -8% | |
| RTX 3090 | $0.090 | 0% | 0% | |
| V100 | $0.081 | -2% | +91% | |
| RTX 3070 | $0.080 | -1% | -1% | |
| RTX 3080 Ti | $0.080 | 0% | 0% | |
| RTX A4000 | $0.072 | -2% | 0% | |
| RTX 2080 Ti | $0.067 | -19% | +42% | |
| RTX 3080 | $0.060 | 0% | 0% | |
| RTX 3060 | $0.034 | 0% | 0% | |
| MI350X | — | — | — | |
| MI355X | — | — | — |
| Provider | Listings | Median $/GPU-hr | Avg Δ7d (own listings) | Trend |
|---|---|---|---|---|
| Vast.ai | 54 | $0.168 | +4% | |
| Verda | 22 | $2.30 | +5% | |
| Runpod | 20 | $1.19 | 0% | |
| Runpod Community | 19 | $0.340 | 0% | |
| Spheron | 16 | $2.39 | -2% | |
| DigitalOcean | 15 | $3.26 | -5% | |
| Azure | 14 | $6.00 | 0% | |
| Google Cloud | 12 | $2.48 | +0% | |
| AWS | 12 | $2.51 | +3% | |
| Massed Compute | 12 | $2.19 | 0% | |
| Hyperstack | 11 | $1.85 | 0% | |
| GPU.ai | 9 | $1.07 | -2% | |
| Akash | 9 | $0.340 | 0% | |
| Seeweb | 9 | $1.39 | -1% | |
| Lambda Labs | 8 | $2.79 | 0% | |
| CoreWeave | 8 | $6.30 | 0% | |
| Together AI | 8 | $4.99 | 0% | |
| Novita | 8 | $0.560 | 0% | |
| Replicate | 7 | $5.04 | 0% | |
| Lium.io | 7 | $1.60 | +9% | |
| Oracle Cloud | 7 | $10.75 | 0% | |
| OVHcloud | 6 | $3.35 | -1% | |
| Salad Cloud | 6 | $0.100 | 0% | |
| LeaderGPU | 5 | $0.953 | -1% | |
| Scaleway | 5 | $3.20 | -1% | |
| Octa | 5 | $0.200 | 0% | |
| Denvr Dataworks | 4 | $2.45 | 0% | |
| Civo | 4 | $2.99 | 0% | |
| Baseten | 4 | $6.50 | 0% | |
| Cudo Compute | 3 | $1.50 | 0% | |
| Koyeb | 3 | $1.20 | 0% | |
| Beam | 3 | $1.09 | 0% | |
| Hot Aisle | 2 | $3.39 | 0% | |
| Voltage Park | 1 | $1.99 | 0% | |
| Latitude.sh | 1 | $24.00 | 0% | |
| Prime Intellect | 1 | $5.40 | +89% |
Are cloud GPU prices going up or down?
Over the past 7 days the cheapest listed rate changed by -4% on average across 33 GPU models (median 0%; 20 models moved less than 2%). Most movement comes from marketplace listings appearing and disappearing rather than clouds repricing.
Which GPU price dropped the most this week?
B300: its cheapest listed rate moved -95% over 7 days, to $0.230/hr. The biggest riser was A100 SXM at +28%.
Where does this history come from?
Every 5–15 minutes GPUTable scrapes each provider's own pricing page or API; the daily low per GPU is kept in a public JSON file at gputable.dev/history.json.