- Prices from public sources
- Updated daily
- Compare by chip and term
- Buy at or below
These are public prices to rent by the hour. To see what is listed on Silicon Liquid for the next 26 weeks, by GPU and week, open the availability calendar.
These are public prices to rent by the hour. To see what is listed on Silicon Liquid for the next 26 weeks, by GPU and week, open the availability calendar.
Updated daily · last run 2 Nov, 20:08 UTC
An H100 hour costs $3.27 at the median provider today. The cheapest is $1.80 on Hyperai; the hyperscalers' median is $9.90.
43 providers, 4,751 offers from 10 sources. Prices are USD per GPU-hour, on demand unless marked. Reserved prices by term →
Need capacity at a better price? Post a request and get quotes →
| Chip | Cheapest | Median | 7 days | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Blackwell | |||||||||||
| GB300 | $13.18Verda (DataCrunch) | $15.59 | +2.5% | ||||||||
| B300 | $0.23Akash | $9.84 | +0.8% | ||||||||
| GB200 | $8.00GMI Cloud | $16.00 | 0.0% | ||||||||
| B200 | $3.49Prime Intellect | $8.19 | −5.1% | ||||||||
| Hopper | |||||||||||
| H200 | $3.14Prime Intellect | $5.04 | −0.7% | ||||||||
| H200 NVL | $2.67Vast.ai | $3.79 | +1.4% | ||||||||
| H100 | $1.80Hyperai | $3.27 | −0.6% | ||||||||
| H100 NVL | $2.61Vast.ai | $3.06 | −0.5% | ||||||||
| GH200 | $1.99Vultr | $2.75 | −2.3% | ||||||||
| AMD Instinct | |||||||||||
| MI355X | $2.59Vultr | $5.59 | 0.0% | ||||||||
| MI350X | $6.16DigitalOcean | $6.16 | 0.0% | ||||||||
| MI325X | $3.80DigitalOcean | $4.21 | 0.0% | ||||||||
| MI300X | $2.39RunPod | $3.29 | −1.7% | ||||||||
| Ampere | |||||||||||
| A100 80 GB | $0.47Vast.ai | $1.79 | +0.1% | ||||||||
| A100 40 GB | $0.40Vast.ai | $1.99 | −0.5% | ||||||||
| Inference | |||||||||||
| L40S | $0.43Vast.ai | $1.53 | +1.1% | ||||||||
| L40 | $0.36Vast.ai | $0.91 | +2.3% | ||||||||
| A40 | $0.47Vast.ai | $0.61 | +3.1% | ||||||||
| L4 | $0.09Vast.ai | $0.59 | −5.8% | ||||||||
| A10 | $0.23Vast.ai | $1.49 | +0.1% | ||||||||
| A10G | $0.18Akash | $0.18 | 0.0% | ||||||||
| T4 | $0.03Vast.ai | $0.22 | +1.6% | ||||||||
| V100 | $0.07Vast.ai | $0.55 | −0.4% | ||||||||
| Workstation | |||||||||||
| RTX PRO 6000 | $0.80Hyperai | $2.19 | −0.1% | ||||||||
| RTX 6000 Ada | $0.46Vast.ai | $0.84 | −0.5% | ||||||||
| RTX A6000 | $0.27Vast.ai | $0.54 | +0.9% | ||||||||
| Consumer | |||||||||||
| RTX 5090 | $0.35Hyperai | $0.80 | −6.5% | ||||||||
| RTX 4090 | $0.33Vast.ai | $0.48 | −7.6% | ||||||||
| RTX 3090 | $0.11Vast.ai | $0.20 | −7.8% | ||||||||
On-demand median per GPU-hour in each region today, one vote per provider; under each, the move since the day before, like for like. Prices from 56 countries.
| Region | H100 | H200 | B200 |
|---|---|---|---|
| North America | $6.88+1.2% | $7.91+9.4% | $8.37+5.3% |
One provider, one source, seven days apart - a provider joining is never a move.
Providers' own public price lists and live marketplace books, collected every day at 10:00 UTC.
Vultr·Scaleway·Vast.ai·Lium·Akash Console·Azure Retail Prices·OVHcloud catalog·Oracle Cloud price list·Verda (DataCrunch) API·gpurentalprices.com
| Latin America | — | $16.430.0% | — |
| Europe | $3.19−3.8% | $5.19+2.3% | — |
| Middle East | $16.70−5.0% | $10.990.0% | $6.061 seller |
| Africa | $12.56−10.1% | $9.610.0% | — |
| South Asia | — | — | — |
| East Asia | $9.630.0% | $8.65+21.1% | — |
| Southeast Asia | — | $12.400.0% | — |
| Oceania | $17.820.0% | $14.570.0% | — |
| $5.24 → $4.19 |
| −20.0% |
| RTX 6000 AdaLium | $1.04 → $0.84 | −19.2% |
| B200Lium | $7.50 → $6.08 | −18.9% |
| L40Vast.ai | $0.35 → $0.41 | +17.3% |
Not counted as moves - a week-on-week jump past +60% is a source revising its data far more often than a provider repricing:
Median on-demand price over dense BF16 throughput, and over memory.
| Chip | $ / PFLOP-hr | $ / GB-hr |
|---|---|---|
| MI300X | $2.52 | $0.017 |
| GH200 | $2.78 | $0.029 |
| MI325X | $3.22 | $0.016 |
| H100 | $3.30 | $0.041 |
| B200 | $3.64 | $0.045 |
| H100 NVL | $3.66 | $0.033 |
| A40 | $4.09 | $0.013 |
| L40S | $4.24 | $0.032 |
| H200 NVL | $4.54 | $0.027 |
| L4 | $4.88 | $0.025 |
| L40 | $5.00 | $0.019 |
| H200 | $5.10 | $0.036 |
| A100 80 GB | $5.74 | $0.022 |
| A100 40 GB | $6.38 | $0.050 |
| GB200 | $6.40 | $0.086 |
| A10 | $11.88 | $0.062 |
Throughput is the vendor datasheet figure without sparsity; H100 is taken at its SXM figure. Chips without a settled figure (B300, GB300, MI350X, MI355X, workstation and consumer cards) are left out.