Market
The GPU cloud market, from the records.
Prices, configurations and terms for 1,060 public GPU offerings from 44 providers, normalised to US dollars per GPU-hour so they compare like for like. Every number links back to the offerings behind it.
- Offerings
- 1,060 775 with a public price
- Providers
- 44 clouds, neoclouds and marketplaces
- GPU models
- 75 NVIDIA, AMD, Intel and more
- Countries
- 30 64 regions
Prices as recorded by Kovara. Newest check 26 Oct 2026; half of the records were checked within the last 33 days. How prices are collected
GPU prices
On-demand price per GPU-hour, for every GPU with at least three prices from two or more providers. The bar runs from the lowest to the highest price on one scale; the mark is the median across providers, each provider counted once.
| GPU | From | Median | Range | Providers | Lowest at |
|---|---|---|---|---|---|
| H100 SXMNVIDIA · 80 GB | $1.99 | $3.35 | 2041 offers | Voltage Park | |
| H200 SXMNVIDIA · 141 GB | $2.60 | $4.47 | 1324 offers | GMI Cloud | |
| L40S 48GBNVIDIA · 48 GB | $0.97 | $1.50 | 926 offers | Massed Compute | |
| RTX PRO 6000 Blackwell Server EditionNVIDIA · 96 GB | $1.80 | $1.96 | 921 offers | Nebius | |
| A100 80GB SXMNVIDIA · 80 GB | $1.38 | $1.78 | 914 offers | Massed Compute | |
| RTX A6000 48GBNVIDIA · 48 GB | $0.35 | $0.55 | 826 offers | Thunder Compute | |
| B300 SXMNVIDIA · 288 GB | $6.60 | $7.67 | 812 offers | Massed Compute | |
| B200 SXM 180GBNVIDIA · 180 GB | $5.43 | $6.84 | 712 offers | Massed Compute | |
| L40 48GBNVIDIA · 48 GB | $0.79 | $0.86 | 58 offers | Thunder Compute | |
| A100 80GB PCIeNVIDIA · 80 GB | $1.59 | $3.39 | 481 offers | Runpod | |
| GeForce RTX 4090 Founders / referenceNVIDIA · 24 GB | $0.16 | $0.55 | 47 offers | SaladCloud | |
| RTX 6000 Ada 48GBNVIDIA · 48 GB | $0.79 | $0.99 | 47 offers | Massed Compute | |
| H100 PCIe 80GBNVIDIA · 80 GB | $2.49 | $3.04 | 44 offers | Civo | |
| H100 NVL 94GBNVIDIA · 94 GB | $2.92 | $3.19 | 360 offers | Massed Compute |
Bigger nodes, per-GPU price
H100 SXM, on-demand price per GPU-hour by GPUs per instance: the median across providers, each provider counted once.
What spot saves
Median spot or preemptible price against the on-demand median, same GPU. Spot capacity can be taken back at short notice.
- H100 SXM$2.18
$3.35−35% - H200 SXM$2.55
$4.47−43% - L40S 48GB$0.83
$1.50−45% - RTX PRO 6000 Blackwell Server Edition$1.20
$1.96−39% - A100 80GB SXM$1.19
$1.78−33% - RTX A6000 48GB$0.36
$0.55−35%
Where the offerings are
Offerings by area of the region they run in. Multi-region and unlocated offerings are left out.
Availability
Available means available now.
A provider’s page saying “in stock” is not enough. Kovara shows an offering as available only with fresh evidence: a confirmation within 30 minutes, a live capacity reading within 15, or a provider’s own attestation until it expires. Everything else is shown as “Listed”, with how long ago it was checked.
Providers by offerings tracked
| Provider | Offerings | GPU models |
|---|---|---|
| Microsoft Azure | 244 | 4 |
| SaladCloud | 71 | 20 |
| Massed Compute | 67 | 14 |
| CoreWeave | 52 | 11 |
| Verda | 48 | 12 |
| Runpod | 44 | 21 |
| DigitalOcean | 39 | 10 |
| Crusoe Cloud | 36 | 9 |
| Paperspace | 35 | 13 |
| Exoscale | 30 | 6 |
| Lambda | 26 | 11 |
| Hyperstack | 25 | 12 |
Latest recorded prices
| GPU | Provider | Per GPU-hour |
|---|