What GPU clouds charge per GPU-hour
7 of 8 GPU rental providers publish a per-GPU-hour price on the pages we read, read 8 October 2026.
Publishes a per-GPU-hour price
H100 on-demand price per GPU-hour
A100 on-demand price per GPU-hour
RTX 4090 price per GPU-hour
Kind of provider
Reserved, spot or commitment terms
The full board, one provider per row
Each row is one provider. Click an answer to see the sentence behind it, quoted from the page with the date we read it. Where it says "not stated", the page we read is silent on that question.
| Provider | Publishes a per-GPU-hour price | H100 on-demand price per GPU-hour | A100 on-demand price per GPU-hour | RTX 4090 price per GPU-hour | Kind of provider | Reserved, spot or commitment terms |
|---|---|---|---|---|---|---|
| Akash | publishes a price | 2.71 USD per GPU-hourNot addressed in Akash GPU pricing page | not statedNot addressed in Akash GPU pricing page Not addressed in Akash, The Bid: H100 rental price in September 2026 | not statedNot addressed in Akash GPU pricing page Not addressed in Akash, The Bid: H100 rental price in September 2026 | marketplace of third-party hosts | the same providers sell the identical hardware for materially less through spot and…Not addressed in Akash GPU pricing page |
| GMI Cloud | publishes a price | 2.60 USD per GPU-hour | not statedNot addressed in GMI Cloud pricing | not statedNot addressed in GMI Cloud pricing | GPU cloud | Reduce unit GPU costs through reserved capacity and sustained deployments. |
| Nosana | not statedNot addressed in Nosana home page | not statedNot addressed in Nosana home page | not statedNot addressed in Nosana home page | not statedNot addressed in Nosana home page | marketplace of third-party hosts | not statedNot addressed in Nosana home page |
| Oracle Cloud | publishes a price | 10.00 USD per GPU-hour | 4.00 USD per GPU-hour | not statedNot addressed in Oracle Cloud GPU price list | not statedNot addressed in Oracle Cloud GPU price list | Dedicated AI clusters require a minimum commitment of 1 unit-hour per cluster for hosting… |
| RunPod | publishes a price | 3.49 USD per GPU-hour | 1.59 USD per GPU-hour | 0.74 USD per GPU-hour | GPU cloud | For reserved capacity and contract pricing, speak to our enterprise sales team. |
| SaladCloud | publishes a price | not statedNot addressed in SaladCloud pricing | not statedNot addressed in SaladCloud pricing | 0.33 USD per GPU-hour | marketplace of third-party hosts | Deploying more than 100 GPUs, long-running jobs, or committed spend? Sales will price it… |
| Spheron | publishes a price | 2.64 USD per GPU-hour | 1.43 USD per GPU-hour | 0.58 USD per GPU-hour | marketplace of third-party hosts | Per-minute billing, no commitments, and instant deployment from certified data centers… |
| Voltage Park | publishes a price | 1.99 USD per GPU-hour | not statedNot addressed in Voltage Park pricing | not statedNot addressed in Voltage Park pricing | GPU cloud | not statedNot addressed in Voltage Park pricing |
Where each answer was read (54)
- Akash, Akash GPU pricing page, read 2026-10-08, Publishes a per-GPU-hour price:
Prices are per GPU-hour in USD, derived from recent provider bids on the network and weighted by each provider's GPU count.
- Akash, Akash GPU pricing page, read 2026-10-08, H100 on-demand price per GPU-hour: Left open by the page (1,234 characters, read in full 2026-10-08).
- Akash, Akash GPU pricing page, read 2026-10-08, A100 on-demand price per GPU-hour: Left open by the page (1,234 characters, read in full 2026-10-08).
- Akash, Akash GPU pricing page, read 2026-10-08, RTX 4090 price per GPU-hour: Left open by the page (1,234 characters, read in full 2026-10-08).
- Akash, Akash GPU pricing page, read 2026-10-08, Kind of provider:
Prices are per GPU-hour in USD, derived from recent provider bids on the network and weighted by each provider's GPU count.
- Akash, Akash GPU pricing page, read 2026-10-08, Reserved, spot or commitment terms: Left open by the page (1,234 characters, read in full 2026-10-08).
- Akash, Akash, The Bid: H100 rental price in September 2026, read 2026-10-08, Publishes a per-GPU-hour price:
Akash Marketplace SXM5 Marketplace ~$2.71
- Akash, Akash, The Bid: H100 rental price in September 2026, read 2026-10-08, H100 on-demand price per GPU-hour:
Akash Marketplace SXM5 Marketplace ~$2.71
- Akash, Akash, The Bid: H100 rental price in September 2026, read 2026-10-08, A100 on-demand price per GPU-hour: Left open by the page (15,872 characters, read in full 2026-10-08).
- Akash, Akash, The Bid: H100 rental price in September 2026, read 2026-10-08, RTX 4090 price per GPU-hour: Left open by the page (15,872 characters, read in full 2026-10-08).
- Akash, Akash, The Bid: H100 rental price in September 2026, read 2026-10-08, Kind of provider:
A marketplace GPU cloud, like Akash, lets multiple providers bid to fill capacity, which tends to push rates toward the market floor.
- Akash, Akash, The Bid: H100 rental price in September 2026, read 2026-10-08, Reserved, spot or commitment terms:
the same providers sell the identical hardware for materially less through spot and multi-year commitments, covered next
- GMI Cloud, GMI Cloud pricing, read 2026-10-08, Publishes a per-GPU-hour price:
NVIDIA H100 GPU from $2.60 /GPU-hour
- GMI Cloud, GMI Cloud pricing, read 2026-10-08, H100 on-demand price per GPU-hour:
NVIDIA H100 GPU from $2.60 /GPU-hour
- GMI Cloud, GMI Cloud pricing, read 2026-10-08, A100 on-demand price per GPU-hour: Left open by the page (2,839 characters, read in full 2026-10-08).
- GMI Cloud, GMI Cloud pricing, read 2026-10-08, RTX 4090 price per GPU-hour: Left open by the page (2,839 characters, read in full 2026-10-08).
- GMI Cloud, GMI Cloud pricing, read 2026-10-08, Kind of provider:
Train models and run inference on dedicated NVIDIA GPUs inside GMI-operated data centers.
- GMI Cloud, GMI Cloud pricing, read 2026-10-08, Reserved, spot or commitment terms:
Reduce unit GPU costs through reserved capacity and sustained deployments.
- Nosana, Nosana home page, read 2026-10-08, Publishes a per-GPU-hour price: Left open by the page (6,079 characters, read in full 2026-10-08).
- Nosana, Nosana home page, read 2026-10-08, H100 on-demand price per GPU-hour: Left open by the page (6,079 characters, read in full 2026-10-08).
- Nosana, Nosana home page, read 2026-10-08, A100 on-demand price per GPU-hour: Left open by the page (6,079 characters, read in full 2026-10-08).
- Nosana, Nosana home page, read 2026-10-08, RTX 4090 price per GPU-hour: Left open by the page (6,079 characters, read in full 2026-10-08).
- Nosana, Nosana home page, read 2026-10-08, Kind of provider:
OpenGPU is a GPU marketplace for AI inference and training, built on a free-market model.
- Nosana, Nosana home page, read 2026-10-08, Reserved, spot or commitment terms: Left open by the page (6,079 characters, read in full 2026-10-08).
- Oracle Cloud, Oracle Cloud GPU price list, read 2026-10-08, Publishes a per-GPU-hour price:
BM.GPU.H100.8 — $10.00 GPU per hour
- Oracle Cloud, Oracle Cloud GPU price list, read 2026-10-08, H100 on-demand price per GPU-hour:
BM.GPU.H100.8 — $10.00 GPU per hour
- Oracle Cloud, Oracle Cloud GPU price list, read 2026-10-08, A100 on-demand price per GPU-hour:
BM.GPU.A100-v2.8 (8x NVIDIA A100 80GB Tensor Core) — $4.00 GPU per hour
- Oracle Cloud, Oracle Cloud GPU price list, read 2026-10-08, RTX 4090 price per GPU-hour: Left open by the page (114,244 characters, read in full 2026-10-08).
- Oracle Cloud, Oracle Cloud GPU price list, read 2026-10-08, Kind of provider: Left open by the page (114,244 characters, read in full 2026-10-08).
- Oracle Cloud, Oracle Cloud GPU price list, read 2026-10-08, Reserved, spot or commitment terms:
Dedicated AI clusters require a minimum commitment of 1 unit-hour per cluster for hosting models.
- RunPod, RunPod GPU instance pricing, read 2026-10-08, Publishes a per-GPU-hour price:
80 GB VRAM 125 GB RAM 20 vCPUs $3.49/hr Deploy H100 SXM
- RunPod, RunPod GPU instance pricing, read 2026-10-08, H100 on-demand price per GPU-hour:
H100 SXM 80 GB VRAM 125 GB RAM 20 vCPUs $3.49/hr Deploy
- RunPod, RunPod GPU instance pricing, read 2026-10-08, A100 on-demand price per GPU-hour:
A100 SXM 80 GB VRAM 125 GB RAM 16 vCPUs $1.59/hr Deploy
- RunPod, RunPod GPU instance pricing, read 2026-10-08, RTX 4090 price per GPU-hour:
RTX 4090 24 GB VRAM 41 GB RAM 6 vCPUs $0.74/hr Deploy
- RunPod, RunPod GPU instance pricing, read 2026-10-08, Kind of provider:
Thousands of GPUs across 30+ regions. Simple pricing plans for teams of all sizes, designed to scale with you.
- RunPod, RunPod GPU instance pricing, read 2026-10-08, Reserved, spot or commitment terms:
For reserved capacity and contract pricing, speak to our enterprise sales team.
- SaladCloud, SaladCloud pricing, read 2026-10-08, Publishes a per-GPU-hour price:
10 × RTX 4090 at Lowest priority · $0.160/GPU/hr · vCPU and RAM included · billed per second, only while running
- SaladCloud, SaladCloud pricing, read 2026-10-08, H100 on-demand price per GPU-hour: Left open by the page (6,169 characters, read in full 2026-10-08).
- SaladCloud, SaladCloud pricing, read 2026-10-08, A100 on-demand price per GPU-hour: Left open by the page (6,169 characters, read in full 2026-10-08).
- SaladCloud, SaladCloud pricing, read 2026-10-08, RTX 4090 price per GPU-hour:
RTX 4090 24 GB $0.330 $0.273 $0.217 $0.160 $116.80
- SaladCloud, SaladCloud pricing, read 2026-10-08, Kind of provider:
GPUs online and schedulable at medium and high priority. Higher priority can take over lower-priority workloads, so the high figure is never smaller than the medium figure.
- SaladCloud, SaladCloud pricing, read 2026-10-08, Reserved, spot or commitment terms:
Deploying more than 100 GPUs, long-running jobs, or committed spend? Sales will price it below list.
- Spheron, Spheron pricing, read 2026-10-08, Publishes a per-GPU-hour price:
Cheapest live on-demand rate per GPU on the Spheron marketplace, with the spot rate below it where one is available.
- Spheron, Spheron pricing, read 2026-10-08, H100 on-demand price per GPU-hour:
H100 HOPPER 80 GB 360 GB RAM 60 vCPUs 1600 GB NVMe $2.64 /HR SPOT $2.20
- Spheron, Spheron pricing, read 2026-10-08, A100 on-demand price per GPU-hour:
A100 AMPERE 80 GB 120 GB RAM 28 vCPUs 850 GB NVMe $1.43 /HR SPOT $1.23
- Spheron, Spheron pricing, read 2026-10-08, RTX 4090 price per GPU-hour:
RTX 4090 ADA 24 GB 24 GB RAM 8 vCPUs 500 GB NVMe $0.58 /HR
- Spheron, Spheron pricing, read 2026-10-08, Kind of provider:
Cheapest live on-demand rate per GPU on the Spheron marketplace, with the spot rate below it where one is available.
- Spheron, Spheron pricing, read 2026-10-08, Reserved, spot or commitment terms:
Per-minute billing, no commitments, and instant deployment from certified data centers worldwide.
- Voltage Park, Voltage Park pricing, read 2026-10-08, Publishes a per-GPU-hour price:
You can rent one H100 GPU for 1 hour starting at $1.99 without a contract.
- Voltage Park, Voltage Park pricing, read 2026-10-08, H100 on-demand price per GPU-hour:
You can rent one H100 GPU for 1 hour starting at $1.99 without a contract.
- Voltage Park, Voltage Park pricing, read 2026-10-08, A100 on-demand price per GPU-hour: Left open by the page (2,692 characters, read in full 2026-10-08).
- Voltage Park, Voltage Park pricing, read 2026-10-08, RTX 4090 price per GPU-hour: Left open by the page (2,692 characters, read in full 2026-10-08).
- Voltage Park, Voltage Park pricing, read 2026-10-08, Kind of provider:
32-8000+ H100 GPUs in a single configuration.
- Voltage Park, Voltage Park pricing, read 2026-10-08, Reserved, spot or commitment terms: Left open by the page (2,692 characters, read in full 2026-10-08).
Not counted on this board
We read 2 named documents that turned out not to be what this board asks about; why is given for each below. They count toward the reading, not the answers.
- getdeploying: H100 rental prices: getdeploying is a third-party price index across providers, not a provider's own pricing page
- computeprices: Salad RTX 4090: computeprices is a third-party price tracker, not SaladCloud's own pricing page
How we read and count
Each provider's own pricing pages were read through estate-fetch and stored whole (sha256 on every row).
For every page, six questions were answered from its own words: publishes a per-GPU-hour price, h100 on-demand price per GPU-hour, a100 on-demand price per GPU-hour, rTX 4090 price per GPU-hour, kind of provider, reserved, spot or commitment terms.
Every stated answer carries the sentence it rests on, checked against the stored text; 'not stated' means the page read does not address the question.
A provider with several pages reads 'varies' where they disagree.
8 providers, 9 pages.
Not read yet, so not counted: 12 named documents.
23 sources, with links
- Vast.ai pricing (not yet read)
- RunPod pricing (not yet read)
- RunPod GPU instance pricing (the publisher's pricing pages, cited for reporting)
- Lambda pricing (not yet read)
- CoreWeave pricing (not yet read)
- Nebius prices (not yet read)
- Together AI pricing (not yet read)
- SaladCloud pricing (the publisher's pricing pages, cited for reporting)
- Paperspace (DigitalOcean) pricing (not yet read)
- GMI Cloud pricing (the publisher's pricing pages, cited for reporting)
- Voltage Park pricing (the publisher's pricing pages, cited for reporting)
- Spheron pricing (the publisher's pricing pages, cited for reporting)
- Akash GPU pricing (the publisher's pricing pages, cited for reporting)
- AWS EC2 P5 (not yet read)
- Google Cloud GPU pricing (not yet read)
- Azure Linux VM pricing (not yet read)
- Oracle Cloud price list (GPU) (the publisher's pricing pages, cited for reporting)
- Nosana (the publisher's pricing pages, cited for reporting)
- Fireworks pricing (not yet read)
- OpenAI API pricing (not yet read)
- Akash: H100 rental price (The Bid) (the publisher's pricing pages, cited for reporting)
- getdeploying: H100 rental prices (read, set aside)
- computeprices: Salad RTX 4090 (read, set aside)
Use this figure
7 of 8 GPU rental providers publish a per-GPU-hour price on the pages we read, read 8 October 2026.
Publishes a per-GPU-hour price, by provider: publishes a price 7 of 8, not stated 1 of 8 (8 providers, read 8 October 2026).
H100 on-demand price per GPU-hour, by provider: US dollars a GPU-hour 6 of 8, not stated 2 of 8 (8 providers, read 8 October 2026).
A100 on-demand price per GPU-hour, by provider: not stated 5 of 8, US dollars a GPU-hour 3 of 8 (8 providers, read 8 October 2026).
RTX 4090 price per GPU-hour, by provider: not stated 5 of 8, US dollars a GPU-hour 3 of 8 (8 providers, read 8 October 2026).
Kind of provider, by provider: marketplace of third-party hosts 4 of 8, GPU cloud 3 of 8, not stated 1 of 8 (8 providers, read 8 October 2026).
Reserved, spot or commitment terms, by provider: stated 6 of 8, not stated 2 of 8 (8 providers, read 8 October 2026).
To cite: "What GPU clouds charge per GPU-hour", updated 2026-10-08, https://rentoutmygpu.com/datasets/gpu-rental-prices-by-provider/.
Get the CSV (54 answers, compiled 2026-10-08)