GPU price index
GPU rental prices: same card, very different bill
Live Updated Oct 17, 2026 · 9 GPU models · prices in USD per GPU-hour
The cheapest H100 SXM we track today is $1.79 an hour (Vast.ai, community). The median of 18 offers is $3.65. Book at the median and you pay 2.0× the cheapest list price for the same model, so check the terms (on-demand, spot or marketplace) and compare before you commit.
The same GPU can cost far more on one cloud than another, and prices move week to week. This index tracks the listed hourly price for each card across providers, the cheapest and median offer, and how both moved. Check it before you book capacity. If you own a card, the calculator below estimates what renting it out could bring in after electricity.
Disclosure: Some links on this page are affiliate links. If you buy through them, we may earn a commission at no extra cost to you. It never changes our verdicts or rankings. How we make money.
H100 rental prices fell sharply through 2025 and rose again in 2026: several providers raised list prices between August and October 2026. If you rent GPUs, a few cents an hour across a cluster and a month is real money; if you own one, the same numbers show what it's worth on a marketplace.
Cheapest vs median, per GPU-hour
- H100 SXM$1.79 / $3.65
- H100 PCIe$1.66 / $2.62
- H200$2.25 / $5.00
- B200$3.74 / $6.69
- A100 80GB$0.80 / $1.85
- L40S$0.43 / $1.43
- RTX 4090$0.34 / $0.41
- RTX 5090$0.47 / $0.69
- RTX 3090$0.15 / $0.22
Cheapest offer Median offer. Shared scale, $0 to $6.69 an hour. List prices; offer types differ (see the table).
GPU rental prices by model
| GPU | VRAM | Cheapest | Median | 7 days | 30 days | Offers |
|---|---|---|---|---|---|---|
| H100 SXM | 80 GB | $1.79 Vast.ai · community | $3.65 | — | — | 18 |
| H100 PCIe | 80 GB | $1.66 Latitude.sh via Shadeform · on-demand | $2.62 | — | — | 10 |
| H200 | 141 GB | $2.25 Excess Supply via Shadeform · on-demand | $5.00 | — | — | 15 |
| B200 | 180 GB | $3.74 Boost Run via Shadeform · on-demand | $6.69 | — | — | 13 |
| A100 80GB | 80 GB | $0.80 Vast.ai · community | $1.85 | — | — | 16 |
| L40S | 48 GB | $0.43 Vast.ai · community | $1.43 | — | — | 15 |
| RTX 4090 | 24 GB | $0.34 RunPod · community | $0.41 | — | — | 6 |
| RTX 5090 | 32 GB | $0.47 Vast.ai · community | $0.69 | — | — | 5 |
| RTX 3090 | 24 GB | $0.15 Vast.ai · community | $0.22 | — | — | 3 |
Price changes: magenta means renting got dearer, volt cheaper. "—" means no reading on that update yet, not zero.
Calculator
Own a GPU? See what renters pay for it
Per month (730 hours). Income uses the median rental price for that GPU in this index times the host's share after marketplace fees; electricity counts only the hours it's rented, at the card's rated power. It leaves out the rest of the machine, internet, wear, hardware cost, taxes and idle time between renters, so treat it as a ceiling, not a forecast. The $0.17 default is a placeholder: use your own rate from your bill.
Every offer we track
H100 SXM 18 offers · from $1.79/hr
| Provider | Price/hr | Type | Seen |
|---|---|---|---|
| Nebius | $0.79 | spot | Oct 17, 2026 |
| Vast.ai | $1.79 | community | Oct 17, 2026 |
| Horizon via Shadeform | $1.95 | on-demand | Oct 17, 2026 |
| Voltage Park | $1.99 | on-demand | Oct 17, 2026 |
| CoreWeave | $2.44 | spot | Oct 17, 2026 |
| Denvr Dataworks via Shadeform | $2.50 | on-demand | Oct 17, 2026 |
| RunPod | $2.69 | community | Oct 17, 2026 |
| Massed Compute via Shadeform | $3.14 | on-demand | Oct 17, 2026 |
| Hyperstack via Shadeform | $3.20 | on-demand | Oct 17, 2026 |
| Verda (DataCrunch) via Shadeform | $3.31 | on-demand | Oct 17, 2026 |
| Lambda | $3.99 | on-demand | Oct 17, 2026 |
| RunPod | $3.99 | on-demand | Oct 17, 2026 |
| Together AI | $3.99 | on-demand | Oct 17, 2026 |
| DigitalOcean | $4.41 | on-demand | Oct 17, 2026 |
| Nebius | $4.50 | on-demand | Oct 17, 2026 |
| CoreWeave | $6.16 | on-demand | Oct 17, 2026 |
| AWS | $6.88 | on-demand | Oct 17, 2026 |
| Azure | $12.29 | on-demand | Oct 17, 2026 |
H100 PCIe 10 offers · from $1.66/hr
| Provider | Price/hr | Type | Seen |
|---|---|---|---|
| Latitude.sh via Shadeform | $1.66 | on-demand | Oct 17, 2026 |
| RunPod | $1.99 | community | Oct 17, 2026 |
| Vast.ai | $2.00 | community | Oct 17, 2026 |
| IMWT via Shadeform | $2.48 | on-demand | Oct 17, 2026 |
| Hyperstack via Shadeform | $2.50 | on-demand | Oct 17, 2026 |
| Massed Compute via Shadeform | $2.73 | on-demand | Oct 17, 2026 |
| RunPod | $2.89 | on-demand | Oct 17, 2026 |
| Lambda | $3.29 | on-demand | Oct 17, 2026 |
| Scaleway via Shadeform | $3.30 | on-demand | Oct 17, 2026 |
| Paperspace via Shadeform | $5.99 | on-demand | Oct 17, 2026 |
H200 15 offers · from $2.25/hr
| Provider | Price/hr | Type | Seen |
|---|---|---|---|
| Nebius | $0.79 | spot | Oct 17, 2026 |
| Excess Supply via Shadeform | $2.25 | on-demand | Oct 17, 2026 |
| Boost Run via Shadeform | $2.45 | on-demand | Oct 17, 2026 |
| CoreWeave | $2.58 | spot | Oct 17, 2026 |
| Horizon via Shadeform | $3.00 | on-demand | Oct 17, 2026 |
| RunPod | $3.59 | community | Oct 17, 2026 |
| Verda (DataCrunch) via Shadeform | $4.21 | on-demand | Oct 17, 2026 |
| DigitalOcean | $4.47 | on-demand | Oct 17, 2026 |
| Vast.ai | $5.00 | community | Oct 17, 2026 |
| RunPod | $5.29 | on-demand | Oct 17, 2026 |
| Nebius | $5.40 | on-demand | Oct 17, 2026 |
| Together AI | $5.99 | on-demand | Oct 17, 2026 |
| CoreWeave | $6.31 | on-demand | Oct 17, 2026 |
| AWS | $7.91 | on-demand | Oct 17, 2026 |
| Azure | $10.60 | on-demand | Oct 17, 2026 |
B200 13 offers · from $3.74/hr
| Provider | Price/hr | Type | Seen |
|---|---|---|---|
| Nebius | $0.99 | spot | Oct 17, 2026 |
| Boost Run via Shadeform | $3.74 | on-demand | Oct 17, 2026 |
| CoreWeave | $4.26 | spot | Oct 17, 2026 |
| Excess Supply via Shadeform | $4.51 | on-demand | Oct 17, 2026 |
| RunPod | $5.98 | community | Oct 17, 2026 |
| Vast.ai | $6.25 | community | Oct 17, 2026 |
| Verda (DataCrunch) via Shadeform | $6.31 | on-demand | Oct 17, 2026 |
| Lambda | $6.69 | on-demand | Oct 17, 2026 |
| RunPod | $7.99 | on-demand | Oct 17, 2026 |
| Together AI | $8.19 | on-demand | Oct 17, 2026 |
| Nebius | $8.50 | on-demand | Oct 17, 2026 |
| CoreWeave | $8.60 | on-demand | Oct 17, 2026 |
| AWS | $14.24 | on-demand | Oct 17, 2026 |
A100 80GB 16 offers · from $0.80/hr
| Provider | Price/hr | Type | Seen |
|---|---|---|---|
| Vast.ai | $0.80 | community | Oct 17, 2026 |
| CoreWeave | $1.19 | spot | Oct 17, 2026 |
| RunPod | $1.19 | community | Oct 17, 2026 |
| Excess Supply via Shadeform | $1.35 | on-demand | Oct 17, 2026 |
| Hyperstack via Shadeform | $1.35 | on-demand | Oct 17, 2026 |
| Massed Compute via Shadeform | $1.35 | on-demand | Oct 17, 2026 |
| Denvr Dataworks via Shadeform | $1.50 | on-demand | Oct 17, 2026 |
| RunPod | $1.79 | on-demand | Oct 17, 2026 |
| Verda (DataCrunch) via Shadeform | $1.85 | on-demand | Oct 17, 2026 |
| Crusoe via Shadeform | $2.00 | on-demand | Oct 17, 2026 |
| Vultr via Shadeform | $2.60 | on-demand | Oct 17, 2026 |
| CoreWeave | $2.70 | on-demand | Oct 17, 2026 |
| Lambda | $2.79 | on-demand | Oct 17, 2026 |
| Paperspace via Shadeform | $3.22 | on-demand | Oct 17, 2026 |
| AWS | $3.43 | on-demand | Oct 17, 2026 |
| Azure | $4.10 | on-demand | Oct 17, 2026 |
L40S 15 offers · from $0.43/hr
| Provider | Price/hr | Type | Seen |
|---|---|---|---|
| Vast.ai | $0.43 | community | Oct 17, 2026 |
| Boost Run via Shadeform | $0.69 | on-demand | Oct 17, 2026 |
| Latitude.sh via Shadeform | $0.74 | on-demand | Oct 17, 2026 |
| Nebius | $0.74 | spot | Oct 17, 2026 |
| RunPod | $0.79 | community | Oct 17, 2026 |
| Massed Compute via Shadeform | $0.88 | on-demand | Oct 17, 2026 |
| CoreWeave | $0.99 | spot | Oct 17, 2026 |
| RunPod | $1.09 | on-demand | Oct 17, 2026 |
| Verda (DataCrunch) via Shadeform | $1.43 | on-demand | Oct 17, 2026 |
| Crusoe via Shadeform | $1.50 | on-demand | Oct 17, 2026 |
| Nebius | $1.55 | on-demand | Oct 17, 2026 |
| DigitalOcean | $1.57 | on-demand | Oct 17, 2026 |
| Scaleway via Shadeform | $1.70 | on-demand | Oct 17, 2026 |
| AWS | $1.86 | on-demand | Oct 17, 2026 |
| CoreWeave | $2.25 | on-demand | Oct 17, 2026 |
RTX 4090 6 offers · from $0.34/hr
| Provider | Price/hr | Type | Seen |
|---|---|---|---|
| RunPod | $0.34 | community | Oct 17, 2026 |
| Excess Supply via Shadeform | $0.40 | on-demand | Oct 17, 2026 |
| Phyntec via Shadeform | $0.40 | on-demand | Oct 17, 2026 |
| Vast.ai | $0.41 | community | Oct 17, 2026 |
| Boost Run via Shadeform | $0.49 | on-demand | Oct 17, 2026 |
| RunPod | $0.89 | on-demand | Oct 17, 2026 |
RTX 5090 5 offers · from $0.47/hr
| Provider | Price/hr | Type | Seen |
|---|---|---|---|
| Vast.ai | $0.47 | community | Oct 17, 2026 |
| Excess Supply via Shadeform | $0.65 | on-demand | Oct 17, 2026 |
| RunPod | $0.69 | community | Oct 17, 2026 |
| Phyntec via Shadeform | $0.70 | on-demand | Oct 17, 2026 |
| RunPod | $1.19 | on-demand | Oct 17, 2026 |
How we collect this
Our system reads each provider's public pricing page (no logins, no API keys) and records the listed hourly price per GPU, whether it's on-demand, spot or a marketplace listing, and when it was seen. "Cheapest" is the lowest current offer; "median" is the middle offer across providers. Changes compare today's median with 7 and 30 days ago.
- List prices, not quotes. Big reservations, commitments and sales discounts aren't public.
- Spot and marketplace offers can disappear (interruptible instances stop when someone outbids you).
- Like for like matters. SXM and PCIe cards, VRAM and the CPU, RAM and storage around the GPU all change the price.
Sources
- console.vast.ai/api/v0/bundles/
- api.runpod.io/graphql
- lambda.ai/pricing
- nebius.com/prices
- coreweave.com/pricing
- together.ai/pricing
- digitalocean.com/pricing/gpu-droplets
- voltagepark.com/pricing
- api.shadeform.ai/v1/instances/types
- b0.p.awsstatic.com/pricing/2.0/meteredUnitMaps/ec2/USD/curre
- prices.azure.com/api/retail/prices
Before you rent
FAQ
What is the cheapest GPU cloud?
It depends on the card and how much interruption you can accept. Marketplace and community-cloud offers (Vast.ai, RunPod Community Cloud) are usually cheapest; on-demand capacity from Lambda, Nebius and the hyperscalers costs more. The table shows the current cheapest offer per GPU.
How much does an H100 cost per hour?
On 17 October 2026, RunPod listed an H100 SXM at $3.99 an hour on Secure Cloud and $2.69 on Community Cloud, and DigitalOcean listed $4.41 on demand. The index above shows today's prices across providers.
Can I make money renting out my GPU?
Possibly, on marketplaces such as Vast.ai, if your card is in demand and stays rented. Income depends on utilisation, and electricity, hardware wear and idle time eat into it. Use the calculator for a ceiling, not a forecast.