Cloud GPU prices per hour, by model and by provider.
Scraped daily from provider APIs and price pages, kept for 90 days. Each model page shows the cheapest live offer from every provider today, the last-seen price where a provider lists nothing today, and the daily floor and median since July. Prices are what the provider published at the time scraped; availability, egress and storage are not included.
Models with live offers
18
18 seen in 90 days
Providers
7
7 tracked
Newest scrape
Nov 16, 2026
daily at 20:45 UTC
H100 floor today
$0.67
cheapest live offer, any provider
By model
Every GPU model, cheapest live price first
| Model | VRAM | Floor today | Median today | Providers live | Last seen |
|---|---|---|---|---|---|
| RTX 4080 | 16 GB | $0.16 | $0.27 | 4 | Nov 16, 2026 |
| A40 | 48 GB | $0.16 | $0.47 | 4 | Nov 16, 2026 |
| A100 SXM4 80GB | 80 GB | $0.27 | $1.10 | 4 | Nov 16, 2026 |
| H100 PCIe 80GB | 80 GB | $0.67 | $2.14 | 4 | Nov 16, 2026 |
| H100 SXM5 80GB | 80 GB | $0.67 | $1.79 | 4 | Nov 16, 2026 |
| RTX 4090 | 24 GB | $0.00 | $0.37 | 3 | Nov 16, 2026 |
| A100 SXM4 40GB | 40 GB | $0.20 | $0.53 | 3 | Nov 16, 2026 |
| L40S | 48 GB | $0.27 | $0.77 | 3 | Nov 16, 2026 |
| B200 SXM6 180GB | 180 GB | $4.26 | $8.75 | 3 | Nov 16, 2026 |
| RTX 3080 | 10 GB | $0.07 | $0.13 | 2 | Nov 16, 2026 |
| RTX 3090 | 24 GB | $0.08 | $0.16 | 2 | Nov 16, 2026 |
| RTX 3090 Ti | 24 GB | $0.12 | $0.26 | 2 | Nov 16, 2026 |
| RTX 5080 | 16 GB | $0.12 | $0.24 | 2 | Nov 16, 2026 |
| A100 PCIe 80GB | 80 GB | $0.15 | $0.87 | 2 | Nov 16, 2026 |
| RTX 5090 | 32 GB | $0.24 | $0.53 | 2 | Nov 16, 2026 |
| V100 SXM2 32GB | 32 GB | $0.79 | $2.30 | 2 | Nov 16, 2026 |
| H200 SXM 141GB | 141 GB | $1.32 | $6.11 | 2 | Nov 16, 2026 |
| GH200 96GB | 96 GB | $2.59 | $6.50 | 2 | Nov 16, 2026 |
By provider
Every provider, live catalog size first
Questions
What people search for before renting a GPU
How do GPU cloud prices compare across providers?
On Nov 16, 2026 AltStreet's scrape held live offers from 7 providers across 18 GPU models. The same model is priced differently by pricing type (on-demand against spot), by provider type (a hyperscaler, a specialist cloud, a decentralized marketplace of individual hosts) and by what is bundled with the instance. Each model page lists every provider's cheapest live offer with its type and region so the comparison is like for like.
Where can I find a GPU cloud comparison tool for pricing and availability?
This page and the model pages under it are that tool for price: daily scrapes, 90 days of history, and the last-seen price where a provider has nothing live. Availability is not scraped; a listed price can be sold out, which is why each row links to the provider. The yield dashboard adds utilization and payback math on top of the prices.
What is the cheapest cloud GPU right now?
RTX 4090 at $0.00 an hour was the lowest live price on Nov 16, 2026, across 3 providers. Cheapest per hour is not cheapest per unit of work; the model pages carry VRAM and release date beside the price.
How much is an H100 per hour?
The cheapest live H100 PCIe 80GB offer was $0.67 an hour on Nov 16, 2026, with a median of $2.14 across 4 providers. The H100 page shows each provider's price and the 90-day floor.
Are these the prices I will actually pay?
They are the hourly rates each provider published at the time scraped, before egress, storage, reserved or committed discounts, and before availability. Treat them as the comparison, then confirm the live quote on the provider's page.
The hand-written comparison with the methodology and the cost calculator: cloud GPU pricing comparison, yield dashboard, AI infrastructure calculator. Prices are descriptive data, not an offer; confirm on the provider's page before renting.
