On-demand GPU instance pricing

Billed per minute with expandable configurations for any project

RTX A6000 On-demand
$0.27
/hr
Pay as you go
Expandable vCPUs and RAM
North American Data Centers
Get started
1-2 x A100 (80GB) On-demand
$0.78
/hr
Pay as you go
Expandable vCPUs and RAM
North American Data Centers
Get started
1-2 x H100 PCIe On-demand
$1.38
/GPU/hr
Pay as you go
Expandable vCPUs and RAM
North American Data Centers
Get started
1-8 x A100 80GB NVLink On-demand
$1.79
/GPU/hr
Best for long-running jobs with high uptime
18 vCPUs 90 GB RAM included per GPU
Expandable disk
Premium networking
Get started
1-8 x H100 PCIe NVLink On-demand
$2.49
/GPU/hr
Production ready
18 vCPUs 90GB RAM included per GPU
Expandable disk
Premium networking
Get started

Add-ons

Additional storage
(100GB included, billed while running)
$0.10
/GB/mo
Snapshots
(Long-term storage)
$0.05
/GB/mo
Additional cores
(4 vCPUs included, 8GB RAM per vCPU)
$0.06
/GB/mo
Additional memory
(32GB included)
8GB
included per vCPU

See how much you save

Frequently asked questions

What is Thunder Compute?

A GPU cloud platform for developers. Thunder Compute is low-cost, easy-to-use, and scalable.

Who should use Thunder Compute cloud GPUs?

Data scientists, startups, and researchers who need reliable, affordable GPU cloud instances for their work

How is Thunder Compute cheaper than competitors?

Within our infrastructure we use proprietary software optimizations to improve scheduling efficiency. The result is we waste less and pass on the savings. You don't have to think about it; it's the same GPU at a lower price.

Why does Thunder Compute have separate prototyping and production pricing?

We provide stronger performance guarantees for production mode. Performance may vary on prototyping mode. If you are deploying inference servers to production or need a larger config, use production mode. Otherwise, prototyping meets most people's needs.

Does Thunder Compute offer persistent storage?

Yes. We offer persistent storage in the form of snapshots. When you finish your work you can take a snapshot of the instance and restore from it another time. Snapshots are billed hourly according to size.

Get the world's
cheapest GPUs

Low prices, developer-first features, simple UX. Start building today.

Get started