Find compute
Find the compute your work needs
Looking to rent H100 or H200 GPUs? Compare selected options by price, memory, region and GPUs per instance, then check the configuration with the provider.
Compare GPU options
Start with the memory your workload needs and where your data can run. Compare these selected A100, H100 and H200 options, then check that the model and configuration suit your workload. The same memory size does not imply the same performance.
Checked 10 November 2026. Use “Filter rows” to look for a GPU, provider or place, such as H100, Finland or Canada.
| Provider | GPU and memory | Location | Access and next step | Price basis | Evidence |
|---|---|---|---|---|---|
| Hyperstack | NVIDIA A100 PCIe 80 GB per GPU GPUs per instance: 1, 2, 4 or 8 |
Quebec, Canada CANADA-1 |
Self-serve virtual machine Check in Hyperstack First VM checklist |
US$1.35 per GPU-hour On-demand list rate; billed by the minute. Compare workload compatibility and performance, not memory size alone. 24-hour GPU-charge example: 1 GPU: US$32.40; 8 GPUs: US$259.20. Calculated as rate × GPU count × 24 hours, not a full bill or a performance comparison. Other charges and tax may apply. What this rate includesThe GPU VM's allocated CPU cores, RAM and local root and ephemeral disks are included. Public IP addresses, attached shared volumes, snapshots and object storage are billed separately. Check the complete configuration before deploying. Billing documentation checked 10 November 2026. Check Hyperstack’s resource pricing. |
|
| Hyperstack | NVIDIA H100 PCIe 80 GB per GPU GPUs per instance: 1, 2, 4 or 8 |
Quebec, Canada CANADA-1 |
Self-serve virtual machine Check in Hyperstack First VM checklist |
US$2.50 per GPU-hour On-demand list rate; billed by the minute. 24-hour GPU-charge example: 1 GPU: US$60.00; 8 GPUs: US$480.00. Calculated as rate × GPU count × 24 hours, not a full bill or a performance comparison. Other charges and tax may apply. What this rate includesThe GPU VM's allocated CPU cores, RAM and local root and ephemeral disks are included. Public IP addresses, attached shared volumes, snapshots and object storage are billed separately. Check the complete configuration before deploying. Billing documentation checked 10 November 2026. Check Hyperstack’s resource pricing. |
|
| Hyperstack | NVIDIA H100 SXM5 80 GB per GPU GPUs per instance: 8 |
Quebec, Canada or Texas, United States CANADA-1 / US-1 |
Self-serve virtual machine Check in Hyperstack First VM checklist |
US$3.20 per GPU-hour Eight-GPU VM: US$25.60 per hour. On-demand list rate; billed by the minute. 24-hour GPU-charge example: 8 GPUs: US$614.40. Calculated as rate × GPU count × 24 hours, not a full bill or a performance comparison. Other charges and tax may apply. What this rate includesThe GPU VM's allocated CPU cores, RAM and local root and ephemeral disks are included. Public IP addresses, attached shared volumes, snapshots and object storage are billed separately. Check the complete configuration before deploying. Billing documentation checked 10 November 2026. Check Hyperstack’s resource pricing. |
|
| Hyperstack | NVIDIA H200 SXM5 141 GB per GPU GPUs per instance: 8 |
Quebec, Canada CANADA-1 |
Self-serve virtual machine Check in Hyperstack First VM checklist |
US$3.99 per GPU-hour Eight-GPU VM: US$31.92 per hour. On-demand list rate; billed by the minute. 24-hour GPU-charge example: 8 GPUs: US$766.08. Calculated as rate × GPU count × 24 hours, not a full bill or a performance comparison. Other charges and tax may apply. What this rate includesThe GPU VM's allocated CPU cores, RAM and local root and ephemeral disks are included. Public IP addresses, attached shared volumes, snapshots and object storage are billed separately. Check the complete configuration before deploying. Billing documentation checked 10 November 2026. Check Hyperstack’s resource pricing. |
|
| Nebius | NVIDIA H100 NVLink 80 GB per GPU GPUs per instance: 1 or 8 |
Finland eu-north1 |
Self-serve virtual machine Start with Nebius |
US$4.50 per GPU-hour On-demand listed rate; billed by the second, excluding tax. Storage charged separately; check the full configuration quote. 24-hour GPU-charge example: 1 GPU: US$108.00; 8 GPUs: US$864.00. Calculated as rate × GPU count × 24 hours, not a full bill or a performance comparison. Other charges and tax may apply. |
Sources and dateChecked . Region locations · GPU specifications and regions · GPU rates and billing · Self-service access |
| Nebius | NVIDIA H200 NVLink 141 GB per GPU GPUs per instance: 1 or 8 |
Finland, France or Kansas City, United States eu-north1 / eu-west1 / us-central1 |
Self-serve virtual machine Start with Nebius |
US$5.40 per GPU-hour On-demand listed rate; billed by the second, excluding tax. Storage charged separately; check the full configuration quote. 24-hour GPU-charge example: 1 GPU: US$129.60; 8 GPUs: US$1036.80. Calculated as rate × GPU count × 24 hours, not a full bill or a performance comparison. Other charges and tax may apply. |
Sources and dateChecked . Region locations · GPU specifications and regions · GPU rates and billing · Self-service access |
| Lambda | NVIDIA H100 SXM or PCIe 80 GB per GPU GPUs per instance: SXM: 1, 2, 4 or 8; PCIe: 1 |
Choose a region in the console | Self-serve on-demand instance Check in Lambda |
Check the current official price The pricing page displays rates effective 16 November 2026. No future rate is presented here as today’s price. |
These are published options, not a live stock check or a ranking. Compare the whole instance bill: GPU count, CPU, RAM, storage, data transfer and tax can change the total. Confirm the selected GPU and region in the provider’s console before committing.
Need help narrowing it down? Compare the forms of access, use the workload chooser, or take the buyer’s checklist with you.
Other routes into compute
Free, funded and research access
Explore programmes and institutional routes. Check each route’s eligibility and current status.
Find a programme →Find a provider
Investigate suppliers and follow their official pages for current terms and availability.
Browse providers →Help choosing GPU access
A practical guide to the questions to ask before committing.
Read the guide →Check before applying or buying
- Suitability. Establish the workload, software, memory, data location, timing and support you need.
- Eligibility and restrictions. Check geography, affiliation, permitted uses and data requirements at the named source.
- Effort and commitment. Establish application steps, allocation periods, onboarding work, contracts and obligations.
- Current position. Confirm the route is open and the required service can be supplied. A historical listing cannot settle this.
- Alternatives. Consider a different access route, a smaller workload or better use of resources already available to you.
Go a little deeper
Use the funding questions, service routes and glossary. Eligibility, allocation and live stock still need confirming at the source.
Researching the wider system? Explore the compute economy. Found incorrect or missing evidence? Suggest a correction.