> For the complete documentation index, see [llms.txt](https://docs.nebulablock.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.nebulablock.com/products/gpu-cloud.md).

# GPU Cloud

Rent NVIDIA B300, B200, H200, H100, A100, L40S and RTX GPUs by the hour across Canada, the US, and Europe.

Rent GPU and CPU compute by the hour for training, fine-tuning, inference, rendering, and simulation. Pick your hardware and OS image, attach an SSH key, and the machine is reachable in minutes.

## Hardware

| GPU                    | Typical use                                                     |
| ---------------------- | --------------------------------------------------------------- |
| **NVIDIA B300 / B200** | Frontier-scale training and inference on Blackwell              |
| **NVIDIA H200**        | Large-model training and high-throughput inference              |
| **NVIDIA H100**        | Large-scale training, available as VM, container, or bare metal |
| **NVIDIA A100**        | AI/ML training and data analytics                               |
| **NVIDIA L40 / L40S**  | Cost-effective inference and graphics workloads                 |
| **NVIDIA RTX series**  | Budget-friendly rendering, visualization, and smaller jobs      |

Configurations run from a single GPU up to 8-GPU nodes (10 in some regions). Availability changes constantly — the [Deploy](https://console.nebulablock.com/deploy) page in the console shows what is in stock right now, and so does [List Products](/api-reference/platform-api/list-products.md).

How much you can run at once is capped by your account's [hourly spending limit](/getting-started/get-started/spending-limits.md). New accounts start above the price of the dearest single-GPU configuration, so any of those can be launched immediately; multi-GPU nodes need a limit that has grown with your payment history, or an increase requested from the billing page — usually granted within a few hours.

## Regions

Instances are available in **Canada**, the **United States**, **Finland**, **France**, and **Norway**. Canadian capacity is what backs [Sovereign AI Compute](https://www.nebulablock.com/ai-sovereign-compute) for workloads with data-residency requirements.

## Form factors

* **Virtual machines** — the default, available across every region and GPU type.
* **Containers** — lighter-weight, for workloads that do not need a full VM.
* **Bare metal** — dedicated hardware with no hypervisor, offered on H100.

## Key features

* **Secure access.** SSH key authentication, or a generated username and password where a configuration does not support keys.
* **Firewall rules.** Manage per-instance rules from the console or the [Platform API](/api-reference/platform-api/list-products/firewall-rules.md). Not available on every configuration — check `firewall_supported` on the instance first.
* **Lifecycle control.** Start, stop, reboot, and delete instances from the console or the API.
* **High-speed networking.** Dedicated bandwidth for moving large datasets in and out.

## Access and billing

* **Tier requirement:** CPU instances need Tier 2, GPU instances need Tier 3 (a $10 deposit). See [Tiers and Rate Limits](/account/tiers-and-limits.md).
* **Billing:** hourly, deducted from your credit balance for as long as the instance exists.
* Rates per configuration are on the [pricing page](https://www.nebulablock.com/pricing/gpu-instances).

> **Important:** Instances are billed while they exist, and are deleted automatically if your credit balance runs out. Turn on auto-pay for anything long-running — see [Billing](/getting-started/get-started/billing.md).

## See also

* [Quickstart](/products/gpu-cloud/quickstart.md)
* [SSH Keys](/products/gpu-cloud/ssh-keys.md)
* [Instances API](/api-reference/platform-api/list-products.md)
* [Glossary](/resources/glossary.md)