Your first GPU, start to finish

1. Create an account

Email and password. No card required to sign up — you only pay when you buy credit.

Create account

2. Add credit

Credit packs start at $25 and never expire. Larger packs carry a bonus. Billing draws down from this balance per second while an instance runs.

3. Deploy an instance

In the console, choose Deploy. Pick three things:

• GPU — start with an RTX 4090 at $0.44/hr if you are just trying the platform
• Region — whichever is closest to your data
• Image — ubuntu-22-cuda-12 covers most PyTorch work

The provisioning log streams while the node builds. You are not billed until it reaches running.

4. Connect

Once the instance is running the console shows its address. Connect as root:

Confirm the GPU is visible before you start real work:

And that the framework can see it:

5. Destroy it when you are done

This is the step people forget, and it is the one that costs money. Destroying releases the hardware, erases the disk and stops billing immediately.

If you want to come back to the same machine tomorrow, stop it instead — compute billing halts and you keep paying only for the root volume.

Next

• API reference for scripting all of the above
• Storage for keeping datasets between runs
• GPU comparison for choosing the right card