Credit system
DeployGPU uses a prepaid credit model. You add credits to your account and they are drawn down as you use GPU instances and volumes. All prices are in USD. There are no contracts or minimum commitments. Offer prices include any platform fee, and the deploy page shows the full breakdown before you pay. See DeployGPU pricing for the complete explanation of how a rate is built.How charges work
GPU deployments
Launching an instance reserves one hour of credit at the offer’s rate, and that reservation is what pays for the first hour once the instance is running. After the first hour, billing continues in hourly blocks, charged as each new hour starts. When an instance stops, whether you stopped it or it failed, the unused seconds of the current hour are settled back to your balance right away. If your balance cannot cover the next full hour, the instance does not stop immediately. It switches to per-minute billing and keeps running until your balance reaches zero, then it stops and its disk is deleted, the same as any other stop. Three emails mark this path: one when your running instances have about two hours of paid runtime left in total, one when per-minute billing actually starts, and one when the instance is stopped.Volumes
Volumes are billed hourly by size, whether or not they are attached to a running instance. Billing starts the moment you create a volume and continues until you delete it. If your balance cannot pay for the next volume-hour, the volume is deleted permanently, together with everything on it. No warning email is sent for this. A volume worth keeping needs either a credit buffer you watch yourself or auto top-up turned on. There is no second chance once a volume is gone.Balance warning emails
DeployGPU sends three emails as a balance runs down: a low-balance warning while roughly two hours of paid runtime remain across your running instances, a notice when per-minute drain billing actually starts on a deployment, and a final notice when a deployment is stopped because the balance reached zero. A separate low-balance email also goes out when a charge leaves your balance below $2, unless you have turned billing alerts off in your settings. Volumes do not get any of these emails, which is why the buffer matters more for them.Vast.ai bandwidth
Vast.ai instances bill data transfer separately from the hourly rate. This applies only to Vast.ai’s dashboard-only marketplace, not to QuantaCloud, Massed Compute, Verda or Hyperstack.Payment methods
Adding credits
Go to Billing in the sidebar. You can:- Click a quick amount button (25, 100, 500)
- Enter a custom amount
Saved cards
Under Payment Methods, you can:- View all saved payment methods
- Mark a card as default (used for auto top-up)
- Add new cards
- Remove cards (keep at least one while auto top-up is active)
Auto top-up
Keep your balance funded automatically:- Enable Auto Top-Up on the billing page
- Set a threshold: when your balance falls below this amount, a top-up is triggered
- Set a top-up amount: how much to add each time
Transaction history
The billing page includes a full transaction log with:
Export your full transaction history as CSV using the export button. DeployGPU does not issue invoices. A receipt email and the CSV export are the record for expensing or reconciliation.
Refunds
A launch that fails during provisioning is refunded in full and automatically, regardless of the reason: the reserved first hour goes straight back to your balance. Stopping an instance before it reaches Running also refunds the whole reserved hour. Once an instance is Running, stopping it refunds the unused seconds of the current hour, settled the moment teardown starts. See the stop instance guide for a worked example.What happens when credits run out
Auto top-up, set up above, is the most reliable way to avoid both outcomes.