Auto Provision works on Verda, Hyperstack and QuantaCloud. Orders deploy plain GPU VMs: app templates and volumes cannot be added, and orders are placed and managed in the dashboard only, not through the API.
Place an order
The page has two tabs, New order and My orders. A new order is four sections on one screen.1
Choose a GPU
Pick a GPU family, the model if the family has more than one, and the number of GPUs: 1, 2, 4 or 8.
2
Providers and machine types
Tick the machine types you accept, up to 20, across Verda, Hyperstack and QuantaCloud. Each provider shows how its stock is watched: “Live · checked every second” for Verda and Hyperstack, and “Live · checked every minute” for QuantaCloud. If the checks fall behind, the badge reads “Stock checks delayed”. Verda’s confidential-computing machine types start unticked.
3
Order settings
Choose an OS image for each provider (the default is an image with CUDA and Docker), at least one SSH key, and the disk per machine: 100, 200, 500, 1,000 or 2,000 GB, default 100. Max price per hour is for the whole machine, not per GPU, with fees included, from $0.01 to $10,000. It starts at the highest price among the machine types you ticked, rounded up to a whole dollar. Start and End are optional and use your time zone. An empty Start means now, and an empty End means the order waits until it is filled or cancelled.
4
Order summary
Check the summary and click Start auto-provision. The order appears under My orders.
If the GPU is in stock now
The page shows “In stock now” with a Deploy now button for each machine that matches. Deploy now uses the order settings on the page. If someone else gets the machine first, the order keeps waiting for the next one. Deploy now cannot be combined with a start time in the future: clear the start time first.If it is not in stock
The page shows “Not in stock right now” and, for each machine type, the price now, the 30-day price range, how often it came into stock, how long it usually lasts and when it was last in stock, followed by the range you can expect to pay per hour.When a machine comes into stock
DeployGPU deploys the machine when its price, fees included, is at or below your max price and your balance covers one hour at that price. The machine is then an ordinary deployment: it shows under Deployments, bills like any other instance, and you connect to it with the SSH keys on the order.- If the provider runs out before the create goes through, nothing is charged and the order keeps waiting.
- If your balance is short, that machine is skipped, the order keeps waiting and you get an email, at most once a day per order.
- A price above your max, or a provider in maintenance, also leaves the order waiting.
Order states
Paused orders
An order pauses when:- One of its SSH keys is deleted.
- A deploy does not finish within about 35 minutes, or the machine stops before it is ready.
- The last machine type on the order has no OS image left.
Cancel an order
Cancel stops the order from waiting for stock, and it cannot be undone. If the order is Deploying, cancel only asks to stop it: a machine that was already deployed keeps running, and you stop it from Deployments.Billing
- You do not need a balance to place an order. The order summary warns you when your balance is below one hour at your max price.
- Waiting is free.
- When a machine is deployed, the first hour is reserved from your balance and the machine is billed like any other deployment. See Billing.
- Each open order can deploy its own machine, and each machine is billed separately.
Emails
No email is sent when you place or cancel an order.
Limits
- 5 open orders per account. Waiting, deploying and paused orders count.
- One machine per order.
- Verda, Hyperstack and QuantaCloud only. Massed Compute and Vast.ai are not part of Auto Provision.
- No app templates and no volumes on orders.
- Dashboard only. API keys are refused on the Auto Provision routes.