The GPUs on DeployGPU come from five providers. Four of them, QuantaCloud, Massed Compute, Verda and Hyperstack, supply full virtual machines that you launch from the dashboard or the API. The fifth, Vast.ai, supplies containers in the dashboard only, labelled Testing Only.
One account and one balance for all 5 providers
DeployGPU holds the account with each provider and launches machines on it for you. You never connect your own cloud account, and every machine is paid from the same prepaid balance at the hourly price shown, which includes any platform fee. The offers come from the same scrapers that feed GpuPerHour, which check every provider DeployGPU can launch on every minute. QuantaCloud is run by Quanta Cloud LLC, the company behind DeployGPU, as the about page explains.
The 4 VM providers side by side
The differences that matter day to day are the login, the disk and whether data can outlive the machine:
| Provider | SSH user | Disk | Volumes | Worth knowing |
|---|---|---|---|---|
| QuantaCloud | ubuntu | Fixed per offer | No | RSA or Ed25519 keys only. Run by the same company as DeployGPU |
| Massed Compute | Ubuntu, capital U | Fixed per offer | No | US offers. On demand only |
| Verda | root | You choose, up to 500 GB | Yes | Finland. Power off and on through the API, with billing continuing |
| Hyperstack | ubuntu | Fixed per offer | Yes | All inbound ports open. SSH can lag behind Running. Power off and on through the API, with billing continuing |
All four run the same app templates (JupyterLab with PyTorch, ComfyUI, Open WebUI with Ollama, and a plain GPU VM), use port 22 for SSH, and come in sizes of 1, 2, 4 or 8 GPUs per VM. Stopping deletes the VM and its disk on every one of them.
Vast.ai containers in the dashboard
Vast.ai offers appear in the dashboard marketplace labelled Testing Only. They are containers rather than VMs, reached through a proxy SSH port shown on the deployment, and they cannot be launched through the API or with the app templates. Vast.ai bandwidth is billed separately from the hourly price. GPU VMs vs container pods explains what changes when you rent a container instead of a VM.
Which provider has which GPU right now
This table counts the launchable VM offers per GPU family and names the providers that have them, refreshed every 15 minutes. Stock moves through the day, so the provider pages below and the marketplace are the place to check before a launch. To compare these providers with the rest of the market, see all providers on GpuPerHour.
| GPU | Memory | Offers now | GPUs per VM | Providers | From (per GPU-hr) | Details |
|---|---|---|---|---|---|---|
| B200 | 192 GB | 1 | 8x | Verda | $8.03 | Offers |
| H200 | 141 GB | 1 | 4x | Verda | $5.44 | Offers |
| H100 | 80 / 94 GB | 8 | 1x, 2x, 4x, 8x | Hyperstack, Massed Compute, QuantaCloud | $2.50 | Offers |
| A100 | 80 GB | 14 | 1x, 2x, 8x | Hyperstack, Massed Compute, QuantaCloud | $1.35 | Offers |
| RTX PRO 6000 Blackwell | 96 GB | 6 | 1x, 8x | Massed Compute, QuantaCloud, Verda | $2.19 | Offers |
| L40S | 48 GB | 8 | 1x, 2x, 4x | Massed Compute, QuantaCloud | $0.97 | Offers |
| L40 | 48 GB | 13 | 1x, 2x, 4x | Hyperstack, Massed Compute, QuantaCloud | $0.86 | Offers |
| RTX 6000 Ada | 48 GB | 12 | 1x, 2x, 4x | Massed Compute, QuantaCloud | $0.78 | Offers |
| RTX A6000 | 48 GB | 51 | 1x, 2x, 4x, 8x | Hyperstack, Massed Compute, QuantaCloud | $0.48 | Offers |