Pause & Resume
Pause a running GPU box to stop paying for it without tearing it down, then resume it later — on the same provider, or a different one — with your workspace restored.
Provider coverage
The full round trip — pause, then resume, on the same provider or a different one — works on DataCrunch, Hyperstack and VoltagePark. On RunPod, pause and resume both work, except resume fails for good if the original host has no free GPUs. On Vast.ai and SimplePod, resume doesn't work at all — pause is fine, but there's no path back, so don't pause a box there if you need it back. Akash, Hot Aisle, Vultr and Massed Compute offer neither. See /providers/capabilities for the full, current table.
What pause actually does
Hitting Pause on a deployment does three things in order: it snapshots the workspace directory the box is running, confirms the snapshot succeeded, then stops billing and closes the lease with the provider. The instance itself goes away — pause is not a suspended, still-reserved VM sitting there waiting for you. What survives is the snapshot.
That ordering is deliberate: billing only stops after the data is safely captured, so a snapshot failure blocks the pause instead of silently losing the workspace along with the box.
What resume actually does
Resume is not un-pausing the same machine — there is no machine left to un-pause. It rents a fresh instance, matched to the same GPU model and count as the one you paused (or a specific offer you pick yourself, including one on a different provider), and restores your snapshot onto it. If the resume target has a template attached, that service is relaunched too, so you land back on a running ComfyUI or Jupyter instance rather than just a restored disk.
This is why a paused deployment is not billed while it sits paused: there is no box behind it. And it's why resuming onto a different provider works at all — the mechanism is identical to a manual snapshot-and-restore, just wired to one button.
Where the round trip actually works
Pause runs the same way on every provider we support, because snapshotting a workspace and closing a lease don't depend on what happens next. Resume is the part that varies — it needs to rent a fresh matching box and restore onto it, and that depends on the provider.
The full pause-then-resume round trip works end to end on DataCrunch, Hyperstack and VoltagePark. On RunPod, pause and resume both work, with one terminal failure mode: if the original host has no free GPUs when you resume, that resume fails for good. On Vast.ai and SimplePod, resume doesn't work at all — pause is fine, but there is no path back to the same box, so pausing there means you're redeploying from your snapshot, not resuming. Akash, Hot Aisle, Vultr and Massed Compute offer neither today. Check the provider before you pause something you're counting on getting back, not after — the provider capability matrix has the full table, kept current independently of this page.
When you'd use this instead of a snapshot backup
If you just want to stop paying for a box you'll get back to later, pause is the one-click version of what you'd otherwise do by hand: take a snapshot, close the deployment, and restore onto a new one when you're ready. Reach for a plain workspace backup instead when you want to keep working on the current box while also holding a restore point, or when you're moving to a provider outside the three the round trip supports and are driving the snapshot-and-restore steps yourself.
FAQ
Does a paused deployment keep billing?
No. Pause closes the provider lease along with taking the snapshot, so there is no running instance behind a paused deployment and nothing to bill for while it's paused.
Can I resume onto a different GPU or provider?
Yes, where the round trip is supported. Resume rents a fresh instance rather than restarting the paused one, so you can point it at a different offer — including a different provider — and it restores your snapshot onto that new box.
Which providers support pause and resume?
The full round trip works on DataCrunch, Hyperstack and VoltagePark. RunPod supports both, but resume has one terminal failure mode: if the original host has no free GPUs when you resume, it fails for good. Vast.ai and SimplePod support pause only — resume doesn't work on either, so pausing there means redeploying from your snapshot rather than getting the same box back. Akash, Hot Aisle, Vultr and Massed Compute support neither. See the provider capability matrix for the current, authoritative table.
What happens if I pause on a provider where resume can fail or doesn't work?
The pause itself still succeeds — your workspace is snapshotted and billing stops. On RunPod, resume can fail if the original host has run out of free GPUs. On Vast.ai and SimplePod, resume doesn't work at all, so plan on redeploying from your snapshot rather than getting the same box back. Take a workspace backup instead if you're on one of those providers and expect to come back to the same setup.
Is resume the same as un-pausing the same machine?
No. The original instance is gone once paused — resume rents a new one and restores your snapshot onto it, which is also what makes cross-provider resume possible in the first place.
