Skip to main content

Billing

A bill has exactly three lines, itemised on each instance's Billing tab.

Billing detail

1. Compute

Metered per second, priced per hour.

  • Starts when the instance reaches Running
  • Ends when it is stopped or destroyed
  • Ninety minutes costs 1.5 hours — no rounding up to whole hours

The rate is locked at order time and holds until that instance is destroyed. A rising market never follows a running instance.

2. Storage

Priced per GB-month of disk.

Three things catch people out:

  1. It starts the moment the instance is created — including the minutes spent pulling the image
  2. It continues after you stop the instance — the disk is still allocated
  3. Only destroying ends it — there is no other way

:::warning A concrete number A 16GB instance at typical rates costs a few tens of cents per stopped day. Small — but a stopped month is several dollars for something you never used.

Destroy what you have finished with. Stopping only pays off if you are genuinely coming back soon. :::

3. Network

Ingress and egress metered separately, per GB.

Everyday training, inference and weight downloads are normally a small share of the bill. The one to watch is bulk export — estimate before moving hundreds of GB out.

Nothing else

  • No instance creation fee
  • No port fees
  • No image licensing fees
  • No minimum spend, no minimum rental period

When the balance runs out

An ordered sequence:

StageWhat happens
Balance lowWarning email telling you how long the current burn rate lasts
Balance exhaustedInstances stop automatically — not destroyed; data intact
While stoppedStorage keeps billing and the balance can go negative
Negative past the set periodInstances are destroyed and the disk released

Topping up at any point restores everything.

:::tip The dashboard shows "runway at current usage" It projects from your rate right now. Worth a glance after starting something new. :::

Multi-GPU machines

Multi-GPU nodes rent whole. The market shows a per-GPU hourly rate; the charge is that rate times the GPU count. The order drawer's estimate gives the final figure.

Disputes

The figures on the Billing tab come from the platform's metering and are reconciled line by line against upstream. If a charge looks wrong:

  1. Note the instance number and the time range
  2. Send them to Telegram support

We can walk through compute, storage and network item by item.