Open Cloud Console

Understand infrastructure estimates

Read machine, GPU, and disk estimates before launch and distinguish live estimates from reconciled charges.

Pneum.ai shows an infrastructure estimate before launch and for eligible existing agents. It helps compare configurations; it is not the final provider bill.

Estimate components

Component When it bills
Machine While the provider instance is running.
GPU While an attached GPU instance is running.
Boot disk While the disk exists, including when the VM is stopped.

The total hourly estimate combines those components. The panel also identifies the machine, region, zone when fixed, image, disk size, and pricing source.

Pricing sources

The source can be a live provider catalog, configured pricing, or a configured fallback. A fallback is useful when a provider catalog is temporarily unavailable but may be less precise than a current regional lookup.

Location, operating-system licensing, sustained-use rules, egress, taxes, credits, and delayed provider corrections can cause actual cost to differ.

Monthly planning

Convert an hourly running estimate into a planning amount with the expected runtime window:

text
estimated monthly runtime = hourly total × expected running hours

Add the subscription and expected model spend separately. Do not multiply disk cost only by running hours if you plan to retain a stopped agent all month.

Stopped-agent estimate

The stop confirmation isolates the estimated boot-disk cost that remains after compute shuts down. Use it to decide whether retaining the environment is worth more than exporting work and terminating it.

Estimate versus reconciled cost

The billing page distinguishes current estimates from provider-exported, reconciled usage. Provider exports arrive later and can be corrected. Reconciled values are the stronger evidence for invoice totals.

If an estimate fails to load, confirm that the provider, location, and machine selection are valid. Then see billing troubleshooting.

Was this guide helpful?

Your response helps us keep product instructions useful.