Skip to main content
NEO GPU rents NVIDIA H200 instances through the /neo-gpus API group, billed per subscription cycle or on-demand (hourly). An instance uses exactly one billing mode.
Every create places a real paid order. On-demand instances accumulate hourly charges until deleted.

biznetgio_gpu_instance

For hourly billing, replace subscription with:
Exactly one of subscription or on_demand must be set - validation fails otherwise. Import with the numeric account id.
Changing product_id, select_os, or keypair_id after creation does nothing useful, and worse than baremetal’s version of this problem: Update only ever calls the API for rebuild_trigger and reserve_additional_hours_trigger, then re-reads the instance from the API and overwrites those three fields back to their real (unchanged) values before saving state. The result is a permanent, unresolvable plan diff - every future terraform plan shows the same change again, forever, because it’s never actually applied and never actually discarded. If you need a different package, OS, or keypair, destroy and recreate the resource.

biznetgio_gpu_keypair

Update is a no-op. Import with the keypair id.

Data sources

For real pricing and how to pick a specific package instead of trusting [0], see the Product Catalog page.

biznetgio_gpu_products

GPU catalog with per-product flavors. products items: product_id, name, description, category_name, and flavors (flavor_id, name).

biznetgio_gpu_console

Mints a console access session for an instance.
Returns url (sensitive). Side-effecting: every read may create a new session - do not use it in places evaluated during plan diffs.

biznetgio_gpu_graph

Monitoring graph data for an instance.
Returns graph and raw (sensitive).