A self-contained script that walks through the entire lifecycle: find the cheapest GPU offer, procure it, wait for it to be ready, and print SSH access details.

What this does

  1. Get offers — queries all vendors for H100 x1 under $5/GPU/hr
  2. Pick cheapest — selects the lowest-priced offer
  3. Procure — provisions the instance with your SSH key
  4. Poll — waits up to 30 minutes for the instance to be ready
  5. Clean up — code to delete the resource (commented out by default)

Customizing

Change the YAML spec to match your needs. See the YAML Spec guide for all available options: