Can you rent a GPU by the minute?
Check the provider’s published billing increment and minimum duration. A price expressed per hour can still be charged in smaller increments; the price unit alone does not establish the billing unit.
In our latest check: 8 providers we track sell serverless GPUs that bill only for running time, across 17 GPU models. The cheapest serverless H100 works out at $2.50 per busy hour (Koyeb), against $1.80 for the cheapest on-demand H100 you keep running (HyperAI).
For a billing increment in seconds, billable runtime is actual runtime rounded up to the next increment. The job cost calculator makes this rounding explicit.
Related questions
- What is the difference between serverless GPU and instance pricing?
- What is a GPU hour?
- Is a GPU cloud deposit part of the rental price?
Numbers on this page come from our verified snapshot of 21 November 2026. Full table on the homepage; method in the methodology.