If you actually control the machines (own them or colo):
List them on Vast.ai or RunPod.
H100s usually pull $2–4/hr, A100s a bit less, B200s higher when demand is there.
Just keep them online and priced competitively.
If these are just cloud credits (Lambda, CoreWeave, etc.):
You can’t really “sell” the credits themselves, most providers ban that.
Better to use the GPUs yourself and charge people for the actual work:
Fine-tuning models
Running inference APIs
Batch jobs for other people
That’s where the real money is with high end cards right now. Renting raw hours only works well if the hardware is under your control
This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
By continuing to use this site, you are consenting to our use of cookies.