GPU cloud comparison, one table

The same H100 SXM 80GB rented from $2.07/hr to $6.16/hr on 2026-11-11 depending on provider and tier; H200 from $2.90/hr; B200 from $6.00/hr. List prices converge whenever vendors resell the same upstream capacity — the real differences live in preemption, cold start and availability. rent.dev compares at dispatch time, not on a static page.

rent.dev — one endpoint for any compute an agent needs.

Questions

How do GPU cloud providers compare on price?

On-demand H100 SXM ranged from $2.07/hr to $6.16/hr across seven providers on 2026-11-11; the gap between tiers (spot, preemptible, marketplace) is larger than the gap between vendors' list prices.

Is spot cheaper than on-demand GPU cloud?

Usually yes: interruptible and marketplace tiers ran 40-60% below on-demand for the same chip, in exchange for reclaim risk that checkpointed jobs absorb.

Which GPU cloud has the best availability?

Availability is the hidden variable: no provider publishes live inventory, so the only honest comparison is at dispatch. rent.dev probes capacity before routing a job.