LLM RAM Calculator → Hardware Guides

Cheapest H100 Rental Price — Comparison Table (2026)

H100 80GB SXM5 prices range from $2.49/hr (Vast.ai spot) to $4.99/hr (CoreWeave on-demand). H100 offers 2× the MFU of A100 for transformer workloads. This table is verified weekly.

Provider price comparison

ProviderSpot $/hrOn-demand $/hrNotes
Vast.ai$2.49$3.2Variable availability; spot can be hard to get
RunPod$2.89$3.49Good availability; secure cloud tier
Hyperstack$3.2European; NVLink clusters available
Lambda Labs$3.99Reliable; popular for training
CoreWeave$4.99Highest reliability; enterprise SLAs
Don't have the hardware? Rent it.

Verified GPU rental prices across all major providers. Start in minutes, pay by the hour.

Compare GPU prices →LLM RAM calculatorSelf-host vs API calculator

Frequently Asked Questions

Is the H100 worth 2× the A100 price for inference?

For inference specifically: the H100 offers about 1.5–2× the throughput (tokens/second) of an A100 at similar batch sizes, due to faster memory bandwidth and FP8 support. If your cost model is $/token rather than $/hr, the H100 is often comparable or slightly better value for large-batch inference.

Data verified 2026-08-20. GPU rental prices change frequently — verify before purchasing. See methodology.