CoreWeave GPU pricing 2026
Founded 2017, originally a crypto-mining cluster operator, now a top-tier Nvidia partner with H100, H200, GB200, and B200 clusters and an InfiniBand-first architecture.
Published rate card
Per-GPU per-hour rates pulled from the vendor pricing page linked above, Last verified June 2026. Rates exclude storage, egress, and any managed-service uplift. Reserved-capacity contracts typically improve on these rates.
| GPU | Configuration | Per GPU per hour |
|---|---|---|
| H100 SXM | HGX H100 8x node, per-GPU | $6.155 |
| H100 PCIe | Per-GPU | $4.760 |
| A100 80GB SXM | HGX A100 8x node, per-GPU | $2.210 |
| A100 80GB PCIe | Per-GPU | $2.210 |
| L40 | Per-GPU | $1.650 |
| A40 | Per-GPU | $1.280 |
Hidden costs to watch
- InfiniBand interconnect is included on HGX nodes but egress to non-peered networks is metered.
- Object storage, distributed file system, and persistent volumes billed separately on top of GPU compute.
- Reserved-capacity contracts (1-3 years) are the practical entry point for large clusters; on-demand is capacity-permitting.
What CoreWeave is best for
Large H100 / H200 / GB200 clusters for foundation-model training and high-utilisation inference.
See the GPU cloud buying guideWorked example
Acme Vision Co. (illustrative example, not a real company) needs to train a 10 billion-parameter vision-language model on a fixed 8-GPU cluster for 30 days at 18 hours per day duty cycle.
At CoreWeave's cheapest published rate of $6.16 per GPU-hour, the run costs $26,590 for raw GPU compute, before storage, egress, and MLOps overhead. Add the typical 25 percent year-one uplift and the modelled spend is $33,237. Use the calculator on the homepage to model your own GPU class, cluster size, and duty cycle.
Last verified June 2026. CoreWeave rates change frequently. Always obtain a vendor quote before purchase.
Visit CoreWeave pricing page