AWS vs CoreWeave: GPU pricing
Public on-demand list prices per GPU-hour, side by side. Updated 2026-06-10 — see methodology.
At a glance
CoreWeave is cheaper on 4 of 4 GPU types both providers list.
| GPU | AWS | CoreWeave | Cheaper |
|---|---|---|---|
| H200 p5en.48xlarge · HGX H200 | $7.91 | $6.31 | CoreWeave by 20% |
| H100 p5.48xlarge · HGX H100 | $6.88 | $6.16 | CoreWeave by 10% |
| A100 80GB p4de.24xlarge · HGX A100 | $3.43 | $2.70 | CoreWeave by 21% |
| L40S g6e.48xlarge · L40S | $3.77 | $2.25 | CoreWeave by 40% |
On-demand list prices per single GPU per hour; committed/reserved rates are lower on both providers. Verify with each provider before purchasing.
FAQ
Is AWS or CoreWeave cheaper for GPUs?+
CoreWeave lists lower public on-demand prices on 4 of the 4 GPU types both providers offer (as of 2026-06-10). Actual costs depend on commitment terms and negotiated rates.
What's the difference between AWS and CoreWeave?+
AWS is a hyperscaler with a broad managed-services ecosystem; CoreWeave is a GPU-focused neocloud with aggressive pricing. Hyperscalers bundle global regions and enterprise SLAs; neoclouds typically undercut them on raw GPU compute.