AWS vs CoreWeave
AWS and CoreWeave split the 2 GPU configurations both list with on-demand pricing: each is cheaper on 1, checked 13 October 2026. The biggest single gap is the H200 (SXM) at $6.30 on CoreWeave vs $7.91. Platform facts come from each provider's documentation.
20.3% cheaper at CoreWeave.
checked 13 October 2026. Price alone does not establish equivalent capacity or performance.
CoreWeave
cheaper on 1 of 2 sharedPrices are per GPU-hour for the same GPU variant. Published node sizes are shown with their rates; whole-node costs differ when node sizes differ. These are listed rates, not a guarantee of stock.
Price per GPU, side by side (13 October 2026)
| GPU (on-demand $/hr) | AWS | CoreWeave | Price diff |
|---|---|---|---|
| L40S | $1.86single GPU | $2.258-GPU node | -$0.39 (-17.3%) AWS |
| H200 (SXM) | $7.918-GPU node | $6.308-GPU node | -$1.61 (-20.3%) CoreWeave |
Cheapest on-demand rate; the cheaper one is highlighted. Price diff is the per-GPU-hour saving on the cheaper side (named), relative to the higher price; – means one side has no comparable recorded rate in our data, which is not proof that it does not offer the GPU. Spot/marketplace tiers shown on the individual pages.
Only AWS lists these GPUs in our data
| GPU | AWS, $/GPU-hour |
|---|---|
| L4 | $0.80single GPU |
| A100 SXM 40GB | $2.748-GPU node |
| A100 SXM 80GB | $3.438-GPU node |
| H100 SXM | $6.888-GPU node |
Only CoreWeave lists these GPUs in our data
| GPU | CoreWeave, $/GPU-hour |
|---|---|
| L40 | $1.258-GPU node |
| RTX PRO 6000 Server | $2.508-GPU node |
| A100 | $2.708-GPU node |
| H100 | $6.168-GPU node |
| GH200 | $6.50single GPU |
| B200 | $8.608-GPU node |
Platform facts
| AWS | CoreWeave | |
|---|---|---|
| Billing increment | per-second | not published |
| Egress | $0.09/GB first 10TB after free tier | free |
| Storage /GB-mo | $0.08/GB-month gp3 example region | $0.070/GB/month distributed file storage; $0.06/GB/month hot object storage |
| Spot offered | yes | not published |
| Reserved offered | yes self-serve | sales only |
| Notebook support | one-click Jupyter | templates |
From each provider's documentation, checked on 7 July 2026 unless a later date is shown; sources on the fees matrix and enterprise matrix.
FAQ
Each is cheaper on 1 of the 2 matching GPU configurations; 0 have equal rates, checked 13 October 2026. Compare the configuration you need in the table above.
The H200 (SXM): $6.30 per GPU-hour on CoreWeave vs $7.91 on AWS, a difference of $1.61 per hour, checked 13 October 2026.
We have an on-demand rate for 4 GPUs at AWS with no comparable recorded rate at CoreWeave: L4, A100 SXM 40GB, A100 SXM 80GB, H100 SXM. At CoreWeave, 6 with no comparable recorded rate at AWS: L40, RTX PRO 6000 Server, A100, H100, GH200, B200. A missing rate means our data has none, not that the provider cannot supply it.
Compare more rental options: the screener · price movers · methodology.