Amazon Web Services vs Google Cloud
As of , Amazon Web Services lists the lower on-demand rate on 5 of the 8 GPUs both offer on-demand; Google Cloud wins on 3. On the priciest shared card both list on-demand, NVIDIA B200: Amazon Web Services $14.24 (us-east-1) vs Google Cloud $16.11 (us-west1) per GPU-hour on-demand — Amazon Web Services is $1.87 (12%) cheaper.
On-demand, per GPU-hour
| GPU |  Amazon Web Services |  Google Cloud | Difference | Cheaper |
|---|---|---|---|---|
| NVIDIA B200 median $6.94 · 40 providers · lowest Hostinger $4.50 from | $14.24 us-east-1 captured Deploy ↗ | $16.11 us-west1 · GPU only captured Deploy ↗ | $1.87 (12%) | Amazon Web Services |
| NVIDIA H200 median $4.56 · 56 providers · lowest Dataoorts $2.29 | $7.91 us-east-1 captured Deploy ↗ | $9.31 us-west1 · GPU only captured Deploy ↗ | $1.40 (15%) | Amazon Web Services |
| NVIDIA H100 SXM median $3.29 · 41 providers · lowest Cudo Compute $1.79 GPU only | $6.88 us-east-1 captured Deploy ↗ | $9.80 us-west8 · GPU only captured Deploy ↗ | $2.92 (30%) | Amazon Web Services |
| NVIDIA A100 80GB median $1.99 · 63 providers · lowest Yotta Labs $0.92 | $3.43 us-east-1 captured Deploy ↗ | $3.93 us-central1 · GPU only captured Deploy ↗ | $0.50 (13%) | Amazon Web Services |
| NVIDIA V100 median $0.83 · 13 providers · lowest Verda $0.24 | $3.06 us-east-1 captured Deploy ↗ | $2.48 us-west4 · GPU only captured Deploy ↗ | $0.58 (19%) | Google Cloud |
| NVIDIA A100 40GB median $1.99 · 22 providers · lowest Trooper.AI $0.69 from | $2.74 us-east-1 captured Deploy ↗ | $2.93 us-central1 · GPU only captured Deploy ↗ | $0.19 (6%) | Amazon Web Services |
| NVIDIA L4 median $0.80 · 23 providers · lowest Hikube $0.28 GPU only | $0.80 us-east-1 captured Deploy ↗ | $0.56 us-east4 · GPU only captured Deploy ↗ | $0.25 (31%) | Google Cloud |
| NVIDIA T4 median $0.55 · 19 providers · lowest immers.cloud $0.24 | $0.53 us-east-1 captured Deploy ↗ | $0.35 us-east5 · GPU only captured Deploy ↗ | $0.18 (33%) | Google Cloud |
Spot / interruptible, per GPU-hour
| GPU | Amazon Web Services | Google Cloud |
|---|---|---|
| NVIDIA B200 | — | $1.45 captured |
| NVIDIA H200 | — | $5.27 captured |
| NVIDIA H100 SXM | — | $1.06 captured |
| NVIDIA A100 80GB | — | $0.40 captured |
| NVIDIA A100 40GB | — | $0.29 captured |
How this is compared
Cloud GPU rental prices of Amazon Web Services and Google Cloud, compared per GPU-hour on the GPUs both list. Where a provider has several tiers or regions, its cheapest on-demand rate is used and labelled. Community capacity (RunPod Community Cloud) and marketplace floors are shown only where a provider has nothing else, and never decide which is cheaper.
GPU by GPU
- NVIDIA B200: Amazon Web Services $14.24 (us-east-1) vs Google Cloud $16.11 (us-west1) per GPU-hour on-demand — Amazon Web Services is $1.87 (12%) cheaper.
- NVIDIA H200: Amazon Web Services $7.91 (us-east-1) vs Google Cloud $9.31 (us-west1) per GPU-hour on-demand — Amazon Web Services is $1.40 (15%) cheaper.
- NVIDIA H100 SXM: Amazon Web Services $6.88 (us-east-1) vs Google Cloud $9.80 (us-west8) per GPU-hour on-demand — Amazon Web Services is $2.92 (30%) cheaper.
- NVIDIA A100 80GB: Amazon Web Services $3.43 (us-east-1) vs Google Cloud $3.93 (us-central1) per GPU-hour on-demand — Amazon Web Services is $0.50 (13%) cheaper.
- NVIDIA V100: Amazon Web Services $3.06 (us-east-1) vs Google Cloud $2.48 (us-west4) per GPU-hour on-demand — Google Cloud is $0.58 (19%) cheaper.
- NVIDIA A100 40GB: Amazon Web Services $2.74 (us-east-1) vs Google Cloud $2.93 (us-central1) per GPU-hour on-demand — Amazon Web Services is $0.19 (6%) cheaper.
- NVIDIA L4: Amazon Web Services $0.80 (us-east-1) vs Google Cloud $0.56 (us-east4) per GPU-hour on-demand — Google Cloud is $0.25 (31%) cheaper.
- NVIDIA T4: Amazon Web Services $0.53 (us-east-1) vs Google Cloud $0.35 (us-east5) per GPU-hour on-demand — Google Cloud is $0.18 (33%) cheaper.
Not compared: GPUs only one of the two lists, reserved/committed tiers, and anything either provider prices only on request. See each provider's page for its full list.
Data updated: