Hyperscaler · 13 GPU models tracked
Amazon Web Services
$0.53–$17.80per GPU-hour, the cheapest row to the dearest
Amazon EC2 on-demand rates for Linux instances in us-east-1 (N. Virginia). Most GPU instances carry several GPUs and are priced per instance; the rate here is that price divided by the GPU count. Spot rates are not included.
As of , Amazon Web Services lists 13 GPU models we track, from $0.53 per GPU-hour (NVIDIA T4, on-demand) to $17.80 (NVIDIA B300, on-demand).
| Website | aws.amazon.com |
|---|---|
| Pricing page | aws.amazon.com/ec2/pricing/on-demand/ |
| Regions and tiers | us-east-1 — as labelled in the rates below |
| Models tracked | 13 |
Price by GPU
As of , per GPU-hour: Amazon Web Services’s on-demand rate for each card it lists, a cheaper market it also sells, and where the on-demand rate stands among the providers we track.
| GPU | Amazon Web Services on-demand, per GPU-hour | Against the market (median rate card) |
|---|---|---|
| NVIDIA B300 | $17.80 us-east-1, 8-GPU node captured Deploy ↗ | 98% above the median $8.99 · 26 providers 25 of 26 list lower · lowest Deep Infra $5.89 · compare |
| NVIDIA B200 | $14.24 us-east-1, 8-GPU node captured Deploy ↗ | 105% above the median $6.94 · 40 providers 38 of 40 list lower · lowest Hostinger $4.50 from · compare |
| NVIDIA H200 | $7.91 us-east-1, 8-GPU node captured Deploy ↗ | 73% above the median $4.56 · 56 providers 47 of 56 list lower · lowest Dataoorts $2.29 · compare |
| NVIDIA H100 SXM | $6.88 us-east-1, 8-GPU node captured Deploy ↗ | 109% above the median $3.29 · 41 providers 36 of 41 list lower · lowest Cudo Compute $1.79 GPU only · compare |
| NVIDIA A100 80GB | $3.43 us-east-1, 8-GPU node captured Deploy ↗ | 72% above the median $1.99 · 63 providers 52 of 63 list lower · lowest Yotta Labs $0.92 · compare |
| NVIDIA RTX PRO 6000 Blackwell | $3.36 us-east-1 captured Deploy ↗ | 53% above the median $2.20 · 42 providers 35 of 42 list lower · lowest Hostinger $0.65 from · compare |
| NVIDIA V100 | $3.06 us-east-1 captured Deploy ↗ | 269% above the median $0.83 · 13 providers 11 of 13 list lower · lowest Verda $0.24 · compare |
| NVIDIA A100 40GB | $2.74 us-east-1, 8-GPU node captured Deploy ↗ | 38% above the median $1.99 · 22 providers 16 of 22 list lower · lowest Trooper.AI $0.69 from · compare |
| NVIDIA RTX PRO 4500 Blackwell | $2.52 us-east-1 captured Deploy ↗ | 193% above the median $0.86 · 6 providers 5 of 6 list lower · lowest GPU.ai $0.72 from · compare |
| NVIDIA L40S | $1.86 us-east-1 captured Deploy ↗ | 31% above the median $1.42 · 40 providers 27 of 40 list lower · lowest CloudRift $0.63 · compare |
| NVIDIA A10G | $1.01 us-east-1 captured Deploy ↗ | 11% below the median $1.13 · 4 providers the lowest on-demand rate of the 4 |
| NVIDIA L4 | $0.80 us-east-1 captured Deploy ↗ | 1% above the median $0.80 · 23 providers 15 of 23 list lower · lowest Hikube $0.28 GPU only · compare |
| NVIDIA T4 | $0.53 us-east-1 captured Deploy ↗ | 4% below the median $0.55 · 19 providers 6 of 19 list lower · lowest immers.cloud $0.24 · compare |
The same, in words
- Amazon Web Services lists the NVIDIA B300 from $17.80 per GPU-hour on-demand (us-east-1, 8-GPU node). The on-demand rate is unchanged since first read, Sep 25, 2026 (7-day change: none). Against the market: 98% above the median rate card ($8.99 across 26 providers); 25 of the 26 providers we track list a lower on-demand rate, the lowest Deep Infra at $5.89 (Deep Infra’s rates, Amazon Web Services vs Deep Infra).
- Amazon Web Services lists the NVIDIA B200 from $14.24 per GPU-hour on-demand (us-east-1, 8-GPU node). The on-demand rate is unchanged since first read, Sep 7, 2026 (7-day change: none). Against the market: 105% above the median rate card ($6.94 across 40 providers); 38 of the 40 providers we track list a lower on-demand rate, the lowest Hostinger at $4.50 (Hostinger’s rates, Amazon Web Services vs Hostinger).
- Amazon Web Services lists the NVIDIA H200 from $7.91 per GPU-hour on-demand (us-east-1, 8-GPU node). The on-demand rate is unchanged since first read, Sep 7, 2026 (7-day change: none). Against the market: 73% above the median rate card ($4.56 across 56 providers); 47 of the 56 providers we track list a lower on-demand rate, the lowest Dataoorts at $2.29 (Dataoorts’s rates, Amazon Web Services vs Dataoorts).
- Amazon Web Services lists the NVIDIA H100 SXM from $6.88 per GPU-hour on-demand (us-east-1, 8-GPU node). The on-demand rate is unchanged since first read, Sep 7, 2026 (7-day change: none). Against the market: 109% above the median rate card ($3.29 across 41 providers); 36 of the 41 providers we track list a lower on-demand rate, the lowest Cudo Compute at $1.79 (Cudo Compute’s rates, Amazon Web Services vs Cudo Compute).
- Amazon Web Services lists the NVIDIA A100 80GB from $3.43 per GPU-hour on-demand (us-east-1, 8-GPU node). The on-demand rate is unchanged since first read, Sep 7, 2026 (7-day change: none). Against the market: 72% above the median rate card ($1.99 across 63 providers); 52 of the 63 providers we track list a lower on-demand rate, the lowest Yotta Labs at $0.92 (Yotta Labs’s rates, Amazon Web Services vs Yotta Labs).
- Amazon Web Services lists the NVIDIA RTX PRO 6000 Blackwell from $3.36 per GPU-hour on-demand (us-east-1). The on-demand rate is unchanged since first read, Sep 27, 2026 (7-day change: none). Against the market: 53% above the median rate card ($2.20 across 42 providers); 35 of the 42 providers we track list a lower on-demand rate, the lowest Hostinger at $0.65 (Hostinger’s rates, Amazon Web Services vs Hostinger).
- Amazon Web Services lists the NVIDIA V100 from $3.06 per GPU-hour on-demand (us-east-1). The on-demand rate is unchanged since first read, Sep 23, 2026 (7-day change: none). Against the market: 269% above the median rate card ($0.83 across 13 providers); 11 of the 13 providers we track list a lower on-demand rate, the lowest Verda at $0.24 (Verda’s rates, Amazon Web Services vs Verda).
- Amazon Web Services lists the NVIDIA A100 40GB from $2.74 per GPU-hour on-demand (us-east-1, 8-GPU node). The on-demand rate is unchanged since first read, Sep 7, 2026 (7-day change: none). Against the market: 38% above the median rate card ($1.99 across 22 providers); 16 of the 22 providers we track list a lower on-demand rate, the lowest Trooper.AI at $0.69 (Trooper.AI’s rates, Amazon Web Services vs Trooper.AI).
- Amazon Web Services lists the NVIDIA RTX PRO 4500 Blackwell from $2.52 per GPU-hour on-demand (us-east-1). The on-demand rate is unchanged since first read, Sep 27, 2026 (7-day change: none). Against the market: 193% above the median rate card ($0.86 across 6 providers); 5 of the 6 providers we track list a lower on-demand rate, the lowest GPU.ai at $0.72 (GPU.ai’s rates, Amazon Web Services vs GPU.ai).
- Amazon Web Services lists the NVIDIA L40S from $1.86 per GPU-hour on-demand (us-east-1). The on-demand rate is unchanged since first read, Sep 7, 2026 (7-day change: none). Against the market: 31% above the median rate card ($1.42 across 40 providers); 27 of the 40 providers we track list a lower on-demand rate, the lowest CloudRift at $0.63 (CloudRift’s rates, Amazon Web Services vs CloudRift).
- Amazon Web Services lists the NVIDIA A10G from $1.01 per GPU-hour on-demand (us-east-1). The on-demand rate is unchanged since first read, Sep 27, 2026 (7-day change: none). Against the market: the lowest on-demand rate of the 4 providers we track for this card, 11% below the median rate card ($1.13 across 4 providers).
- Amazon Web Services lists the NVIDIA L4 from $0.80 per GPU-hour on-demand (us-east-1). The on-demand rate is unchanged since first read, Sep 23, 2026 (7-day change: none). Against the market: 1% above the median rate card ($0.80 across 23 providers); 15 of the 23 providers we track list a lower on-demand rate, the lowest Hikube at $0.28 (Hikube’s rates, Amazon Web Services vs Hikube).
- Amazon Web Services lists the NVIDIA T4 from $0.53 per GPU-hour on-demand (us-east-1). The on-demand rate is unchanged since first read, Sep 23, 2026 (7-day change: none). Against the market: 4% below the median rate card ($0.55 across 19 providers); 6 of the 19 providers we track list a lower on-demand rate, the lowest immers.cloud at $0.24 (immers.cloud’s rates, Amazon Web Services vs immers.cloud).
All listed prices
One table per price type, never mixed. Every rate normalized to one GPU for one hour; a price opens the page it was read from, and 7d marks a rate that moved in the past seven days. The day each figure was first listed is in this page’s data, below.
On-demand
| GPU · what is sold | $ / GPU-hour | Read | Source |
|---|---|---|---|
| NVIDIA T4APIus-east-1 | $0.53 | captured | Deploy ↗ |
| NVIDIA L4APIus-east-1 | $0.80 | captured | Deploy ↗ |
| NVIDIA A10GAPIus-east-1 | $1.01 | captured | Deploy ↗ |
| NVIDIA L40SAPIus-east-1 | $1.86 | captured | Deploy ↗ |
| NVIDIA RTX PRO 4500 BlackwellAPIus-east-1 | $2.52 | captured | Deploy ↗ |
| NVIDIA A100 40GBAPI$21.96/node ÷ 8 · us-east-1 | $2.74 | captured | Deploy ↗ |
| NVIDIA V100APIus-east-1 | $3.06 | captured | Deploy ↗ |
| NVIDIA RTX PRO 6000 BlackwellAPIus-east-1 | $3.36 | captured | Deploy ↗ |
| NVIDIA A100 80GBAPI$27.45/node ÷ 8 · us-east-1 | $3.43 | captured | Deploy ↗ |
| NVIDIA H100 SXMAPI$55.04/node ÷ 8 · us-east-1 | $6.88 | captured | Deploy ↗ |
| NVIDIA H200API$63.30/node ÷ 8 · us-east-1 | $7.91 | captured | Deploy ↗ |
| NVIDIA B200API$113.93/node ÷ 8 · us-east-1 | $14.24 | captured | Deploy ↗ |
| NVIDIA B300API$142.42/node ÷ 8 · us-east-1 | $17.80 | captured | Deploy ↗ |
Ask an AI agent about Amazon Web Services’s rates: Claude, Cursor, VS Code and ChatGPT can read them through the GPU Economy MCP server - free, no key, every answer citing the page and read time.
What else Amazon Web Services bills
The GPU-hour is not the whole bill. What Amazon Web Services's own pages say about the rest, read by hand and checked on ; each line links the page it was read from. "Not published" means we looked and the pages do not say. Every provider side by side.
| What | What Amazon Web Services states | Source |
|---|---|---|
| In the rate | Instance price covers vCPU, RAM and local NVMe instance store (p5.48xlarge: 192 vCPU, 2 TiB, 8x3.84 TB NVMe); EBS volumes billed separately our reading | source checked |
| Data out | 100 GB/month free across AWS, then $0.09/GB for the first 10 TB/month out to internet, US East (N. Virginia) | source checked |
| Storage | EBS gp3 $0.08/GB-month in US East (N. Virginia); gp2 $0.10 | source checked |
| While stopped | Stopped instance not billed for usage; attached EBS volumes (and Elastic IPs) keep billing | source checked |
| Billing step | Per second, one-minute minimum each time the instance starts (Linux) | source checked |
Every rate on this page as data: /providers/aws/prices.json — the same fields as prices.json, narrowed to this provider. CC BY 4.0, no key, no rate limit. On your own page: embed this table (/embed/providers/aws, an iframe, no script).
Compare Amazon Web Services with
Every provider that lists at least two of the same GPUs on-demand, most shared GPUs first; each opens the side-by-side page.
- EmpirioLabs 11 shared GPUs
- RunPod 11 shared GPUs
- Modal 10 shared GPUs
- GPU.ai 9 shared GPUs
- Oracle Cloud (OCI) 9 shared GPUs
- Vast.ai 9 shared GPUs
- Verda 9 shared GPUs
- Cerebrium 8 shared GPUs
- E2E Networks 8 shared GPUs
- Google Cloud 8 shared GPUs
- Sesterce 8 shared GPUs
- Hugging Face 7 shared GPUs
- Lium 7 shared GPUs
- Microsoft Azure 7 shared GPUs
- Civo 6 shared GPUs
- CoreWeave 6 shared GPUs
- HexGrid 6 shared GPUs
- Hyperstack 6 shared GPUs
- Jarvislabs 6 shared GPUs
- Koyeb 6 shared GPUs
- Lightning AI 6 shared GPUs
- Nebius 6 shared GPUs
- Northflank 6 shared GPUs
- Seeweb 6 shared GPUs
- Barrack AI 5 shared GPUs
- Baseten 5 shared GPUs
- Beam 5 shared GPUs
- Crusoe 5 shared GPUs
- Gcore 5 shared GPUs
- Geodd 5 shared GPUs
- Hikube 5 shared GPUs
- Lambda 5 shared GPUs
- Lyceum 5 shared GPUs
- Massed Compute 5 shared GPUs
- OpenRelay 5 shared GPUs
- Paperspace 5 shared GPUs
- Runware 5 shared GPUs
- Yotta Labs 5 shared GPUs
- airgpu 4 shared GPUs
- Daytona 4 shared GPUs
- Deep Infra 4 shared GPUs
- fal 4 shared GPUs
- FarmGPU 4 shared GPUs
- FPT AI Factory 4 shared GPUs
- Hostinger 4 shared GPUs
- Limestone Networks 4 shared GPUs
- Neysa 4 shared GPUs
- OVHcloud 4 shared GPUs
- packet.ai 4 shared GPUs
- Tata Communications 4 shared GPUs
- Together AI 4 shared GPUs
- Trooper.AI 4 shared GPUs
- Bitdeer AI 3 shared GPUs
- Cloud.ru Evolution 3 shared GPUs
- CloudGPU 3 shared GPUs
- CloudRift 3 shared GPUs
- Cudo Compute 3 shared GPUs
- Dataoorts 3 shared GPUs
- Denvr Dataworks 3 shared GPUs
- Elice Cloud 3 shared GPUs
- Fireworks AI 3 shared GPUs
- HPC-AI.COM 3 shared GPUs
- Hyperbolic 3 shared GPUs
- immers.cloud 3 shared GPUs
- Infomaniak 3 shared GPUs
- QuantaCloud 3 shared GPUs
- Replicate 3 shared GPUs
- Selectel 3 shared GPUs
- Servercore 3 shared GPUs
- STACKIT 3 shared GPUs
- SynpixCloud 3 shared GPUs
- VESSL AI 3 shared GPUs
- Airon 2 shared GPUs
- Atlas Cloud 2 shared GPUs
- Bamboozle Web Services 2 shared GPUs
- Canopy Wave 2 shared GPUs
- Cato Digital 2 shared GPUs
- centron 2 shared GPUs
- Clore.ai 2 shared GPUs
- Enverge 2 shared GPUs
- FlexAI 2 shared GPUs
- FluidCore 2 shared GPUs
- GMI Cloud 2 shared GPUs
- Inferless 2 shared GPUs
- Leafcloud 2 shared GPUs
- MimicPC 2 shared GPUs
- RunComfy 2 shared GPUs
- Taiga Cloud (Northern Data Group) 2 shared GPUs
- Thunder Compute 2 shared GPUs
- Vagon 2 shared GPUs
- Yandex Cloud 2 shared GPUs
Data updated: