AWS (Amazon EC2)
Check the cited vendor page before you rent from AWS (Amazon EC2).
Amazon Web Services offers EC2 accelerated computing instances with NVIDIA GPUs, including P5 (H100), P5e/P5en (H200), P6 (B200/B300) and P4d (A100), billed per second on-demand, with Spot and Capacity Blocks.
- Address
- Online, Global, United States
- Website
- aws.amazon.com/
- GPU models
- B300, B200, H200, H100, A100, L40S
- Lowest on-demand H100 rate
- 688 US cents/GPU-h
- Lowest on-demand A100 rate
- 274 US cents/GPU-h
- Billing
- per second
- Compute regions
- US East (N. Virginia), US East (Ohio), US West (Oregon)
- Spot capacity
- Yes
- Kubernetes
- Yes
- Documentation
- docs.aws.amazon.com/ec2/
- Source note
- Facts from the vendor's own website (https://aws.amazon.com/ec2/pricing/on-demand/), fetched 2026-10-29.
Photos
Questions
Where is AWS (Amazon EC2)?
How do I contact AWS (Amazon EC2)?
More GPU cloud providers in Online
GPU neocloudsOnline
CoreWeave
AI cloud offering NVIDIA GPU instances (GB300/GB200 NVL72, B200, H200, H100, L40S, A100 and more) on demand and spot, plus managed Kubernetes (CKS), SUNK,…
- Lowest on-demand H100 rate
- 615 US cents/GPU-h
- Lowest on-demand A100 rate
- 270 US cents/GPU-h
Bare-metal GPU clustersOnline
GPU Mart
GPU hosting brand of Database Mart LLC offering dedicated GPU servers and GPU VPS (H100, A100, RTX Pro 6000, RTX 5090/4090 and others) from owned data centers…
GPU neocloudsOnline
Lambda
GPU cloud offering on-demand NVIDIA B200, H100, A100, GH200 and other GPU instances, 1-Click Clusters of 16 to 2,000+ GPUs, private cloud, and managed…
- Lowest on-demand H100 rate
- 329 US cents/GPU-h
- Lowest on-demand A100 rate
- 199 US cents/GPU-h
HyperscalersOnline
Microsoft Azure (GPU VMs)
Microsoft Azure offers GPU-optimized virtual machine families (NC, ND, NV) with NVIDIA GB300, GB200, H200, H100, A100, RTX PRO 6000, A10 and T4 GPUs and AMD…
- Lowest on-demand H100 rate
- 698 US cents/GPU-h
- Lowest on-demand A100 rate
- 340 US cents/GPU-h
GPU neocloudsOnline
Mithril (formerly Foundry / ML Foundry)
Self-serve GPU cloud from Foundry Technologies offering NVIDIA H100 and H200 instances via market-priced spot bids and flexible reservations, with API, CLI and…
GPU neocloudsOnline
Oblivus
Offers on-demand and reserved NVIDIA GPU instances for AI training, inference and rendering, with per-minute billing, an API and deployment in Montreal and…
- Lowest on-demand H100 rate
- 350 US cents/GPU-h
- Lowest on-demand A100 rate
- 147 US cents/GPU-h