Hyperscaler
Microsoft Azure
OpenAI's landlord. Big H100/H200 and MI300X fleets with enterprise Microsoft integration.
Azure operates some of the largest AI clusters on earth for OpenAI, and sells ND-series (NVIDIA) and the industry's most visible MI300X deployment (ND MI300X v5). Pricing is hyperscaler-typical; the MI300X instances are the interesting value play for memory-bound inference.
GPU pricing on Azure
On-demand list per GPU-hour, last reviewed 2026-09. Methodology.
| GPU | $/GPU-hr | Notes |
|---|---|---|
| A100 80GB | $3.67 | ND A100 v4 ÷ 8 |
| MI300X | $5.42 | ND MI300X v5 ÷ 8 |
| H100 SXM | $5.70 | ND H100 v5 ÷ 8 |
Strengths
- MI300X at scale
- InfiniBand-backed ND clusters
- Microsoft enterprise stack
Watch for
- Quota friction
- Complex SKU naming