Azure GPU pricing 2026
Azure exposes GPUs through ND (training, InfiniBand), NC (inference and training), and NV (visualisation) families. ND H100 v5 is the headline 8x H100 SKU with InfiniBand.
Published rate card
Per-GPU per-hour rates pulled from the vendor pricing page linked above, Last verified June 2026. Rates exclude storage, egress, and any managed-service uplift. Reserved-capacity contracts typically improve on these rates.
| GPU | Configuration | Per GPU per hour |
|---|---|---|
| H100 80GB (ND96isr H100 v5) | On-demand, per-GPU equiv | $12.290 |
| H100 80GB (ND96isr H100 v5) | Spot, per-GPU equiv | $9.000 |
| A100 80GB (ND96amsr A100 v4) | On-demand, per-GPU equiv | $4.780 |
| A100 40GB (NC A100 v4) | On-demand | $3.400 |
| T4 (NCasT4_v3) | On-demand | $0.526 |
Hidden costs to watch
- Quota approval is the rate-limiting step for H100 SKUs; the public retail price is the floor, not the ceiling, for what teams pay before discount.
- Managed Disks, Azure NetApp Files, and bandwidth out of Azure billed separately.
- Reserved Instances (1 or 3 year) and Savings Plans are the practical path to predictable cost.
- Spot VMs use eviction-based pricing; production training jobs need checkpointing discipline.
What Azure is best for
Microsoft-anchored estates that need GPU compute alongside Entra ID, Azure AI Foundry, and Synapse data infrastructure.
See the GPU cloud buying guideWorked example
Acme Vision Co. (illustrative example, not a real company) needs to train a 10 billion-parameter vision-language model on a fixed 8-GPU cluster for 30 days at 18 hours per day duty cycle.
At Azure's cheapest published rate of $9.00 per GPU-hour, the run costs $38,880 for raw GPU compute, before storage, egress, and MLOps overhead. Add the typical 25 percent year-one uplift and the modelled spend is $48,600. Use the calculator on the homepage to model your own GPU class, cluster size, and duty cycle.
Last verified June 2026. Azure rates change frequently. Always obtain a vendor quote before purchase.
Visit Azure pricing page