NVIDIA H100 Lease Rates
Normalized pricing indices for NVIDIA H100 clusters, tracking the dislocation between Hyperscale and Tier 2 Cloud pricing.
What is the lease rate for NVIDIA H100 in 2026?
The average lease benchmark for NVIDIA H100 is $2.67 / GPU-hour, with lower-quartile pricing near $1.58 / GPU-hour and upper-quartile pricing near $3.82 / GPU-hour.
Key Data Points
- Spot Rate Low: $1.58 / GPU-hour
- Hyperscale High: $3.82 / GPU-hour
- Monthly Reserve: $1,657.51 / GPU-month
- Trend: Bearish (Blackwell Impact)
The "Legacy Premium" Analysis
Current data indicates a Price Dispersion of 142% for NVIDIA H100. While Hyperscalers (AWS, Azure, GCP) maintain rates near $3.82 / GPU-hour due to enterprise service layers, specialized "Neoclouds" are offering identical silicon near $1.58 / GPU-hour to fill capacity.
Regional Variance: Availability Zones
| Region | Cluster Latency | Power Cost Basis | Avg. Rate |
|---|---|---|---|
| US-East (Virginia) | ~20ms | $0.08/kWh | $2.94 |
| US-Central (Ohio/Texas) | ~35ms | $0.05/kWh | $2.54 |
| EU-West (London) | ~15ms | $0.18/kWh | $3.34 |
For inference workloads not requiring sub-millisecond latency, aiming for **US-Central** availability zones can yield a net cost reduction of 15% purely on lease rate arbitrage.
The 'Pricing Vacuum' Report
- Why NVIDIA H100 pricing is crashing 40% faster than A100s.
- The 'Shadow Banking' loops keeping Hyperscale rates high.
- Full list of Tier 2 providers with immediate availability.
Need format proof first? View sample output
Explore More
Related Tools
GPU Residual/LTV Calculator
Calculate GPU depreciation and residual values
Open Speed-to-Power Watchlist