The two clocks
behind every
GPU rental quote.
When capacity starts and how long you commit are different decisions. A useful price comparison needs both.
Read the guide ↗THE COMPUTE BRIEF14 SEPTEMBER 2026
The two clocks
H100 COMMITMENT CURVESUSD / GPU-hour
Verda · $3.25 on-demand → $2.99 at 12 monthsTogether AI · $3.99 on-demand → $3.19 at 6 monthsHyperstack · on-demand onlyLambda · on-demand onlyHyperstack · reserved from $2.72 (term not published)
Off this scale: Microsoft Azure ND96isr H100 v5 (8x H100 SXM), per GPU: $12.29 on-demand, $7.87 at 1-year, $5.40 at 3-year, $4.92 at 5-year.
Equal spacing by category; connecting lines are guides, not prices for unlisted terms. Configurations, regions, networking and tax treatment differ; see the buyer guide before comparing.
A longer commitment and a later start are separate pricing decisions.
Understand before you commit.
All guides ↗What belongs in a GPU rental price comparison?
The fields that turn a list of prices into a comparison you can actually use.
A forward-start GPU quote is not a price forecast
Separate an offer available today from a prediction about what future capacity might cost.
A lower hourly GPU price can still cost more
A worked example of utilization, commitments and the true cost of useful compute.
THE PRICING LABSee the two time axes
See the two time axes
working together.
Explore published commitment discounts and see which future-start prices remain unavailable.
Explore the data ↗VERDA H100 PRICING CALCULATED
| Start / Term | 1 month | 3 months | 12 months |
|---|---|---|---|
| Public tariff* | $3.1850 | $3.1525 | $2.9900 |
| +3 months | — | — | — |
| +6 months | — | — | — |
USD / GPU-hour · *No guaranteed start date.
— = Not published in this dataset. Source · 14 Sep 2026
The conditions behind the number matter.
Contract terms, sources and comparison assumptions belong alongside the price. The Pricing Lab distinguishes published inputs, calculated rates and missing forward-start observations.
Read our editorial policy ↗