GPU rental price per hour
Weekly median on-demand rental price per H100 GPU-hour across the cloud providers GetDeploying tracks; H200 and B200 alongside.
Reason through thisThe reading
The H100's median rent fell by about a sixth into the spring and has since climbed back above where it stood a year before; newer chips rent for more. No sign of the glut that follows an overbuild.
58 readings drawn from getdeploying.nvidia_h100.on_demand_median_usd_per_gpu_hour.w, one of 3 series listed under How it is measured; 174 observations in the evidence.
Source: GetDeploying GPU rental price history (weekly median on-demand price per GPU-hour). GetDeploying, GPU rental price history (CC BY 4.0).
The numbers
| as of | reading | value | record |
|---|---|---|---|
| 2026-10-26 | nvidia h100 | $3.85 | measured |
| 2026-10-19 | nvidia h100 | $3.99 | measured |
| 2026-10-12 | nvidia h100 | $3.80 | measured |
| 2026-10-05 | nvidia h100 | $3.41 | measured |
| 2026-09-28 | nvidia h100 | $3.32 | measured |
| 2026-09-21 | nvidia h100 | $3.23 | measured |
| 2026-09-14 | nvidia h100 | $3.15 | measured |
| 2026-09-07 | nvidia h100 | $3.20 | measured |
| 2026-08-31 | nvidia h100 | $3.14 | measured |
| 2026-08-24 | nvidia h100 | $3.20 | measured |
| 2026-08-17 | nvidia h100 | $3.25 | measured |
| 2026-08-10 | nvidia h100 | $3.29 | measured |
| 2026-08-03 | nvidia h100 | $3.25 | measured |
| 2026-07-27 | nvidia h100 | $3.11 | measured |
| 2026-07-20 | nvidia h100 | $2.99 | measured |
| 2026-07-13 | nvidia h100 | $2.89 | measured |
| 2026-07-06 | nvidia h100 | $2.99 | measured |
| 2026-06-29 | nvidia h100 | $2.87 | measured |
| 2026-06-22 | nvidia h100 | $2.73 | measured |
| 2026-06-15 | nvidia h100 | $2.80 | measured |
| 2026-06-08 | nvidia h100 | $2.76 | measured |
| 2026-06-01 | nvidia h100 | $2.94 | measured |
| 2026-05-25 | nvidia h100 | $2.80 | measured |
| 2026-05-18 | nvidia h100 | $2.86 | measured |
| 2026-05-11 | nvidia h100 | $2.64 | measured |
| 2026-05-04 | nvidia h100 | $2.59 | measured |
| 2026-04-27 | nvidia h100 | $2.50 | measured |
| 2026-04-20 | nvidia h100 | $2.59 | measured |
| 2026-04-13 | nvidia h100 | $2.50 | measured |
| 2026-04-06 | nvidia h100 | $2.49 | measured |
| 2026-03-30 | nvidia h100 | $2.50 | measured |
| 2026-03-23 | nvidia h100 | $2.49 | measured |
| 2026-03-16 | nvidia h100 | $2.49 | measured |
| 2026-03-09 | nvidia h100 | $2.49 | measured |
| 2026-03-02 | nvidia h100 | $2.49 | measured |
| 2026-02-23 | nvidia h100 | $2.53 | measured |
| 2026-02-16 | nvidia h100 | $2.87 | measured |
| 2026-02-09 | nvidia h100 | $2.94 | measured |
| 2026-02-02 | nvidia h100 | $2.94 | measured |
| 2026-01-26 | nvidia h100 | $2.89 | measured |
| 2026-01-19 | nvidia h100 | $2.89 | measured |
| 2026-01-12 | nvidia h100 | $2.69 | measured |
| 2026-01-05 | nvidia h100 | $2.73 | measured |
| 2025-12-29 | nvidia h100 | $2.73 | measured |
| 2025-12-22 | nvidia h100 | $2.81 | measured |
| 2025-12-15 | nvidia h100 | $2.81 | measured |
| 2025-12-08 | nvidia h100 | $2.89 | measured |
| 2025-12-01 | nvidia h100 | $2.89 | measured |
| 2025-11-24 | nvidia h100 | $2.89 | measured |
| 2025-11-17 | nvidia h100 | $2.95 | measured |
| 2025-11-10 | nvidia h100 | $2.95 | measured |
| 2025-11-03 | nvidia h100 | $2.98 | measured |
| 2025-10-27 | nvidia h100 | $2.98 | measured |
| 2025-10-20 | nvidia h100 | $2.98 | measured |
| 2025-10-13 | nvidia h100 | $2.98 | measured |
| 2025-10-06 | nvidia h100 | $2.98 | measured |
| 2025-09-29 | nvidia h100 | $2.95 | measured |
| 2025-09-22 | nvidia h100 | $2.95 | measured |
Why it matters
The cloud layer's price per unit of compute; falling rental prices are the neoclouds' rent eroding, and a crash would be the glut that bandwidth went through in 2001.
What cuts against it
List prices, not what contracted buyers pay; spot and reserved prices diverge from on-demand; one aggregator, and its panel of providers changes week to week; an older chip's rent falls as newer chips arrive, with no glut.
How it is measured
- series
getdeploying.nvidia_h100.on_demand_median_usd_per_gpu_hour.w - series
getdeploying.nvidia_h200.on_demand_median_usd_per_gpu_hour.w - series
getdeploying.nvidia_b200.on_demand_median_usd_per_gpu_hour.w - source GetDeploying GPU rental price history (weekly median on-demand price per GPU-hour) · default tier 3 · CC BY 4.0
- proxy price · cadence weekly
Eight weekly readings, about two months. A move of a quarter an hour or less is noise in a median across some forty providers whose panel changes week to week (a third of four-week moves in the first year exceeded fifteen cents); rising rents mean the owners of GPUs keep more of the value.
Applied to getdeploying.nvidia_h100.on_demand_median_usd_per_gpu_hour.w.
Latest readings
Why this status, and its history
First reading, from GetDeploying's weekly median of on-demand list prices across the cloud providers it tracks, read only once a week has closed. An H100 rented for $3.15 an hour in the week of 14 Sep 2026, against $2.99 eight weeks before (inside the quarter-dollar dead band, so stable) and $2.95 in its first week on record, 22 Sep 2025. One aggregator at tier 3 caps the status at emerging.
- 2026-09-25unmeasured to emergingconf — → 35 · Claude (model)
First reading, from GetDeploying's weekly median of on-demand list prices across the cloud providers it tracks, read only once a week has closed. An H100 rented for $3.15 an hour in the week of 14 Sep 2026, against $2.99 eight weeks before (inside the quarter-dollar dead band, so stable) and $2.95 in its first week on record, 22 Sep 2025. One aggregator at tier 3 caps the status at emerging.
Weighed against it: List prices, not contracted prices; reserved and spot prices sit well below on-demand, so a glut could show there first; the median fell by about a sixth into the spring before recovering.
Why this confidence
Rests on 174 observations from GetDeploying GPU rental price history (weekly median on-demand price per GPU-hour); best evidence tier 3, grade B; current.
Confidence is independent of status: 90–95 multiple strong independent sources; 70–89 good evidence, some ambiguity; 50–69 mixed or hard to operationalise; 0–49 limited or vague evidence.