H100 rental price per hour

NVIDIA H100 rental prices per GPU-hour from suppliers that publish them, collected automatically every day. Last source check 29 Oct 2026 (UTC).

SOURCES CHECKED29 Oct 202646 days of history

Which cloud publishes the cheapest NVIDIA H100? The table below lists every supplier we track that posts a public H100 price, normalised to US dollars per GPU-hour, with on-demand, spot and commitment-term rates side by side. It is collected automatically each day; the check date at the top is the last successful run.

Published prices by supplier

Supplier	Configuration	On-demand	Spot	1 mo	3 mo	6 mo	1 yr	3 yr	5 yr	Basis
Verda ↗	H100 SXM5 80GB · 1 GPU · 30 CPUs · 120GB RAMUSD per GPU-hour, 1-GPU instance	$4.75	$2.37	$4.6511	$4.6036	$4.5562	$4.3663	—	—	CalculatedCalculated: base × (1 − published discount)
Together AI ↗	HGX H100USD per GPU-hour, HGX H100 (8-GPU node pricing quoted per GPU)	$3.99	$1.99	$3.69	$3.45	$3.19	—	—	—	PublishedPublished by duration bucket: 3-30 days → 1 mo, 31-90 days → 3 mo, 91-180 days → 6 mo
Hyperstack ↗	H100 SXM 80GB · up to 24 CPUs/GPUUSD per GPU-hour, on-demand VM, billed per minute	$3.20	—	—	—	—	—	—	—	PublishedReserved from $2.72, term not publishedReserved: starting-from price, term not published
Lambda ↗	H100 SXM 80GB · 26 vCPUsUSD per GPU-hour, on-demand instance, price shown for a 1-GPU instance; larger instances listed separately	$4.29	—	—	—	—	—	—	—	PublishedNo term schedule published
Microsoft Azure ↗	ND96isr H100 v5 (8x H100 SXM) · 8 GPUs per VMUSD per GPU-hour, Standard_ND96isr_H100_v5 (8x H100) in eastus, Linux; VM price divided by 8; reservation total divided by term hours	$12.29	$2.27	—	—	—	$7.87	$5.40	$4.92	PublishedPublished reservation terms

USD per GPU-hour. “—” means not published by that supplier. Azure figures are the VM price divided by eight GPUs and, for reservations, the total divided by the term’s hours. Configurations, regions, networking, storage, minimum quantities and taxes differ; a lower hourly figure is not a cost-per-job verdict.

Commitment curves

H100 COMMITMENT CURVESUSD / GPU-hour

Verda · $4.75 on-demand → $4.37 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)

Verda · H100 SXM5 80GB · Checked 29 Oct 2026 · Calculated from published base + discounts. Together AI · published reserved rates by duration bucket (3-30 days → 1 month, 31-90 days → 3 months, 91-180 days → 6 months). Hyperstack · on-demand only; no term schedule published. Lambda · on-demand only; no term schedule 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.

On-demand price history

H100 ON-DEMAND · DAILYUSD / GPU-hour

HyperstackLambdaTogether AIVerda

One observation per day per supplier; the line joins consecutive daily checks. 14 Sep to 29 Oct, 46 days. Verda is its published on-demand base; other suppliers are their published on-demand rate. A flat line means the public page did not change. Microsoft Azure is tracked but plotted separately in the table because its per-GPU price sits several dollars above this range.

Data and downloads

Every daily observation is published as JSON and CSV, and the full day-by-day history is kept in the public repository. Collection rules, validation and comparison limits are described in the data methodology. Explore commitment terms interactively in the Pricing Lab.