GPU models: 8Providers: 12US price points: 244Market: United StatesList prices: 20 Nov 2026, 07:43 UTCMarket sample: 19 Nov, 19:21 UTC
US GPUs / H100 SXM

H100 SXM rental price in the US

Hopper GPU with 80 GB of HBM3. On-demand, spot and reserved list prices per GPU-hour at AWS, Azure and the neoclouds.

Arch HopperMemory 80 GB HBM3Launched 2022
Request H100 SXM →

On-demand price by provider

Per GPU-hour · ■ neocloud · ■ hyperscaler · daily history from 18 Nov 2026
  • Vast.ai · US-hosted median$1.73
  • Runpod · US$2.69
  • Vultr · US$2.99
  • Thunder Compute · US$3.20
  • Crusoe · US$3.90
  • Together AI · US$3.99
  • DigitalOcean · US$4.41
  • Lambda · US$4.59
  • CoreWeave · US$6.16
  • AWS · N. Virginia, US$6.88
  • Microsoft Azure · Virginia, US$11.06
Cheapest neocloud$2.69/GPU/hrRunpod
Cheapest hyperscaler$6.88/GPU/hrAWS · N. Virginia
Market median, US$1.73/GPU/hrVast.ai, US-hosted offers
Hyperscaler premium+156%AWS vs Runpod

What an H100 SXM costs in the US

GR-H100-US · per GPU-hour
Runpod, on-demandcheapest neocloud$2.69
AWS, on-demandcheapest hyperscaler · N. Virginia$6.88
AWS, 1-yearreserved, per hour over the term$4.33
AWS, 3-yearreserved, per hour over the term$2.97
Together AI, spotcan be stopped at short notice$1.99
Vast.ai, US median4 rentable offers$1.73

+156% AWS over Runpod, on-demand.

Lowest price you can book: $1.99 (Together AI, spot).

Instance shapes differ between providers; compare CPU, memory and networking before you book.

Price by rental term

Per GPU, cheapest US region per instance type. AWS and Azure publish 1- and 3-year prices; neoclouds price longer terms on request. Discounts are against the same row's on-demand price.

ProviderOn-demandSpot1-year3-year
Runpod NeocloudUS · H100 SXM, Community Cloud$2.69–––
Vultr NeocloudUS · vbm-112c-2048gb-8-h100-gpu, 8 GPUs$2.99–––
Thunder Compute NeocloudUS · H100 SXM, 1 GPU$3.20–––
Crusoe NeocloudUS · H100 80GB$3.90–––
Together AI NeocloudUS · HGX H100 cluster$3.99$1.99−50%––
Runpod NeocloudUS · H100 SXM, Secure Cloud$4.39–––
DigitalOcean NeocloudUS · NVIDIA HGX H100 GPU Droplet$4.41–$3.26−26%–
Lambda NeocloudUS · H100 SXM instance$4.59–––
CoreWeave NeocloudUS · HGX H100, 8 GPUs$6.16$2.46−60%––
AWSN. Virginia, US · p5.4xlarge$6.88–$4.33−37%$2.97−57%
AWSN. Virginia, US · p5.48xlarge$6.88–$4.33−37%$2.97−57%
Microsoft AzureVirginia, US · Standard_ND96is_H100_v5$11.06$2.04−82%$7.08−36%$4.86−56%
Microsoft AzureVirginia, US · Standard_ND96isr_H100_v5$12.29$2.27−82%$7.87−36%$5.39−56%
Vast.ai marketplaceUS-hosted · median of offers$1.73$1.67−4%––

Runpod Neocloud

US · H100 SXM, Community Cloud
On-demand$2.69
Spot–
1-year–
3-year–

Vultr Neocloud

US · vbm-112c-2048gb-8-h100-gpu, 8 GPUs
On-demand$2.99
Spot–
1-year–
3-year–

Thunder Compute Neocloud

US · H100 SXM, 1 GPU
On-demand$3.20
Spot–
1-year–
3-year–

Crusoe Neocloud

US · H100 80GB
On-demand$3.90
Spot–
1-year–
3-year–

Together AI Neocloud

US · HGX H100 cluster
On-demand$3.99
Spot$1.99−50%
1-year–
3-year–

Runpod Neocloud

US · H100 SXM, Secure Cloud
On-demand$4.39
Spot–
1-year–
3-year–

DigitalOcean Neocloud

US · NVIDIA HGX H100 GPU Droplet
On-demand$4.41
Spot–
1-year$3.26−26%
3-year–

Lambda Neocloud

US · H100 SXM instance
On-demand$4.59
Spot–
1-year–
3-year–

CoreWeave Neocloud

US · HGX H100, 8 GPUs
On-demand$6.16
Spot$2.46−60%
1-year–
3-year–

AWS

N. Virginia, US · p5.4xlarge
On-demand$6.88
Spot–
1-year$4.33−37%
3-year$2.97−57%

AWS

N. Virginia, US · p5.48xlarge
On-demand$6.88
Spot–
1-year$4.33−37%
3-year$2.97−57%

Microsoft Azure

Virginia, US · Standard_ND96is_H100_v5
On-demand$11.06
Spot$2.04−82%
1-year$7.08−36%
3-year$4.86−56%

Microsoft Azure

Virginia, US · Standard_ND96isr_H100_v5
On-demand$12.29
Spot$2.27−82%
1-year$7.87−36%
3-year$5.39−56%

Vast.ai marketplace

US-hosted · median of offers
On-demand$1.73
Spot$1.67−4%
1-year–
3-year–

Pricing & availability

14 sources · cheapest first

Runpod Neocloud

US · H100 SXM, Community Cloud · 1× GPU · checked 20 Nov 2026, 07:43 UTC
On-demand, per instance$2.69/hr1-year reservednot public
On-demand$2.69 /GPU/hrPrice source ↗

Vultr Neocloud

US · vbm-112c-2048gb-8-h100-gpu, 8 GPUs · 8× GPU · checked 20 Nov 2026, 07:43 UTC
On-demand, per instance$23.92/hr1-year reservednot public
On-demand$2.99 /GPU/hrPrice source ↗

Thunder Compute Neocloud

US · H100 SXM, 1 GPU · 1× GPU · checked 20 Nov 2026, 07:43 UTC
On-demand, per instance$3.20/hr1-year reservednot public
On-demand$3.20 /GPU/hrPrice source ↗

Crusoe Neocloud

US · H100 80GB · 1× GPU · checked 20 Nov 2026, 07:43 UTC
On-demand, per instance$3.90/hr1-year reservednot public
On-demand$3.90 /GPU/hrPrice source ↗

Together AI Neocloud

US · HGX H100 cluster · 1× GPU · checked 20 Nov 2026, 07:43 UTC
On-demand, per instance$3.99/hr1-year reservednot public
On-demand$3.99 /GPU/hrPrice source ↗

Runpod Neocloud

US · H100 SXM, Secure Cloud · 1× GPU · checked 20 Nov 2026, 07:43 UTC
On-demand, per instance$4.39/hr1-year reservednot public
On-demand$4.39 /GPU/hrPrice source ↗

DigitalOcean Neocloud

US · NVIDIA HGX H100 GPU Droplet · 1× GPU · checked 20 Nov 2026, 07:43 UTC
On-demand, per instance$4.41/hr1-year reserved$3.26/GPU/hr
On-demand$4.41 /GPU/hrPrice source ↗

Lambda Neocloud

US · H100 SXM instance · 1× GPU · checked 20 Nov 2026, 07:43 UTC
On-demand, per instance$4.59/hr1-year reservednot public
On-demand$4.59 /GPU/hrPrice source ↗

CoreWeave Neocloud

US · HGX H100, 8 GPUs · 8× GPU · checked 20 Nov 2026, 07:43 UTC
On-demand, per instance$49.24/hr1-year reservednot public
On-demand$6.16 /GPU/hrPrice source ↗

AWS

N. Virginia, US · p5.4xlarge · 1× GPU · checked 20 Nov 2026, 07:43 UTC
On-demand, per instance$6.88/hr1-year reserved$4.33/GPU/hr
On-demand$6.88 /GPU/hrPrice source ↗

AWS

N. Virginia, US · p5.48xlarge · 8× GPU · checked 20 Nov 2026, 07:43 UTC
On-demand, per instance$55.04/hr1-year reserved$4.33/GPU/hr
On-demand$6.88 /GPU/hrPrice source ↗

Microsoft Azure

Virginia, US · Standard_ND96is_H100_v5 · 8× GPU · checked 20 Nov 2026, 07:43 UTC
On-demand, per instance$88.49/hr1-year reserved$7.08/GPU/hr
On-demand$11.06 /GPU/hrPrice source ↗

Microsoft Azure

Virginia, US · Standard_ND96isr_H100_v5 · 8× GPU · checked 20 Nov 2026, 07:43 UTC
On-demand, per instance$98.32/hr1-year reserved$7.87/GPU/hr
On-demand$12.29 /GPU/hrPrice source ↗

Vast.ai marketplace

US-hosted · 4 rentable offers
Cheapest offer$1.73/GPU/hrMedian offer$1.73/GPU/hr
Median, on-demand$1.73 /GPU/hrMarketplace ↗

Use this data

Free to embed and quote with a link

Embed the live price

Updates every day on your page.

Chart for articles

H100 SXM on-demand price per GPU-hour by US providerDownload PNG

Cite

All US data (CSV)

H100 SXM US prices: questions

How much does an H100 SXM cost per hour in the US?

The cheapest public on-demand list price for an H100 SXM in the US is $2.69 per GPU-hour (Runpod), checked 20 Nov 2026, 07:43 UTC. On AWS or Azure it starts at $6.88 (AWS, N. Virginia).

Why is an H100 SXM more expensive on AWS and Azure?

The cheapest hyperscaler price is +156% above the cheapest neocloud ($6.88 vs $2.69). Hyperscalers bundle the GPU with their wider cloud (managed services, compliance, networking, support) and price it as one. Neoclouds sell GPU capacity on its own.

How much does an H100 SXM cost per month in the US?

At the cheapest on-demand list price, one H100 SXM costs about $1,963.70 per month (730 hours), or $15,709.60 for an 8-GPU server. A 3-year reservation on AWS brings it to $2,169.56 per GPU per month.

Where can I rent an H100 SXM on spot or interruptible terms?

The cheapest spot price we found is $1.99 per GPU-hour (Together AI). Spot capacity can be stopped at short notice.