H100 80 GB price per hour
On AWS in September 2026, an hour on an H100 80 GB cost $2.57 on spot, against $8.43 on-demand: 64% less, comparing the same regions. Spot held level with August.
- Spot, per GPU-hour
- $2.57
- On-demand, per GPU-hour
- $8.43
- Spot discount
- 64%
- Cheapest region
- Stockholm
- eu-north-1
- Spot vs August
- 0%
Daily price per GPU-hour, September 2026
The price in effect at the end of each day on p5.48xlarge, the cheapest H100 machine per GPU this month. Each point is the median across that day’s regions.
By month
One row per monthly report. The history grows with each one.
| Month | Cloud | Spot / GPU-hr | On-demand / GPU-hr | Spot discount | vs month before |
|---|---|---|---|---|---|
| September 2026 | AWS | $2.57 | $8.43 | 64% | 0% |
Machines with the H100 80 GB, September 2026
Hourly prices for the whole machine, cheapest per GPU first. “Region gap” is how many times as much the most expensive region charged as the cheapest.
| Machine | GPUs | Spot | Spot / GPU-hr | On-demand | Cheapest region | Region gap | Regions |
|---|---|---|---|---|---|---|---|
| p5.48xlarge | 8 | $20.555 | $2.569 | $67.424 | Stockholm eu-north-1 | 5.8× | 12 |
| p5.4xlarge | 1 | $2.630 | $2.630 | $8.256 | Stockholm eu-north-1 | 9.8× | 9 |
Questions
How much does an H100 cost per hour?
On AWS in September 2026, an hour on an H100 80 GB cost $2.57 on spot, against $8.43 on-demand: 64% less, comparing the same regions. Spot held level with August.
Did the H100 80 GB price change since August?
September 2026 spot: $2.57 per GPU-hour. Spot held level with August. The change compares only regions priced in both months.
Where is H100 80 GB spot capacity cheapest?
In September 2026, the cheapest region for p5.48xlarge was Stockholm (eu-north-1) at $13.44 an hour for the whole 8-GPU machine. The most expensive, São Paulo (sa-east-1), cost 5.8 times as much.
What does an H100 80 GB cost per GB of memory?
$0.032 per GB of GPU memory per hour on spot in September 2026: $2.57 an hour for 80 GB. GPU memory decides how large a model fits, so this is the fair way to compare cards of different sizes.
Prices are what AWS published, not proof that capacity was available at them. Method and coverage are in the September 2026 report.
Get the next report
GPU prices every month, by GPU and by region. One email a month; we’ll ask you to confirm first.
Free. One-click unsubscribe in every issue.
Let the price come to you.
Otium is in closed alpha. Set a deadline for your batch AI work, and it waits for the lowest-cost compute available before then, so you don’t have to watch these prices yourself. Leave your email and we’ll send an invite when a spot opens.
No credit card, no spam — one email when your invite is ready.
Closed alpha — onboarding is gated while we calibrate. Already invited? Sign in.