Paid links may earn Compute Screener a referral fee. They never affect prices, weights or the order.
About NVIDIA A100
The NVIDIA A100 rental price today is $1.73 per GPU-hour, up 0.4% since yesterday, up 0.5% over the past week and down 10.4% over the past month. At that rate one A100 costs about $1,270 for a month of 730 hours. Over the past year it has ranged from $1.65 to $1.94. The index is the price of one benchmark contract: A100 SXM 80GB, on-demand, 8-GPU node, US hardware, CPU-neutral. It is standardized across 11 cloud providers and marketplaces, from Vast.ai at $1.12 to DigitalOcean at $3.20, and is updated every 2 to 3 hours.
The A100 is NVIDIA's Ampere-generation data-center GPU, announced and shipping on May 14, 2020, with more than 54 billion transistors on a 7-nanometer process. An 80 GB version followed on November 16, 2020. It trained the generation of models before the H100: Meta's first LLaMA used 2,048 A100s, and Stable Diffusion was trained on 256. Lenovo has withdrawn it from marketing, but it is still the cheapest GPU on this site to rent.
Specifications
| Spec | A100 SXM 80GB |
|---|
| Architecture | Ampere, 7 nm, more than 54 billion transistors |
| Memory | 80 GB HBM2e, 2,039 GB/s; the 40 GB version uses HBM2 |
| BF16 / FP16 tensor | 312 TFLOPS dense, 624 with sparsity |
| FP8 | not supported |
| NVLink | 600 GB/s |
| Power | 400 W |
| Typical server | HGX or DGX A100, 8 GPUs |
Source: NVIDIA A100, accessed October 24, 2026.
Where the A100 is rented
Google Cloud offered the first A100 machines in the public cloud, called A2, in private alpha from July 7, 2020. AWS P4d became generally available on November 2, 2020. Today AWS rents p4d with 40 GB and p4de with 80 GB, Azure ND A100 v4 with 40 GB and NDm A100 v4 with 80 GB, and Google A2. GetDeploying tracked 46 A100 providers on October 23, 2026.
What this price measures
The index is the posted on-demand price of one benchmark contract, an A100 SXM 80GB in an 8-GPU node of US hardware, rented by the hour from GPU clouds and marketplaces; prices from other regions are counted without a regional adjustment for now. Prices for the 40 GB version and PCIe cards are adjusted to that contract, or left out when there are too few to adjust; reserved and spot prices are not counted.
On October 23, 2026, 8-GPU A100 machines were listed at $1.99 per GPU-hour at Lambda for 40 GB, $2.70 at CoreWeave for 80 GB, $2.74 at AWS for a 40 GB p4d, $3.05 at Oracle, $3.40 at Azure and $3.72 at Google Cloud; the lowest on-demand listing across all configurations was $0.89.
What moves A100 rental prices
- Age. The A100 has no FP8 support and about a third of the H100's dense compute, so it rents mainly for smaller models, fine-tuning and workloads where price per hour matters more than speed.
- Hyperscaler pricing. AWS cut on-demand prices of P4d and P4de by 33% from June 1, 2025 and added A100 capacity in Seoul, Sydney, Canada and London.
- Shortages of newer GPUs. When H100, H200 and Blackwell capacity is hard to find, as in early 2026, some demand moves down to older GPUs.
- Limited new supply. Lenovo no longer markets the A100, though AWS still added A100 capacity in new regions in 2025.
- Export controls. A100 sales to China have needed a US license since August 2022, under the same rule that covered the H100.
Price history
Before October 23, 2026 this index's history is daily from November 7, 2025, rebuilt with this site's methodology from the posted prices recorded in SkyPilot's open price catalog for 5 to 8 providers a day, and weekly before that, from GetDeploying's A100 series; both are scaled to join the live index.
Futures and benchmarks
No A100 futures contract has been filed; CME's filed contracts and ICE's contracts set for listing cover the H100 and B200. The Bloomberg ticker beside the name, SDA100RT, is Silicon Data's A100 index, which measures a different market from this index.
FAQ
Why is the A100 cheaper per hour but dearer per unit of compute? On GetDeploying's October 23, 2026 medians, an A100 rented for about half the hourly price of an H100, $1.88 against $3.85, with about a third of its dense BF16 compute.
What is the difference between the 40 GB and 80 GB versions? The 80 GB version has twice the memory and faster HBM2e, about 2 TB/s against 1.6 TB/s. This index measures the 80 GB SXM version.
How often is this index updated? Every 2 to 3 hours, with one value per day in US Central time that is final when the day closes.
Sources
Text reviewed Oct 24, 2026; the figures in it are as of that date. The figures at the top of this page update every 2 to 3 hours. Not investment advice.
