84
#1 of 8
notable AI models trained on Nvidia A100 chips since 2017 (V100: 69; H100: 28)
What the chart shows
Since 2017, 84 notable AI models have been trained on Nvidia’s A100, more than on any other accelerator. The previous generation, the V100, accounts for 69 and Google’s TPU v3 for 44.
Newer chips have fewer models so far: 28 each for Nvidia’s H100 and Google’s TPU v4, and 4 for the H800, a version of the H100 made for the Chinese market.
Key numbers
| Measure | Value |
|---|---|
| Nvidia A100 | 84 |
| Nvidia V100 | 69 |
| Other accelerators | 54 |
| Google TPU v3 | 44 |
| Nvidia H100 | 28 |
| Google TPU v4 | 28 |
How to read it
- Counts only notable models whose training hardware is known.
- Other groups all remaining accelerators (54 models).
- Running totals at each chip’s latest data point, between late 2019 (P100) and December 2025.
Source: Epoch AI, 2026, via the AI Index Report 2026 (Stanford HAI), figure 1.2.2: Cumulative number of notable AI models trained by accelerator, 2017–25. Chart and text by AI Adoption.
Numbers checked against the published data on 19 October 2026.
Download the data (CSV) · Cite this chart
Numbers checked against the published data on 19 October 2026.
Download the data (CSV) · Cite this chart
Discussion