
How to read this finding
Paper-based usage ratios are not sales or installed-capacity shares. The compute-growth estimate describes selected large experiments and should not be applied to every AI task.
Its research-paper analysis showed GPUs used 78 times as often as Google TPUs and 90 times as often as five named challenger suppliers combined.

Paper-based usage ratios are not sales or installed-capacity shares. The compute-growth estimate describes selected large experiments and should not be applied to every AI task.
Evidence you can use
Historical snapshot: October 2022. Dates and populations are specified per row.
| Measure | Reported value | Definition and source |
|---|---|---|
| GPU usage versus Google TPU | 78x | Research-paper comparison reported in the deck.2022 report, slide 52 (PDF page 52) |
| GPU usage versus five challengers combined | 90x | Graphcore, Habana, Cerebras, SambaNova, and Cambricon in the cited analysis.2022 report, slide 52 (PDF page 52) |
| Large-experiment compute growth | More than 300,000x | Selected large-scale experiments over the prior decade.2022 report, slide 82 (PDF page 82) |
Paper-based usage ratios are not sales or installed-capacity shares. The compute-growth estimate describes selected large experiments and should not be applied to every AI task.
Historical snapshot published October 11, 2022. This web edition was prepared on 2026-10-11 from the online deck and original launch posts. Findings and forecasts retain their original time frame.
Benaich, Nathan, and Ian Hogarth. “How dominant was NVIDIA hardware in the 2022 report?” State of AI Report 2022. Historical report snapshot; web edition prepared 2026-10-11.