Can AI Growth Absorb High Rates? A 18:25 한국어
Jump to the 3 latest entries Google Project Suncatcher — first Gemma 4 response from an orbital TPU, a 1kW proof of concept(1)Anthropic's autonomous red team finds monitor-evading misalignment datasets in every run(2)Epoch AI — open-weight GLM 5.3 exploits the patch gap in all six Firefox releases (about $1,000 per version)(2)

← Back to the record

Epoch AI — global AI compute doubling every 9.1 months, up about 2.5x per year

Date2026-10-29
CategoryCompute
JudgementNeutral
Gauge weight2 / 3
SourceEpoch AI

Summary

In a data insight Epoch AI published on October 29, 2026, the growth rate of compute from quarterly AI chip sales (in H100 equivalents) was put at about 2.5x per year since 2024, with a doubling time of 9.1 months (90% CI 8.4-10.0 months). Quarterly additions rose from 0.8M H100e in Q1 2024 to 6.3M H100e in Q2 2026, of which Nvidia was 3.95M and Google 1.6M. The slowdown trend was stated to be not statistically significant (p=0.26).

A/B Impact Analysis

This record is filed as neutral, as the watch judgement explicitly read it. Growth of about 2.5x per year is not evidence that the compute bottleneck is binding now, and being slower than about 6x per year in 2023 does not make it evidence of acceleration either. Year-on-year growth fell from 2.8x in Q2 2025 to 2.2x in Q2 2026, but the authors stated the slowdown is not statistically significant (p=0.26), the figures are on a sales-timing basis, and operation lags sales by several months.

← Previous entryNext entry →