Global AI Compute Capacity Doubles Every Nine Months, Epoch Finds
AI
Reported by Dr. Kai Wei, Frontier AI Desk
- Quarterly AI compute additions rose from 0.8 M to 6.3 M H100-equivalents between Q1 2024 and Q2 2026.
- The fitted trend indicates a 2.5× yearly increase, with a 9.1-month doubling time (90 % CI 8.4-10.0 months).
- Growth rates have steadied since 2024, despite noisy quarter-to-quarter changes linked to new chip generations.
Epoch AI’s latest analysis shows quarterly AI chip additions growing at a 2.5× annual rate, implying a 9-month doubling time.
«New global AI computing capacity is doubling every 9 months»
Summary and commentary by Aura-V. Facts, figures and opinions belong to the original.
Epoch’s latest forecast, based on sales data from Nvidia, Google, AMD, Amazon, Huawei and Cambricon, estimates that quarterly AI compute capacity additions increased from 0.8 million H100-equivalents (H100e) in Q1 2024 to 6.3 million H100e in Q2 2026. The analysis fits a log-linear regression over ten quarters (Q1 2024-Q2 2026) and derives a 2.5× annual growth rate, corresponding to a 9.1-month doubling time (90 % confidence interval 8.4-10.0 months) Epoch AI Compute & Frontier Forecasting.
Quarterly additions grew faster in 2023, with Nvidia GPUs and Google TPUs showing six-fold higher sales in Q1 2024 versus Q1 2023. Since 2024, the growth of quarterly additions has remained stable, while cumulative capacity growth converged to the same rate, rising 2.6× between Q2 2025 and Q2 2026.
Our read: The near-steady 2.5× yearly expansion suggests that AI infrastructure planners can expect predictable scaling pressures, but the reliance on sales-based metrics and the exclusion of smaller designers introduce uncertainty into precise capacity forecasts.
Interested in implementing this technology inside your company?
Our studio designs custom agent meshes, data pipelines, and outbound GTM systems for forward-looking enterprises.
