Stockatomics Launches Monthly Index Tracking AI Infrastructure Hardware Costs

Summary

Stockatomics’ AI Infrastructure Index measures how much the physical hardware behind AI costs over time. It combines accelerators and server processors, optical transceivers, memory, circuit equipment, power systems, transformers, cables, and cooling equipment used in data centers. The index is set at 100 in January 2023; the page displays a level of 1,212.47 and labels the latest month as August 2026, while a chart on the page is marked October 2026. A reading of 200 represents twice the January 2023 unit cost, not a stock-market value or a dollar price. Stockatomics builds the measure each month by comparing the average value of each hardware unit imported into the United States with its January 2023 level. Components are weighted by their share of US import value in 2025, so accelerators and server processors account for 55.71% and optical transceivers and network optics for 23.84%. The page says the rise can reflect both higher prices caused by demand exceeding supply and a shift toward more powerful, expensive AI systems. New benchmarks are incorporated monthly.