Capital-cycle tempo compression — AI investment cycle duration?

By end of 2028, AI infrastructure investment cycles compress to 18-24 months from planning through major capacity deployment. Hyperscalers like OpenAI, Anthropic, and others launch new model architectures, immediately signal capex needs in earnings calls, secure private credit or public funding within 90 days, and spin up data centers and chip orders within 6 months. Revenue growth from these new models funds the next cycle before previous capex fully depreciates. The US CHIPS Act accelerates domestic chip output, shortening supply delays that previously stretched cycles to 36+ months.

AI investment cycles remain locked at 30-42 month durations through 2028. Hyperscalers plan capex in Q1, negotiate contracts and secure financing through mid-year, begin physical buildout in Q4, and complete deployment 24-30 months later. Chip shortages, real estate permitting delays, and power grid interconnection bottlenecks create persistent 12-18 month gaps between money committed and infrastructure live. Revenue growth from AI services cannot accelerate capex timing because manufacturing and grid constraints stay binding.