A CapEx cycle describes the periodic expansion and contraction of capital expenditure by corporations, typically lasting 5–10 years. In the AI infrastructure context, the current CapEx cycle is driven by hyperscaler spending (Microsoft, Google, Amazon, Meta) on data centers, GPUs, networking equipment, and power infrastructure. Total hyperscaler CapEx exceeded $200 billion in 2024 and is projected to grow through 2027.
CapEx cycles follow a predictable pattern: early investment creates capacity, capacity enables revenue growth, revenue growth justifies more investment, until supply overshoots demand and spending contracts. The key question for AI investors is whether the current cycle resembles the telecom buildout of 1998–2001 (which ended in bust) or the cloud buildout of 2010–2020 (which sustained itself through genuine demand growth). Monitoring the ratio of CapEx to revenue generation across the hyperscaler complex provides the clearest signal of cycle health.
Capital expenditure cycles drive long-term returns in infrastructure-heavy sectors. The current AI CapEx cycle is the largest technology buildout since the telecom boom of 1996-2000, with hyperscalers committing over $200 billion annually. Understanding where we are in the cycle — early expansion, peak, or contraction — determines whether infrastructure stocks are buys or sells.
Historical CapEx cycles follow a predictable pattern: initial under-investment creates bottlenecks and pricing power (where we are now in AI silicon), followed by over-investment that crushes margins. The key question for investors is timing — the gap between peak spending and peak earnings can be 12-18 months, creating significant risk for those who confuse revenue growth with sustainable returns.
CapEx cycles connect to the CapEx Cliff thesis, the 6-Layer Model of AI infrastructure, and Sentinel Tickers that signal cycle turns 9-12 months early.
