“Capex supercycle” is the label for the extraordinary wave of capital spending on AI infrastructure - data centers, GPUs, networking, and electricity - that defines the current boom. What makes it a supercycle rather than ordinary investment is the magnitude and concentration: a handful of companies each committing tens of billions of dollars per year, with the totals rising sharply year over year.
The scale is documented in the companies’ own words. On January 3, 2025 Microsoft president Brad Smith wrote that “in FY 2025, Microsoft is on track to invest approximately 80 billion dollars to build out AI-enabled datacenters,” with more than half in the United States. In September 2025 NVIDIA said it intended to invest up to 100 billion dollars in OpenAI to deploy 10 gigawatts of systems. The Stargate Project, announced in January 2025, was framed as a plan to invest 500 billion dollars over four years. Across the major hyperscalers, combined annual capital spending climbed past 200 billion dollars and kept rising.
Two features distinguish this cycle from past technology booms. First, much of the spend is on hardware that depreciates fast - GPUs are improved every year or two - so the investment must earn a return quickly. Second, the spending is unusually interlinked: chip makers invest in the labs that buy their chips, and labs commit to cloud providers that buy those same chips, blurring the line between customer and investor (see circular financing).
Why a business reader should care: the supercycle is the financial engine of the entire AI era, and it rests on a bet that demand and revenue will eventually justify the outlay. Whether that bet pays off - the question raised by the “$600B” essay and the ROI debate - will determine whether this looks in hindsight like the railroads or like the dot-com fiber glut.