When Meta reported its fourth-quarter and full-year 2024 results on January 29, 2025, the headline for the AI-buildout story was not earnings but spending. Meta’s capital expenditures, including finance leases, came to 39.23 billion dollars for full-year 2024, and the company guided sharply higher for the year ahead: “We anticipate our full year 2025 capital expenditures will be in the range of 60-65 billion dollars.” The jump - roughly 50 to 65 percent in a single year - was driven overwhelmingly by AI: GPUs, data centers, and the power to run them.
Meta’s numbers are a clean illustration of the broader capex supercycle because they come straight from the company’s own financial reporting rather than from a press splash. They show one hyperscaler’s annual infrastructure spend moving from the tens of billions toward the high tens of billions in a year, and they were echoed across the other major cloud and AI companies, whose combined annual capital spending climbed past 200 billion dollars and kept rising.
These figures are also the denominator in the ROI debate. Spending 60 billion dollars a year on infrastructure raises the bar for the revenue and productivity gains that have to follow. When Meta executives were later pressed on the return on this spending, the answers were notably hedged - a reminder that the bet is large and the payoff still being argued over.
Why a business reader should care: Meta’s guidance put a concrete, audited number on how fast AI capital spending was accelerating at a single company. It is the kind of first-party figure - not an announcement, but a financial commitment - that lets a reader gauge the real scale of the buildout and the size of the return it now has to earn.