The five largest cloud and AI companies — Meta, Amazon, Microsoft, Alphabet, and Oracle — will collectively spend an estimated $720 billion on AI infrastructure in 2026. [1] The figure, compiled by Motley Fool from each company's capital expenditure guidance, represents a near-doubling from 2025 levels and dwarfs the combined capex of any industry in any prior year. [2]

Meta alone has committed up to $135 billion, anchored by a multibillion-dollar Nvidia chip deal announced in February. [3] Jensen Huang used Nvidia's GTC conference this week to argue the spending is rational: AI infrastructure is replacing traditional data centers, and the companies buying chips today are building the computing substrate for the next decade. [4] CNBC was less sanguine, reporting that free cash flow for the megacaps could drop up to 90 percent in 2026 as capital expenditure outpaces revenue growth. [5]

The bet is existential in both directions. If AI revenue materializes at the scale these companies project, the infrastructure will have been cheap at twice the price. If it does not, $720 billion in sunk capital becomes the most expensive lesson in corporate history since the telecom bust of 2001.