Big Tech's $3tn of AI commitments dwarfs the capex it reports

Signed leases and purchase contracts worth about five times a year of conventional spending sit in footnotes, not on balance sheets, and nobody has said when they fall due.

The $3tn that Big Tech has committed to the AI build-out is a ceiling on pledges, not a tally of cash spent, and the gap between the two is where the risk sits. A Wall Street Journal analysis, relayed by AI Fire, set the total against about $600bn of conventional capital spending over the latest 12 months, roughly five times as much. The reported capex line understates what these companies have promised to pay.

The Journal's estimate, as relayed by Toobit, covers nine companies: Alphabet, Meta, Microsoft, Amazon, Oracle, Nvidia, Broadcom, SpaceX and AMD. It splits into about $1.2tn of signed leases not yet started and about $1.9tn of purchase commitments. Alphabet alone disclosed about $811bn of purchase commitments and contractual obligations as of June 2026, and Meta about $347bn of leases not yet started.

Most of this does not appear as a liability on the balance sheet under current accounting rules. It sits in footnotes. That is what makes it a credit question and not merely an accounting one. Morgan Stanley's analysis, reported by Axios on August 27, put off-balance-sheet commitments at about $3tn for seven companies, on top of an estimated $770bn of debt and lease obligations already on balance sheets.

There is a caveat, and it cuts both ways. Axios said it is unclear over what period the commitments fall due, and that they are often contracts that could in theory be renegotiated. Companies may be able to walk back some of this if AI demand disappoints.

The measure that settles it is how much of the total turns up in reported capital spending in the coming quarters, and whether any company cuts its commitments.