Big Tech plans record $720 billion AI capex as profits stay out of reach

Amazon, Microsoft, Alphabet and Meta wrapped their latest quarterly earnings calls with the same message: the AI data center buildout is accelerating, and nothing — not Wall Street nerves, not the absence of demonstrated profits — is slowing it down. According to The Verge, which reviewed all four calls, the companies told investors they plan to spend at least $720 billion on capital expenditures this year, most of it on AI infrastructure, on top of at least $1 trillion in spending commitments already arranged for coming years.

The numbers only go up

Each company offered its own version of "more." Alphabet plans to spend between $195 billion and $205 billion this year and, according to CFO Anat Ashkenazi, will increase that "significantly" in 2027. Amazon is budgeting $220 billion, the largest figure in its history, and CEO Andy Jassy said 2027 spending will be "meaningfully more," framing the moment as a "once in a lifetime opportunity" to expand the business.

Meta expects to spend $135 billion to $145 billion this year and gave no figure for next year, though The Verge reports that major financial institutions anticipate an increase. CEO Mark Zuckerberg said Meta's agentic AI push, including the rollout of Muse, only becomes profitable if compute costs come down, and the company expects a "tenfold" gain in compute efficiency next year — an effort he said will itself demand significant investment.

Microsoft, until now the only hyperscaler that had avoided borrowing to fund its buildout, posted $45.2 billion in quarterly capital expenditures against an expected $50 billion — a gap The Verge attributes to "third party capacity delays" rather than restraint. The morning after CFO Amy Hood noted the company had access to debt markets but had "not yet" tapped them, Microsoft sold $25 billion of bonds across ten parts.

Revenue is growing, but not enough

The top line is not the problem. Amazon posted its fastest quarterly growth in five years, Meta's ad revenue is surging, and Google Cloud's revenue more than doubled in a year. As The Verge notes, that may still not be enough to justify the outlay.

Alphabet supplied the sharpest illustration. Philipp Schindler, the company's chief business officer, acknowledged on the call that growth in AI Mode queries dented results because AI Mode monetizes at lower rates than classic search. Search ad revenue grew 12% in the quarter, down from 17% in the same quarter a year earlier — Google's AI search is undermining the core business before the company has worked out how to monetize the replacement. Schindler said ad experiments show "promising results" but offered no specifics. Alphabet shares fell 13% the next day, the worst single-day drop in the company's history, before recovering this week.

Wall Street blinks, then looks away

The spending plans triggered a sharp selloff in tech stocks that eased as attention shifted to the midterm elections, where voters chose candidates who favor data center moratoriums and other restrictions. The Verge speculates that the subsequent stock rebound may reflect investor hopes that new regulation could force a slowdown and buy the industry time to make AI profitable. Notably, the elections went unmentioned on all four earnings calls.

Anthropic shows the strain beyond Big Tech

The cash burn is not confined to the hyperscalers. Anthropic's updated S-1, as reported by The Verge, shows quarterly revenue of $17 billion — up more than 1,000% year over year — against compute costs of $17.5 billion. The startup has committed nearly $900 billion to future compute and infrastructure, plans to build its own data centers, and will send part of that money to Google for TPU chips. Two analysts pressed Alphabet executives on why a company that says it is capacity-constrained sells that capacity to rival labs; Ashkenazi declined to give a clear answer, and The Verge points out that those contracts front Google's costs rather than delivering cash today.

Why it matters

Three years and hundreds of billions of dollars in, no major player has shown the buildout can pay for itself, and none will stop first — slowing down now feels riskier than overspending. Wall Street's brief revolt and quick forgiveness suggest investors will tolerate the burn only while revenue keeps climbing, and Google's search numbers show AI can cannibalize existing profits faster than it creates new ones. With Microsoft now borrowing to fund capex and Anthropic's compute bill exceeding its revenue, the spree is increasingly debt-financed and still waiting for a business model.