Meta's $125B–$145B AI spend lacks clear payoff plan

Meta Platforms, with a market capitalization of $1.7 trillion（約270兆円） and 3.56 billion daily active users across its family of apps, has long been regarded as one of the world's most dominant and successful companies. However, the company now faces a critical test of investor confidence as it enters an unprecedented period of capital intensity tied to artificial intelligence infrastructure.

On its first-quarter earnings call, when an analyst directly asked CEO Mark Zuckerberg what the return on invested capital would be from Meta's AI-related spending over the coming 12 to 24 months, Zuckerberg offered a notably vague response. "That's a very technical question," he replied, and then explained the company's philosophy: build great products and experiences first, then scale and monetize them afterward. This response came in the context of forecasts showing Meta will spend between $125 billion（約20兆円） and $145 billion（約23兆円） on AI data center capital expenditures in 2026 alone, with analysts expecting even larger spending in 2027. To place this in perspective, analysts' consensus view is that Meta will report $145 billion（約23兆円） in EBITDA in 2026—meaning its capex plans will consume nearly all of that annual earnings figure.

The absence of concrete ROI metrics or timelines has created unease among investors, even those inclined to give Meta's management team the benefit of the doubt. While the company's track record of successful capital allocation might justify some trust, the sheer magnitude of this spending warrants more detailed accountability. Investors would have greater confidence if management had articulated which specific metrics it is tracking to gauge the success of the AI spending spree, or at least provided a window into when adequate returns might be expected to materialize.

Meta's primary revenue source remains advertising, and the digital advertising segment is performing strongly, with sales up 33% year over year in the first quarter. The logical rationale for the AI spending would be to enhance targeting and monetization techniques, as well as improve user engagement. However, Meta's recent announcement of a new cloud segment under its Meta Compute initiative—designed to lease excess compute resources to external customers—may suggest the company has overbuilt capacity. While leasing excess capacity can generate revenue quickly, given that global data center demand substantially outstrips supply, this move also underscores a potential capital efficiency problem. Hundreds of billions of dollars are now at stake, and investors would be well-served to scrutinize Meta's capital allocation decisions more critically than they have in the past.

Meta's massive AI spending announcement reflects the hyperscaler's confidence in its long-term competitive position, but it has created a credibility gap with investors. The company operates one of the world's most dominant advertising platforms, with revenue from ad sales soaring 33% year over year in the first quarter, and it has historically earned trust through successful capital allocation. However, the scale of the 2026–2027 spending—potentially consuming the vast majority of annual EBITDA—represents uncharted territory, and Zuckerberg's vague response about ROI on the earnings call has raised questions about whether management has a rigorous framework for justifying these outlays.

The announcement of the Meta Compute cloud initiative suggests the company recognizes it may have overbuilt capacity and is attempting to monetize excess resources while global data center demand remains constrained relative to supply. While this could generate revenue quickly, it also signals that Meta's internal capacity planning may not have been as precise as investors would expect for a company of this scale and sophistication. The core tension remains unresolved: Meta's shareholders are being asked to trust management on a bet that will consume nearly all of 2026 earnings, yet they lack concrete metrics or timelines for when—or how much—profitability will result.