Meta’s Revenue Test Is Whether AI Gains Can Outrun a $140 Billion Buildout

Bank of America expects Meta Platforms to report third-quarter revenue of $63.8 billion, slightly above the $63.3 billion consensus cited in the original report. The difference is modest. The more important question is whether artificial intelligence can lift advertising revenue fast enough to justify a much larger infrastructure budget.

The bank expects Meta’s advertising tools to improve targeting, creative production and return on spending for marketers. It also points to adoption of Muse AI, enterprise demand and the company’s large-language-model strategy as evidence to seek on the earnings call. Meta is scheduled to report third-quarter results on October 21.

Bank of America forecasts third-quarter earnings per share of $3.79, well below the $6.08 consensus, because it expects a one-time $10 billion legal settlement charge involving state attorneys general. Excluding that charge, the bank expects operating income to exceed market expectations. That distinction is material, but investors should wait for Meta’s filing to confirm the amount and accounting treatment.

The fourth-quarter estimates preserve the same split between strong demand and heavy cost. Bank of America projects revenue of $74.1 billion and earnings per share of $8.74, both above the consensus figures cited in the original report. It expects management to guide to $71.5 billion to $74.5 billion of fourth-quarter revenue, which would represent 19% to 24% growth.

Infrastructure is the harder side of the equation. The bank expects Meta to narrow 2026 capital-expenditure guidance to $140 billion to $145 billion and has raised its own 2027 estimate to $205 billion. Those numbers are analyst forecasts, not company results. They frame how much cash may be required for data centers, servers and other AI capacity.

Meta’s second-quarter release showed why investors are willing to entertain the spending. Advertising demand and engagement remained strong, and management continued to link AI systems to better recommendations and ad performance. Yet rising capital expenditure can depress free cash flow and returns if revenue gains arrive more slowly than infrastructure costs.

The earnings call needs to provide three pieces of evidence: reported advertising growth, a measured improvement in monetization attributable to AI and a capital plan tied to useful capacity rather than ambition alone. A revenue beat of $500 million would be encouraging, but it would not by itself validate a spending program measured in hundreds of billions.

The bull case is that Meta’s scale turns AI improvements into revenue across billions of users. The risk is that the infrastructure bill compounds faster than the proof.