Meta's October 21 Earnings Put a $130–145 Billion AI Spending Plan Under the Microscope

As of October 18, 2026, 8:45 p.m. Eastern Time, Meta Platforms has not reported third-quarter earnings. Its next scheduled release is Wednesday, October 21, after the U.S. closing bell, with an earnings call at 4:30 p.m. ET. The latest complete cash-equity trading session was Friday, October 16. Those distinctions matter: a weekend earnings preview is not evidence of an earnings beat, a guidance increase or a new trading-session price reaction.

The immediate question is no longer whether Meta can attract advertisers. Its advertising franchise has already demonstrated substantial pricing and impression growth. The harder question is whether the economic return on its rapidly expanding AI infrastructure can catch up with the cost of the buildout. Investors will be evaluating two businesses inside one income statement: a highly profitable advertising platform and a capital-intensive effort to build the next generation of AI products.

The numbers Meta actually reported

Meta's official July 29 release covered the three months ended June 30, not the quarter due to be released this week. Second-quarter revenue was $60.801 billion, a 28% year-over-year increase. Advertising impressions across the Family of Apps rose 14%, while average price per ad climbed 12%. Average Family daily active people reached 3.60 billion in June, up 3% from a year earlier. These operating measures showed that advertiser demand, inventory and monetization were all contributing to the top line.

But operating economics moved in the other direction. Costs and expenses reached $42.026 billion, up 55% year over year; operating income declined 8% to $18.775 billion; and the GAAP operating margin contracted to 31% from 43%. The expense total included $2.40 billion of legal charges and $1.18 billion of severance costs. Those items explain part of the comparison, but they should not distract from the underlying strategic issue: Meta is committing large recurring resources to engineers, compute and data centers.

Capital expenditures including principal payments on finance leases were $31.08 billion in the second quarter. Operating cash flow was $31.86 billion, while Meta's reported free cash flow was only $784 million. It would be misleading to call that a collapse in the company's operating cash generation; the much narrower residual reflects its investment cycle and Meta's stated cash-flow definition. It is nevertheless relevant to shareholders because a company can grow revenue while leaving less immediately distributable cash.

What Wednesday's guidance must resolve

Management's last published guidance called for third-quarter revenue of $61 billion to $64 billion. The same July statement projected full-year 2026 expenses of $165 billion to $169 billion and capital expenditure, including finance-lease principal, of $130 billion to $145 billion. Those are management forecasts, not results. They should not be compared with a third-party analyst consensus as if the two were interchangeable.

The midpoint of the quarterly revenue range is $62.5 billion. Relative to second-quarter sales of $60.801 billion, that midpoint implies sequential growth of roughly 2.8%, but seasonality makes a sequential comparison less informative than growth against the corresponding prior-year quarter. A result near the upper end would support the argument that advertising can finance the spending program. A result near the lower end, accompanied by rising expense expectations, would make the return-on-capital question harder.

A large portion of AI investment is initially capitalized rather than deducted from current operating profit. Servers and buildings depreciate over their useful lives, so a significant buildout can pressure free cash flow before its full depreciation burden appears in GAAP earnings. Investors therefore need the income statement, cash-flow statement and forward capex guidance together. Treating quarterly earnings per share as a complete verdict on AI economics would miss the timing difference.

Why Meta matters beyond its own shares

Meta is an important customer and reference point for the semiconductor, memory, networking, power and data-center supply chains. A sustained capital-spending plan can support equipment orders even if near-term profits come under pressure. Conversely, a spending slowdown or warning about utilization would weaken the assumptions behind several other listed companies' growth forecasts. This is a transmission mechanism, not proof that one earnings call determines every supplier's sales.

The timing is unusually sensitive. Reuters reported on October 16 that a broad selloff in AI-linked stocks had shifted attention toward earnings and the ability to justify infrastructure expenditure. Investors who once rewarded larger capital budgets may now demand evidence of advertising improvements, subscription income or enterprise revenue sufficient to generate a return. These perspectives are not mutually exclusive: spending can be rational for the business while still lowering short-term shareholder cash flow.

The debate is about returns, not just model capability

The optimistic case begins with Meta's distribution, advertiser relationships and the measurable effect of AI on ad ranking and targeting. Improvements in conversion efficiency may support higher ad prices without requiring a separate consumer AI subscription to be material immediately. New agent and enterprise products might create another revenue stream, but they should be valued on paying customers and recurring revenue rather than product announcements alone.

The skeptical case starts with the denominator of return on invested capital. Building infrastructure quickly increases invested assets. Even if operating profits remain impressive in absolute dollars, the return earned on incremental spend may lag its cost of capital for an extended period. Competition can also force Meta to provide expensive AI features at low prices. Higher interest rates make a delayed payoff less attractive to the market, whether or not Meta faces immediate funding constraints.

Legal and regulatory expense adds another source of noise. GAAP net income and adjusted interpretations can diverge when substantial charges occur. Readers should identify the precise charge, period, tax treatment and accounting presentation before attributing a per-share surprise to the core advertising business. A headline EPS beat driven by accounting may be less informative than a smaller beat accompanied by improving operating margins.

A practical framework for the release

First, place reported third-quarter revenue against Meta's own $61 billion–$64 billion outlook and examine the advertising volume-price mix. Second, reconcile expense growth with any unusual charges rather than assuming every dollar is recurring AI expense. Third, track the new full-year and any preliminary 2027 capital-spending language. Fourth, compare operating cash generated with spending on property, equipment and finance-lease principal. Fifth, ask whether management supplies quantifiable signs of paid AI adoption or merely describes user engagement.

The release after Wednesday's close and the 4:30 p.m. ET call are confirmed events. The next FOMC meeting is scheduled for October 27–28; that later policy decision may alter the discount rate investors apply to long-duration AI projects. No third-quarter result or October rate decision has occurred at this cutoff.

Meta does not need to prove that every AI product is profitable this week. It does need to show that a powerful core business can support the investment pace without leaving the expected shareholder return entirely dependent on distant, unverified monetization. That distinction will likely matter more than the headline revenue beat alone.