Meta keeps beating revenue, and Wall Street still wants receipts for AI

What does a company have to do to get applauded these days? Meta grew ad sales 26% in the third quarter and landed above the top of its own revenue guidance. Yet the real story of the quarter isn't the beat. It's the cash. The market has quietly changed the question: it no longer asks "how big is your AI ambition?" It asks "where is the cash?"

Here is what happened. Meta reported Q3 revenue of $64.60 billion, ahead of the $63.3 billion that Bloomberg's analyst consensus had expected. Then the bill arrived. According to StockTitan's breakdown, Meta spent $34.91 billion on data centers and equipment in the quarter, more than the $32.79 billion its operations generated. Free cash flow was minus $2.12 billion, against plus $10.63 billion a year earlier. Stocktwits reports the shares rose about 3% after hours, but after-hours moves are a first reaction, not a verdict.

The market isn't scared of AI. It's scared of the arithmetic

We have a recent reference for how this audience behaves. After the Q2 report in July, Fortune reported that Meta's shares fell as much as 10% in after-hours trading and were down 7% later, even though revenue beat expectations. Our reading: investors don't object to big capex in itself. They object to capex that outruns the business funding it.

The most telling figure in Q3 isn't the capex, it's the profit line. By StockTitan's calculation, earnings per share excluding one-time items grew only about 2% in a quarter when sales grew 26%. Costs outran revenue. Meanwhile, Meta's 2026 capex guidance is now $135 billion to $145 billion, per Business Insider. That is "narrowed," but it narrowed upward, because the lower end went up.

Mark Zuckerberg says in the release that "our investments are paying off, the core business is strong." Maybe. But "paying off" is a claim. The one hard number in Fortune's July account of Meta's AI stack is an ad system that AI has made 15.7% better at converting. That is real, and we think it is also small next to a spending plan in the $135 billion to $145 billion range.

Meta's problem is that it has no cloud to hide behind

Fortune made a sharp point in July that we think explains the harsher treatment. Microsoft, Amazon and Google can rent their data centers to outside customers, so their capex produces revenue almost immediately. Meta builds the same kind of infrastructure but, for now, mostly consumes it itself. The market is effectively asking: if your rivals can show a meter running, why can't you?

Zuckerberg's answer was that it "would be foolish" to sell all the compute for a short-term profit, and that he expects "a significantly higher margin on selling intelligence rather than selling compute directly." That is a coherent strategy. It is also a promise, because part of that "intelligence" stack, including the 24/7 personal assistant he alluded to, did not exist yet when he described it. In an accountability phase, a promise about a product that doesn't exist is the weakest currency you can pay with.

We'd add a note on direction of travel. Meta's finance chief, Susan Li, said in July that operating cash flow was "the primary pillar" financing its ambitions. By StockTitan's reading, one quarter later that pillar no longer covers the whole build by itself. Meta also issued $24.9 billion of long-term debt in Q2 and bought back no stock, per Fortune. Nothing about that is reckless on its own, but the trajectory points away from a cash-rich ad machine and toward something closer to a capital-intensive utility, a comparison Fortune itself makes.

What we'd watch instead of the headline

Don't read this as "the AI boom is over." The phase changes, the spending doesn't. Our view is that the debate shifts from how much to what it buys.

Three things deserve attention. First, whether Meta announces the cloud business Zuckerberg said it plans to enter, because that would be the clearest route to visible returns. Second, whether profit growth starts to keep pace with sales growth, rather than lagging it by more than 20 points. Third, headcount: by StockTitan's numbers, Meta has 15% fewer employees than a year ago while its spending on data centers and equipment rose 80%. That is a bet that machines beat people, and it will be judged on results.

Our verdict: Meta's ad business is still excellent, and the company doesn't look short of demand for compute. But the grace period for "trust us, it'll pay off" looks like it is ending across Big Tech, and Meta, with no cloud revenue to point at, is simply the one feeling it loudest. If you're an investor, stop asking how much Meta will spend and start asking what each billion returns. If you're a competitor, the lesson is that the next earnings call needs a number next to the vision.