The question investment committees keep circling this week is not whether artificial intelligence demand is real. It is who can afford to pay for it without flinching when borrowing costs spike. Meta’s October 21 earnings report, scheduled after the close, lands at exactly the right moment to answer that question.
Markets spent this week aggressively reassessing the cost of capital behind AI infrastructure expansion, with the Nasdaq Composite falling about 2.1% on Friday alone as investors concluded that demand growth no longer dispels cash-flow anxieties. AI-related corporate debt issuance had already cooled from roughly $113 billion in June to $23 billion in September, according to Morgan Stanley data cited by the Financial Times. Yields on CoreWeave’s bonds pushed above 15% this week as the company’s debt-funded model came under direct scrutiny. That is the backdrop Meta walks into when it reports.
What the Numbers Say
Meta is scheduled to report Q3 2026 results on Wednesday, October 21, after the close. Revenue growth of roughly 23% year-over-year would be solid. But the headline figures are not the point. Investors will closely track expenses and capital spending, because Meta’s Q2 earnings miss came as costs surged, with management guiding to full-year expenses of $165 billion to $169 billion and full-year capex of $130 billion to $145 billion.
Two weeks before earnings, Zuckerberg told a podcast audience on October 12 that Meta was prepared to commit between $100 billion and $200 billion to AI infrastructure before knowing whether its models would succeed. To support those ambitions, Meta is constructing high-capacity computing facilities, including a “Prometheus” cluster in Ohio on the order of a gigawatt and a planned “Hyperion” cluster in Louisiana designed to scale to about five gigawatts. Q3 is the first chance management has to put specific quarterly spending figures alongside that vision.
The Bull Case: No Lender Required
Here is the structural argument that separates Meta from the rest of the AI financing debate. Meta’s position within the hyperscaler group is structurally different: Amazon and Microsoft earn cloud revenue that partially offsets their buildout, while Meta’s income is advertising, meaning every dollar of capex must be justified by what advertising, subscriptions, or transaction fees eventually produce. That sounds like a liability. Bulls argue the opposite.
Zuckerberg has pointed out that the company’s highly profitable core business and strong credit allow Meta to underwrite the costs itself. As of June 30, 2026, Meta held $90.26 billion in cash, cash equivalents, and marketable securities. CoreWeave, by contrast, has seen stress show up in credit insurance markets, with five-year credit default swaps jumping to around 1,100 basis points this week. Meta is the customer, not the borrower, in that relationship.
The Bear Case: Free Cash Flow Is Almost Gone
The concern is equally concrete. Q2 capex hit $31.08 billion, consuming nearly all operating cash flow, and shares fell sharply after-hours as Meta missed EPS estimates despite 28% revenue growth. Free cash flow for the quarter came in at just $784 million.
Sell-side projections now anticipate Meta’s free cash flow turning negative in 2026 and worsening in 2027. Some analysts, including skeptics at firms like BMO, argue Meta shows the least visible return on AI spending among its peers. Analyst confidence has softened heading into the report, with consensus estimates moving lower over recent weeks.
What Investors Are Missing
The overlooked angle is not Meta’s capex itself. It is what a confirmed spending guide does to the two hardware names most exposed to it. Broadcom said AI semiconductor revenue was $5.2 billion in Q3 of fiscal 2025 and later reported $16.7 billion in Q3 of fiscal 2026. Broadcom’s bull case rests partly on how quickly hyperscaler custom silicon programs ramp. If Zuckerberg raises the top end of 2027 capex guidance, Broadcom’s order book gets a direct lift.
Stocks to Watch
Meta (META) is the test case. A Q3 capex figure above $33 billion alongside an upward 2027 spending comment would likely push free cash flow negative and force allocators to decide whether to treat the AI buildout as capital destruction or strategic investment. The answer is not obvious.
Nvidia (NVDA) remains the most direct hardware beneficiary of every dollar Meta commits. Nvidia has long dominated AI training accelerators, with many analysts putting its share near 90% in that segment, even as inference has more alternatives. A confirmed Meta acceleration shores up Nvidia’s forward demand picture at a week when it badly needed reassurance.
Broadcom (AVGO) is the quieter read-through. Broadcom reported 221% AI semiconductor revenue growth in its most recent quarter, and has recently traded around 19x forward earnings by some market data services. Meta’s custom silicon commitment sits directly inside Broadcom’s pipeline. A stronger-than-expected Meta guide is an indirect earnings revision for AVGO before it reports its own results.