Negative $2.12bn Free Cash Flow (Meta Q3 2026, 21 October): Why AI Capex Outran the Cash — and What the New Non-Voting Class C Stock Is For

Meta's third quarter brought in $32.79 billion of operating cash and spent $34.91 billion on capital expenditures, so free cash flow came out at negative $2.12 billion — a quarter in which the AI build-out cost more than the business generated. In the same release, Meta created a new class of non-voting Class C stock and described it in so many words as a way "to finance its growth". The two announcements are connected, and that connection is the part the headline beat on revenue doesn't explain.

What Meta reported

The figures below come from Meta's third-quarter press release as filed with the SEC on 21 October 2026; consensus comparisons are from CNBC's results coverage, citing LSEG.

The reported numbers look worse than the business because of one item. In August Meta agreed with 52 state attorneys general to settle litigation over teen mental health and social media, with payments of up to about $18.59 billion; the third quarter carries a $9.89 billion charge for it. Strip that out and operating income was $22.88 billion and diluted EPS $7.40 — Meta's own reconciliation, not an analyst adjustment.

The operating detail was strong on price and modest on users. Ad impressions rose 10% and the average price per ad rose 14%; daily active people across the apps were 3.64 billion, up 3%. Headcount fell 15% to 66,775. CEO Mark Zuckerberg said in the release that "our investments are paying off, the core business is strong." That is the case the company is making; the cash-flow statement shows what it costs.

Why free cash flow went negative

Free cash flow is the cash the business generates minus what it spends to keep and grow its physical assets. Meta defines capex to include principal payments on finance leases — data-centre capacity it leases rather than buys — because, as the release puts it, together they represent the equipment it needs "regardless of whether we procure" it outright or by lease.

The arithmetic for the quarter is short:

The business didn't shrink. Operating cash flow grew about 9% year on year. What changed is spending: quarterly capex of $34.91 billion is roughly 80% above the $19.37 billion of a year earlier. For the first nine months, free cash flow was $11.05 billion, against $29.51 billion in 2025.

The forward guide matters more than the quarter. Meta narrowed its 2026 capex range to $135–145 billion from $130–145 billion — a higher floor, against an analyst expectation of about $138.3 billion per CNBC. Nine months have already used $85.82 billion. Subtract that from the range and Q4 capex comes to roughly $49–59 billion. That is arithmetic from Meta's own figures, not guidance for the quarter, but it puts one quarter's spending well above Q3's $32.79 billion of operating cash. On those numbers the cash gap gets wider before it narrows, and the settlement payments are still to come.

Three ways to pay, and how the mix has shifted

Any company spending faster than it earns has three ways to cover the difference: run down cash, borrow, or issue equity. Meta has now moved toward the second and third while it cuts back the cash it hands to shareholders.

The buyback line is the clearest change. A year ago Meta was using surplus cash to retire shares; this year it has not bought back any. Debt has risen by about $22 billion since the end of 2025. This is the same move this site traced at SpaceX's $40 billion chip financing — the AI build-out moving from being paid for out of earnings to being paid for with outside money — now at a company that until recently generated more cash than it could use.

What the Class C stock is for

The new share class is the third line. Meta's board approved a publicly listed, non-voting Class C common stock, effective 19 January 2027. Every existing Class A and Class B share will receive one Class C share as a one-time dividend; Meta expects the Class A price to then reflect the effect of a two-for-one split, and says the dividend won't change anyone's voting power. Meta intends to list the Class C shares on Nasdaq.

The release gives the reasons directly. The first is "creating a new class of shares for capital raising, employee compensation, potential acquisitions, and other general corporate purposes", because "the scale of investment required to compete" in AI is significant. The release also says the structure "will continue the founder-led model," and Meta disclosed a Founder Alignment Agreement with Zuckerberg covering his sales of Class C stock, his service at the company and conditions for future Class C dividends.

The mechanism is the separation of money from control. Ordinary shares carry both an economic claim and votes. Each time a company sells voting shares to raise money, or pays employees in them, voting power spreads a little further. A non-voting class keeps the economic claim and drops the vote, so equity can fund growth without moving control. It isn't new: Alphabet created non-voting Class C shares through a stock dividend in 2014.

What matters for shareholders is what happens after the dividend. The dividend itself moves no value — each holder simply has twice as many shares, each worth roughly half as much. Any later issue of new Class C stock, whether sold for cash, used to pay for an acquisition or paid to employees, adds to the share count. A company that has spent years shrinking its share count now has a tool built to grow it.

Where it reaches the indices

The concentration of the US indices is what turns one company's funding choices into a market story. There are two channels.

Demand for hardware. Meta's capex is revenue for its suppliers — chip designers, memory makers and equipment makers that carry large weights in the NAS100 and US500. A higher floor on the 2026 range, and an implied Q4 of $49–59 billion, is more spending for them this year. The memory side of that chain has been volatile all month; see the CXMT selloff in memory stocks. Meta's comment that it expects to keep "investing aggressively" in 2027, with no budget yet, leaves next year's number open.

Supply of shares. For years the largest US companies have been reliable net buyers of their own stock, and that bid has been part of what supports index valuations. Meta's buyback line going from $26.25 billion to zero in nine months, and a share class designed partly for raising capital, cut against that. One company won't move the US500 on this alone. But the same pressure — spending that grows faster than operating cash — applies to the other hyperscalers, and the data-centre lease commitments covered in Microsoft's $329 billion of leases show how big those obligations already are.

What would change the picture, without implying a direction:

• Next week's reports. Alphabet and Microsoft report on Wednesday 28 October and Amazon on Thursday 29 October, per CNBC. Whether their free cash flow holds up against their own capex is the cross-check on whether Meta's quarter is unusual or typical.
• The 2027 capex budget. Meta said it is still building its plan. A number much above 2026 would widen the cash gap; the release's expectation of Reality Labs losses narrowing to about $15 billion next year offsets part of it.
• How the Class C stock is used. The record date, the payment date and the first issuance beyond the dividend will show whether the class is mainly for compensation or for raising capital at scale.

What it does — and doesn't — mean for the dollar

This is not a currency story, and it is worth saying so. Meta's guidance assumes foreign exchange is about a 1% headwind to Q4 revenue growth, which is a small translation effect on a company that earns much of its revenue outside the US — a consequence of the US dollar's level, not a cause of it. The channel to currencies, if there is one, runs through risk appetite: a large AI-led swing in US equities shows up in the risk-sentiment factor the meter already scores, and in the funding currencies that move with it. For how this site approaches stories like this one, see about.

The narrow conclusion is the useful one. Meta's core business grew faster than expected, and its spending grew faster still. This quarter the gap showed up as a negative free-cash-flow print, and the company answered it with more debt, no buybacks and a new share class for raising money. None of that says whether the spending will pay off. It shows that the AI build-out is now big enough to change how its largest buyer pays for things.

Educational macro context only — not investment advice.