Meta Creates Nonvoting Class C Stock to Fund AI After Free Cash Flow Turns Negative

TL;DR

Meta reported third-quarter 2026 revenue of $64.60 billion on October 21, up 26% from a year earlier. In the same quarter its capital expenditures, including finance lease payments, reached $34.91 billion, and its earnings release puts free cash flow at negative $2.12 billion. Its fix is a new publicly listed, nonvoting Class C common stock. Every existing shareholder gets one Class C share per share they already hold, effective January 19, 2027, and after that Meta can sell Class C to raise money for AI. The information statement is direct about why it chose new stock: Zuckerberg would not back a large equity raise that cut his voting power, and debt on the scale management had in mind could have hurt Meta's credit ratings.

The quarter Meta's cash flow went negative

The ad business is fine. Average price per ad rose 14% year over year, ad impressions rose 10%, and 3.64 billion people used a Meta app on an average day in September. Revenue beat the $63.34 billion analysts expected, according to the AP, and Q4 guidance is $73-76 billion.

The spending moved faster. Operating cash flow grew from $30.00 billion to $32.79 billion. Capex including finance lease principal grew from $19.37 billion to $34.91 billion, about 80%. Subtract one from the other and a company that generated $10.63 billion of free cash flow in Q3 2025 burned $2.12 billion in Q3 2026.

Headline profit looks better than it is. Net income rose 264% to $9.86 billion, which sounds like a victory lap until you check that Q3 2025 carried an 87% effective tax rate. This quarter also includes a $9.89 billion charge for Meta's agreement with 52 state attorneys general over teen mental health, a deal that provides for payments of up to about $18.59 billion. Without that charge, operating income would have been $22.88 billion, up from $20.54 billion.

Full-year 2026 capex guidance moved from $130-145 billion to $135-145 billion. Meta gave no 2027 number but said it expects "to continue investing aggressively." Cash and marketable securities stood at $82.45 billion on September 30.

Buybacks stopped, debt arrived

The cash flow statement shows the same squeeze. In the first nine months of 2025, Meta bought back $26.25 billion of Class A stock. In the first nine months of 2026, it bought back none and took in $24.91 billion in net proceeds from new long-term debt. It still paid $1.36 billion in dividends this quarter.

Debt was the obvious next lever, and the information statement explains why Meta didn't keep pulling it. The independent Special Committee "recognized that debt financing alternatives of the magnitude that Company management was considering could have an adverse effect on the Company's credit ratings."

How Class C works

Meta already runs two classes: Class A trades on Nasdaq, and Class B carries extra votes. Class C adds a third that carries economic rights and no vote. The mechanics, from the filing:

• A one-time stock dividend of one Class C share for each outstanding Class A and Class B share, paid after the restated charter takes effect on January 19, 2027. The record and payment dates are still to be set.
• Meta expects Class A to trade as if it had a 2-for-1 split, with Class C priced at roughly the same level. Class A closed at $694.25 on October 20.
• Meta plans to list Class C on Nasdaq and says it "expects to conduct offerings of Class C common stock from time to time in the future," along with using it for employee pay and acquisitions.
• No existing holder's voting power or equity stake changes on the day of the dividend.

A useful way to picture Class C is a concert ticket with no backstage pass: you get the same seat, and the same share of the economics, but no say in the setlist. From January, Meta can print as many tickets as investors will buy without anyone new reaching the mixing desk.

Google pulled the same move in April 2014, paying a nonvoting Class C dividend that kept its founders in control. Meta's version adds a layer aimed at the founder himself.

The Zuckerberg clauses

The charter change needed no public shareholder vote. Zuckerberg and his Chan Zuckerberg Holdings entities held 315.8 million Class B shares, about 56.1% of the combined vote at the October 20 record date, and approved it by written consent. The filing calls that consent "the only shareholder approval required."

The filing also records the negotiation that made Class C the answer. The Special Committee looked at simply selling more Class A, "but Mr. Zuckerberg made it clear during the course of the negotiations with the Special Committee that he would not support a significant equity financing that would have the effect of reducing his voting power in the Company."

A separate Founder Alignment Agreement limits how Zuckerberg can cash out his new nonvoting shares:

• 20% free, 80% "stapled." He can sell 20% of the Class C he receives without restriction. For each stapled share he sells, he must within two business days also sell one Class A or Class B share, or convert a Class B share into Class A.
• Release by tenure. Another 2.5% of the stapled shares unlocks for each full quarter he serves full time on the executive team or as executive chair. All of it is released after 32 quarters.
• Doubling dividends. While he stays in that role, Meta pays a further one-for-one Class C dividend to all shareholders each time the stock doubles from its September 10, 2026 reference price, provided the internal rate of return over that period is at least 10%.

Stapling is the clever part. Without it, Zuckerberg could sell nonvoting shares for liquidity and keep every vote, ending up with control and very little money at risk. With it, a stapled share can only leave if a voting share is sold or converted alongside it, so his votes shrink as his cash-outs grow. Until the tenure releases add up, his liquidity and his control are tied together.

Moelis & Company advised the Special Committee. Morgan Stanley advised Meta.

"Before competitors saturate" the market

The filing's most revealing paragraph is about timing. The committee wrote that "timing is important to secure capital before competitors saturate various funding options," naming recent and expected raises at Alphabet, SpaceX, Anthropic and OpenAI. Moelis told the committee that Alphabet had recently announced an equity raise of about $80 billion.

The committee also says Meta "presently has sufficient capital resources for its current needs." The point is to have the instrument ready before the line at the bank gets longer. Management wanted Class C available "as early as the end of 2026."

What Muse has to earn back

The product all this money is for got a public user number on the call. Zuckerberg said more than 10 million people in the US and Canada have set up accounts for Muse, Meta's personal AI agent, and that about a quarter of them use it on a given day, according to the AP. Muse is expensive to run. Meta expects its cost efficiency to improve more than 10x over the next year, the AP reported, and says that would make the product profitable once revenue reaches scale.

The CFO outlook asks for patience. The investments of Meta Superintelligence Labs (MSL) "will take longer to generate returns than those in our core business." Reality Labs losses are expected to narrow to about $15 billion next year as money moves "primarily" to AI. Headcount was 66,775, down 15% year over year.

Investors didn't mind: shares closed up 0.5% on Wednesday and rose 2.9% after hours, per the AP.

Why builders should care

If you buy tokens from frontier labs or rent GPUs from clouds, this is a supply-side signal. The biggest buyers of compute have run past their own cash flow and are now competing for the same equity and debt dollars, and Meta's board said so on the record. Expect that competition to show up downstream, in what inference costs and how long capacity contracts run.

Muse's numbers are also a real data point for anyone building a consumer agent. Ten million signups with about 25% daily use is strong engagement, yet Meta's own plan to make it profitable depends on a 10x cost cut. Expect the same math in your own agent product: retention arrives before margins do.

Key Takeaways

• Meta's Q3 2026 revenue was $64.60 billion (+26%), but capex of $34.91 billion pushed free cash flow to negative $2.12 billion, against positive $10.63 billion a year earlier.
• Meta bought back no stock in the first nine months of 2026 and raised $24.91 billion in new long-term debt.
• A new nonvoting Class C stock arrives via a one-for-one dividend effective January 19, 2027, and Meta plans to sell Class C in future offerings to fund AI.
• The filing says Zuckerberg refused an equity raise that diluted his vote, and that more debt risked Meta's credit ratings. His group's 56.1% of the vote approved the change by written consent.
• 80% of Zuckerberg's own Class C is "stapled": selling it forces matching sales of voting shares until tenure releases unlock it over 32 quarters.
• Muse has passed 10 million accounts in the US and Canada, with about 25% daily use, and Meta is counting on a 10x cost-efficiency gain to make it profitable at scale.

Sources: Meta Q3 2026 earnings release (Form 8-K, Exhibit 99.1), Meta preliminary information statement on Class C common stock (Schedule 14C), AP: Meta Q3 revenue tops estimates and Zuckerberg says Muse accounts have topped 10 million, Meta Investor Relations, Meta: Introducing Muse, AFP via Phys.org: Google exits $1,000 stock club after special dividend (2014)