Everything About AI
NextFin News - Meta Platforms is creating a new class of non-voting stock, a governance shift unveiled alongside its third-quarter earnings report that is designed to finance the company's artificial-intelligence spending and future acquisitions without weakening Chief Executive Mark Zuckerberg's control. The move, paired with a 2026 capital expenditure outlook now set at $135 billion to $145 billion, marks the most consequential restructuring of voting power at a mega-cap technology company since Alphabet created its own non-voting shares in 2014.
The question the proposal raises is not whether Meta needs more capital. It does, in quantities that are beginning to strain even its formidable cash generation. The question is what kind of company investors are buying. Meta is effectively telling the market that its AI ambitions are too large to be funded from operations alone, yet its founder's control is too important to be shared. That is a bet on permanent founder rule, not a temporary financing bridge.
The Deal: Non-Voting Shares as an AI Funding Tool
Under the plan, Meta would issue a new class of stock that carries economic rights but no voting power. The shares could be used to pay for acquisitions or to raise cash for data centers, servers, and networking gear without touching Zuckerberg's voting stake. He owns roughly 13.5% of Meta's outstanding shares but controls about 61% of the voting power through super-voting Class B stock - a gap that already makes Meta a "controlled company" under Nasdaq rules.
The announcement came Wednesday, October 21, 2026, after the market close, alongside results that showed just how expensive the AI build-out has become. Revenue rose 26% to $64.60 billion, ahead of the roughly $63.3 billion consensus and above the top of the company's own $61 billion to $64 billion guidance range. But total costs and expenses jumped 68% to $51.61 billion. Operating income fell 37% to $12.99 billion, and the operating margin compressed to 20% from 40% a year earlier. Diluted earnings per share came in at $3.83 - well below the $6.76 analysts expected - largely because of a roughly $10 billion legal accrual tied to a multistate settlement over youth-safety claims.
Free cash flow, the metric investors watch most closely on AI spending, turned negative for the quarter. A year earlier, Meta was converting its ad revenue into billions of dollars of free cash every three months. Now the infrastructure bill is arriving faster than the revenue it is meant to generate. By the company's own disclosure, Meta has already spent about $85.8 billion on capital expenditures through the first nine months of 2026 - meaning the full-year $135 billion to $145 billion target requires another roughly $50 billion to $60 billion in the final quarter alone, a pace that would make 2026 the most capital-intensive year in the company's history.
The spending trajectory has ratcheted up all year: $115 billion to $135 billion at the start of 2026, raised to $125 billion to $145 billion in April, tightened to $130 billion to $145 billion in July, and now $135 billion to $145 billion. That top end is roughly double the $72.22 billion the company spent in 2025 - a 101% year-over-year jump that exceeds the annual military budgets of Germany and India. Analysts are modeling roughly $193 billion to $205 billion for 2027, with Bank of America at the top of that range at $205 billion. Total 2026 expenses are guided to $165 billion to $169 billion.
Meta still sits on about $90 billion of cash and marketable securities against roughly $84 billion of long-term debt, giving it borrowing capacity. But debt is not free, and equity issued with voting rights would dilute the founder's control. Non-voting shares offer a third path: raise permanent capital, keep the voting block untouched, and pass the economic risk of the AI build-out to public shareholders who get no say in how it is deployed.
The Alphabet Precedent: A Template From 2014
This is not without precedent. In April 2014, Google executed a 2-for-1 stock split that created non-voting Class C shares (ticker GOOG) alongside the voting Class A shares (GOOGL). The stated purpose was identical to Meta's logic: allow the company to issue shares for acquisitions and employee compensation without diluting the voting control of founders Larry Page and Sergey Brin, who held super-voting Class B stock.
The mechanism worked as designed. Existing Class A holders received one Class C share for each share owned, instantly doubling the share count while leaving voting proportions unchanged. Today, Class C trades at a slight discount to Class A - a persistent "governance discount" that quantifies what the market charges for shares with no voice.
Meta's version differs in one important respect. Google created its non-voting class through a split that rewarded existing shareholders. Meta, by contrast, is creating a class it can issue going forward - to sellers in acquisitions or to new investors - which transfers voting dilution risk entirely onto future holders rather than spreading it across the current base.
Why Control Is the Real Asset Here
To understand why Meta would rather create a two-tier capital structure than issue ordinary shares, look at what control is worth to Zuckerberg. With 61% of the vote on about 13.5% of the economic interest, he can set strategy, hire and fire the board, and block activist campaigns unilaterally. Nasdaq's "controlled company" designation exempts Meta from certain board-independence requirements, though the company voluntarily keeps an independent majority and independent committees.
That concentration of power cuts both ways. It let Meta commit to a multi-hundred-billion-dollar AI build-out without the quarterly pressure a dispersed shareholder base might impose. It also means the person deciding how much to spend answers to no one. Zuckerberg has framed the spending as existential.
As AI usage in our products and businesses continues to ramp, we continue to invest aggressively in infrastructure to meet the demand.
That was Zuckerberg on the second-quarter earnings call, where he added that a "significant portion" of compute would go toward training models and growing the core business. The board that hears that statement is the board he appoints.
The market has rewarded the AI story when it sees revenue coming back. Shares rallied 27% in September - the best month since 2022 - after the debut of the Muse personal AI agent, which topped Apple's App Store charts and introduced paid subscription tiers. But revenue growing more than 25% was not enough to stop the stock from falling 8.6% and 8.0% after the previous two reports, both times on spending fears rather than top-line weakness.
Wednesday's report shows why those fears persist. Meta sold 26% more advertising and other services than a year earlier, and advertisers paid 14% more per ad on 10% more impressions - the core engine is firing. But costs rose 68%, and the company kept just 20 cents of each sales dollar as operating profit, down from 40 cents. The ad business is growing; the AI bill is growing faster.
The scale of that bill is what separates Meta from its peers. Bank of America's tally puts 2026 capex at $135 billion for Meta, versus $185 billion for Alphabet, $190 billion for Microsoft, and $200 billion for Amazon - a combined four-company outlay approaching $710 billion in a single year. No public company has ever sustained spending of that magnitude without either diluting shareholders or leaning hard on debt. Meta is now building a structure that lets it do neither, at least not in the way markets are used to.
The Second-Order Problem: Who Bears the Risk, Who Captures the Reward
The first-order reading of this move is simple: Meta needs money for AI, and non-voting shares are a clever way to get it. The second-order question is harder, and it is the one the market has not fully priced. If Meta funds its AI build-out with non-voting equity, the economic claim on those assets grows faster than the voting claim. Over time, a company can end up with a small voting class that controls a capital structure whose economic risk is held almost entirely by outsiders.
That structure creates a specific agency problem. The controlling shareholder enjoys the upside of a successful AI bet - and the reputational upside of building a leading AI company - while the non-voting holders absorb the dilution if the spending does not earn its keep. There is no historical template for a multi-hundred-billion-dollar AI infrastructure program funded this way, so there is no evidence that non-voting holders will be made whole.
There is also a market-structure consequence. If a meaningful share of Meta's roughly $1.78 trillion market capitalization migrates into a non-voting class, the voting shares could become less liquid and more volatile, while the non-voting class trades at a widening discount. Index providers and governance-focused funds may be restricted from holding non-voting shares, shrinking the buyer base for the new class and raising Meta's cost of capital on exactly the instrument it is creating to lower it.
And there is a precedent for the discount becoming permanent. Alphabet's Class C shares have traded below Class A since their creation. Meta's non-voting class, if it carries no path to conversion and no sunset provision, could see that discount deepen during any period when AI returns disappoint - which is precisely when the company would most want to issue more of them.
The Counter-Case: Founder Control Is Why Meta Can Think in Decades
The strongest argument for this structure is also the simplest: it is why Meta can spend $145 billion in a single year on infrastructure whose payoff may take a decade. Public companies led by dispersed shareholders rarely make decade-scale bets; they make quarter-scale bets. Zuckerberg's control insulates Meta from the pressure to cut AI spending when the next quarter's free cash flow looks thin.
Supporters would point to the alternative. If Meta funded AI through ordinary equity, Zuckerberg's voting stake would dilute toward 50%, then below it, and an activist or hostile bidder could eventually force a change in strategy. If it funded through debt, interest costs would eat the cash flow that currently supports a growing dividend and buyback program. Non-voting equity is the only instrument that finances the build-out without mortgaging either control or the balance sheet.
There is also a real possibility that the market barely notices. Alphabet's Class C shares were absorbed without a sustained governance revolt, and Meta's stock has rallied about 37% over the past year. If the AI spending produces visible revenue - better ad targeting, higher pricing per ad, new paid products - the governance question becomes academic. Shareholders forgive a lot when the return on invested capital is compounding.
What to Watch: The Falsifying Signal
This thesis - that Meta is locking in permanent founder control while socializing the economic risk of its AI bet - rests on one assumption: that the non-voting class is used meaningfully. The falsifying signal is straightforward. If Meta announces the new class but does not issue a material amount within four quarters - if it funds the $135 billion to $145 billion 2026 capex and the modeled 2027 spend from cash flow and debt instead - then this is a contingency tool, not a structural shift, and the governance concerns are overstated.
Conversely, if the company begins issuing non-voting shares for acquisitions or capital raises in 2027, the discount between the two classes and the concentration of voting power become the central governance story of the mega-cap technology complex.
The other signal to watch is the 2027 capex number. The Street is modeling roughly $193 billion to $205 billion. If Meta guides anywhere near that range while free cash flow remains negative or near zero, the funding question moves from theoretical to urgent - and non-voting shares become the most likely answer.
Outlook: Three Scenarios
Base case: Meta announces the new class, uses it sparingly in 2026 for small acquisitions, and funds most capex from cash flow and debt. The stock trades on the AI revenue story, and the governance issue simmers. Time horizon: 12 months.
Upside case: AI-driven ad tools lift revenue growth above 25% and the Muse agent begins generating paid subscriptions at scale. The non-voting class is absorbed easily, the discount stays narrow, and Meta's control structure is vindicated as the price of a decade-scale win. Time horizon: 24 to 36 months.
Downside case: 2027 capex approaches $205 billion, free cash flow stays depressed, and Meta issues a large tranche of non-voting shares. The discount widens, governance-focused funds exit, and the stock re-rates on both lower returns and a higher governance risk premium. Time horizon: 12 to 24 months.
Meta's non-voting stock is not a financing footnote. It is the clearest signal yet that the AI build-out has outgrown what the advertising business can fund - and that the founder who started that build-out has no intention of letting anyone else steer it.
Explore more exclusive insights at nextfin.ai.
