Skip to content
Some content is members-only. Sign in to access.

Who Funds Meta’s AI Build-Out—and Who Keeps the Vote?

New capital broadens investor exposure while Class C shares preserve Zuckerberg’s voting power.

By KAPUALabs

Meta is raising outside money on a very large scale, and it has arranged matters so that the new money carries no say in how the company is run. That is the plain conclusion of the November 2026 record. A profitable advertising business is paying for an expensive AI build-out. The build-out is now big enough to need new equity, preferred stock and bonds. The share structure is being reworked at the same moment so that Mark Zuckerberg's voting control does not move. Whether the spending earns a return is still an open question. Who carries the dilution, and who keeps the authority, is not.

The Engine and the Appetite

The funding base is advertising. More than 97% of revenue comes from digital advertising 5,6,9,39,42,67. Third-quarter revenue was $64.6 billion 51, up 26% year over year 43. Ad impressions rose 10% 56. A $9.89 billion legal charge weighed on earnings all the same 41.

The appetite is larger than the engine. Zuckerberg reportedly confirmed Meta's readiness to spend $100B–$200B on AI infrastructure 40. Quarterly free cash flow is cited as –$2.1 billion 61, against Q3 operating cash flow of $32.8 billion 58. The rise in debt is attributed largely to leased data-center capacity 52.

Meta has therefore gone to the capital markets, in several forms:

The totals quoted for the programme do not agree. One post says Meta is raising $37.5 billion in equity 53. Pershing Square's letter cites a $98B equity-raise figure that it says is only 6% of shares outstanding 54. One source expects 15% to 20% dilution by 2029 64. These probably measure different things, such as completed versus potential issuance, but the material does not settle it. A company that cannot give its owners one consistent number for a raise of this size has a disclosure problem. One source puts the result bluntly: outside shareholders are funding Meta's AI spending 21. A defender, Joseph Carlson, calls the financing transparent and standard for profitable tech firms 40.

Control Without Ownership

The funding works because Zuckerberg's authority does not depend on his stake. He holds roughly 13% of the economics 2,3,7,40 and about 60% of the voting power 1,2,3,4,6,8,10,11,18,40. Other sources put the vote at 60.6% 21,22 and 61% 62,65. Class B shares carry 10 votes each 29, and he owns 99.7% of them 15,16. An older article gives much lower voting figures as of 2023 20. The recent figures are far more widely corroborated, so I prefer them.

The new non-voting Class C stock is the device that keeps this arrangement intact. Its stated purpose is to allow future equity raises without diluting his voting control 46. The deal record shows what the device avoids. Moelis estimated that Meta could raise about $700 billion over years using Class A shares before Zuckerberg's voting power fell below 50% 29. Zuckerberg estimated that selling Class B shares would reduce his voting power to about 53% 29. Non-voting stock removes that ceiling.

The approval process also deserves attention. Class B approval alone is sufficient, with no separate Class A vote 29. Class B also holds a separate-class veto over change-in-control transactions until it falls below 35% of combined voting power 29. A railroad commissioner would recognise the pattern. The party who benefits from the rule is also the party who approves it, and the public shareholders are told afterwards.

A Split in Form, a Control Instrument in Substance

The mechanics are well corroborated and recent. The Board approved the Class C stock on 19 October 2026 32. The certificate takes effect on 19 January 2027 32. Meta declared a stock dividend of one Class C share per Class A share 32. Because each existing share gains a Class C counterpart with the same economic rights 16, several outlets call the result effectively a 2-for-1 split 14,27,34. Meta itself expects the Class A price to reflect the effect of a two-for-one split after the payment date 32.

The same coverage stresses that this is not a traditional split 32. The new shares carry no votes except as required by law or the charter 16,29. Meta said the dividend produces no change to voting power 32. Both statements are true, and together they describe the design. Shareholders see a familiar price adjustment. What they receive is a permanent control feature.

I treat the "locks in control" language some commentators use 15,62 as characterisation, not finding. The charter terms and the Moelis and Zuckerberg estimates support the same point without the adjective.

The Stapling Dispute

The sharpest question is what happens when the founder sells. Of Zuckerberg's Class C holdings, 20% are immediately freely saleable and 80% are stapled 29. The saleable 20% equals about 10% of Meta's overall economic stake 29. Selling a stapled Class C share requires, within two business days after settlement, an offsetting sale of Class A or Class B, a Class B-to-A conversion, or the application of Stapling Credits 29. Credits accrue 1:1 per Class A or B share sold, excluding compliance conversions 29.

Plaintiffs seek to represent a class of Class A stockholders 16. They allege that Zuckerberg could keep about 60.6% voting control after selling the Class C shares he receives 16. The suit concerns a transaction in which a controlling stockholder allegedly received unique benefits 16. These are allegations, and they are untested. Meta's position, as supplied, is that voting power is unchanged 32.

The two sides are not contradicting each other so much as describing different sales. Zuckerberg's own estimate of about 53% applies if he sells Class B 29. The plaintiffs' 60.6% applies if he sells Class C 16. The dispute is whether Class C sales let him turn ownership into cash without the voting erosion that Class B sales would cause. Management and Moelis estimated a 5–20% discount for Class C shares absent an immediate dividend 29. That is a measure of how much the market might price the missing vote.

Two single-source claims call for caution. One says over 90% of voting shares are held indirectly through CZI-affiliated fiduciary vehicles 20. It does not fit the quantified figures elsewhere, such as pledged Class B shares at about 2.1% of voting power as of April 2026 29, and I treat it as unverified. Another says Class C fails the Section 368(a) voting-stock requirement because it is non-voting 29. If that is right, Class C may serve less well as acquisition currency than the headline rationale implies.

Insider selling is visible but modest against this backdrop. Zuckerberg sold 27,474 shares worth about $21.4 million over six months 33. Insiders overall sold 150,973 shares worth $103.1 million over the last quarter 44. One source describes insider sentiment as neutral 30. The routine sales are not the concern. The concern is the stapled Class C holdings, which will govern his future sales.

Leadership turnover adds execution risk to a capital-heavy plan. Meta is losing its longtime head of infrastructure, Santosh Janardhan 38. Former AWS executive David Brown is slated to take over 41, with the change reported for early 2027 25. CNBC placed the transition amid rising investor scrutiny of capital spending at Meta and its large technology rivals 37.

The legal docket also bears on how well management is held to account. Italy's Garante imposed a €805 million penalty 35, and Meta plans to appeal 68. The EU preliminarily found a Digital Services Act breach over addictive design in July 35. Meta settled with several US states at the end of August over the same design concerns 45. A US judge has certified four classes of Facebook users 23, and a trial in a Washington, DC data privacy case is scheduled for July 50. Each item looks small against revenue. Together they are a recurring cost.

What the Market Is Saying

The market has accepted the growth story without fully accepting the bill. JPMorgan's Doug Anmuth holds a $940 target 17,31,34,43,57 and added Meta to the Analyst Focus List in place of Alphabet 60. The stock closed at $628.77 on November 11 34,47. That is well below its 52-week high of $779.82 17. The stock was down 18.22% over 30 days, with a one-year total shareholder return of just 2.60% 48. One post attributes an 8% one-day drop to concerns about the scale of spending and the timeline for profitability 52.

Returns on the new revenue lines are not yet disclosed. Muse and Meta Superintelligence Labs profit is not disclosed on a standalone basis 26. Zuckerberg said at the Q3 briefing that he would not discuss profit until revenue scale ramps up 36. He has said Muse could become the largest revenue source, with transaction fees rather than ads 12. Most of the enterprise and Muse claims rest on single X posts, so they are directional only.

Dilution Arrives in Stages

The share-count effects come in layers, and the timing matters. The Class A leg of about $25bn was done through forward sale agreements covering 40m shares 32. Underwriters later took a further 6m shares 28,32. This is consistent with forwards on 46m new Class A shares settleable at any time up to November 2027 49. One claim says the forward involves no newly issued shares 26, while another warns that new shares from forward sales may dilute 32. I read these as differing on timing. Borrowed shares are sold up front, and new shares arrive only when the forwards settle. That is my reconciliation, not a stated finding.

The preferred leg is a 6.50% Series A mandatory convertible 26. It converts in August 2029 24, at 0.0681 to 0.08 Class A shares per depositary share 29. Holders receive more common shares if the stock price falls 55. Diluted share count was 2.573bn in Q3 24, and the Class C dividend would roughly double the figures 32. The record shows no observed market effect of Class C, because the shares have not yet traded.

The Judgment

Meta's operating strength is real. Its governance is a different matter. The company is asking the public to supply capital for a very large, unproven bet. It has built the structure so that the people supplying it cannot influence the bet. Zuckerberg holds about 60% of the votes on roughly 13% of the economics 1,2,3,4,6,7,8,10,11,18,40, and the instrument chosen for the next stage of financing exists to keep it that way 46. Outside investors carry the dilution and the execution risk. The authority to judge the bet stays with one man.

That is not necessarily unlawful, and the stapling dispute is still to be tested. It is untenable as a long-run bargain, though, unless the terms are fully disclosed. I recommend three things:

Until then, investors are being asked for trust in place of information. A railroad commissioner would have called that a rate without a published schedule.

Takeaways

More from KAPUALabs

See all
| Free

Meta’s AI Build-Out: Ad Gains Versus Funding Strain

By KAPUALabs
/
| Free

Italy’s Meta Ruling: Consent, Remedy and Regulatory Reach

By KAPUALabs
/
| Free

Meta Bull Case Holds on Ads, Bear Case on Spending

By KAPUALabs
/