Meta's $900 Billion Class C Is Not a Second AI Security — the Spread, Not the Listing, Prices the Buildout

The board approved a new class of publicly listed, non-voting Class C common stock, to take effect with an amended and restated certificate of incorporation on January 19, 2027. One Class C share goes to every outstanding Class A and Class B share, payable no later than the fifth business day after the charter takes effect, and Meta expects the Class A price to reflect a two-for-one stock split once that dividend is paid. Two things this is not. It is not a fundraising, because no money changes hands on distribution. It is not a vote, because no shareholder's voting power moves. Class C is a financing shelf that starts life empty, with stated uses of capital raising, employee compensation, potential acquisitions and general corporate purposes.

Meta's third quarter was two businesses running in opposite directions. Advertising and other revenue reached $64.16 billion inside a $64.60 billion top line, up 26%, while costs and expenses rose 68% to $51.61 billion, with research and development alone at $20.05 billion. Strip out the $9.89 billion charge for the August agreement with 52 state attorneys general, which commits the company to payments of up to roughly $18.59 billion, and operating income was $22.88 billion — a 35.4% margin against 40.1% a year earlier. Diluted earnings per share of $3.83 landed well below the $5.70 consensus cited at the time. Advertising cash flow compounds while the buildout's margin claim does not, and that is the gap a new share class is meant to bridge.

Read the Founder Alignment Agreement carefully and the design becomes clear. Mark Zuckerberg's Class C shares are stapled to his Class B. He can sell them only by selling or converting a proportionate number of Class A or Class B shares, or by drawing on accrued Stapling Credits, and a breach is not a fine: the holder group is deemed to have irrevocably converted the same number of Class B into Class A. Meta has engineered a slow, pre-announced supply release instead of an overhang cliff, which is the difference between a discount that measures funding cost and one that merely measures founder intentions.

Before there was a Class C, there was a bond market that had already formed a view of the same risk. Meta's 5-gigawatt campus in Richland Parish, Louisiana sits inside a vehicle owned 80% by Blue Owl Capital and 20% by Meta. Meta put $5.8 billion of equity into it against Blue Owl's $23.0 billion, and the vehicle issued $27.3 billion of SPV debt that never appears on Meta's balance sheet — Meta converts a capital liability into a lease obligation that starts June 1, 2029. The paper priced at roughly 225 basis points over Treasuries, rated A, one notch below Meta's own AA-, with PIMCO anchoring and BlackRock participating. The bonds traded near 91 cents on the dollar by late September, a record low, with the 2049 maturity set against a four-year initial lease flagged as the structural flaw.

None of this is a demand problem. Meta's September-quarter spending became revenue for the companies that sell into the buildout. NVIDIA posted $96.22 billion of revenue and $63.73 billion of operating income in its fiscal second quarter ended July 26, 2026, with $89.0 billion of it from data center. SK hynix, whose American Depositary Shares began trading on Nasdaq in July 2026, reported won-denominated revenue of ₩79.32 trillion in its second quarter at a 76.3% operating margin and carries a $992 billion market value on those ADSs. Micron's fiscal fourth quarter ended September 3 delivered $54.23 billion of revenue, up 30.8% from the prior quarter, at an 86.8% gross margin. The risk is being pushed down the capital chain, not back up it.

Nasdaq rewrote the Nasdaq-100's float rules effective May 1, 2026, replacing the old 10% minimum-float test with a graduated scale: below 33⅓% float, a company's index weight is a fraction of its market capitalization that rises with float, and above it the company is represented at full listed market capitalization. The same notice states that eligibility considers listed and unlisted shares while weighting stays on listed shares, an explicit framework for the multi-class era. Applied to Meta, the arithmetic is anticlimactic. Because the Class C distribution is one-for-one and pro rata, and because the stapled founder block is carved out of Class B which already sat outside the float, Meta's float ratio barely moves. Passive index weight does not change: float-adjusted market capitalization stays near $1.56 trillion on both sides of the dividend.

Meta is not the first to fund AI with equity. It is the first to do it with a new class of shares. The sequence is clear. Wave one was debt: incremental annual debt issuance rose from 9% of capital spending in fiscal 2024 to 32% on a trailing basis by mid-2026, with the six largest hyperscalers carrying roughly $700 billion of debt and about $820 billion of lease commitments that never reach the balance sheet. Wave two was primary equity: Alphabet priced $84.75 billion in June 2026, the largest equity transaction by a listed company, with $44.75 billion earmarked for general corporate purposes including AI capex. Class C is the third wave, and the first to create a tradable instrument rather than sell a block of stock.

Days to quarters. The charter takes effect on January 19, 2027, Meta will set the record and payment dates later, and Class C will begin printing against Class A almost immediately. The first meaningful spread will come from the composition of the two order books, not from the dividend mechanics. Three prints decide whether this is a financing or a curiosity: Meta's fourth-quarter capital spending against the $135-145 billion guide, Alphabet's October 28 report and its 2027 commentary, and SK hynix's October 28 earnings call, which tells the market what high-bandwidth memory is worth in the same 24 hours. Those three prints settle the question of whether the market will pay for AI-buildout equity.

One to three years. Meta gave no 2027 revenue outlook and said only that it expects to sustain strong overall revenue growth through the year, with Reality Labs losses narrowing to about $15 billion as spending is redirected toward AI. It also put no ceiling on 2027 capital spending. The gap between the $140 billion midpoint of the 2026 guide and the roughly $196 billion consensus for 2027 is about $56 billion of funding that has to come from Class C, project finance or new bonds, and Class C is the only one of the three with no coupon, no maturity and no covenant. Whoever subscribes to Class C is underwriting the 2027 buildout without a contractual return.

This argument is deliberately narrower than a call on an AI bubble. The observable claim is narrow: a mega-cap with negative free cash flow, a halted buyback, a raised spending floor and $112.1 billion of debt has pre-created a non-voting equity class, and the founder agreement has been built so the founder cannot flood it. The next honest read on the cost of AI infrastructure equity will be the Class C discount to Class A, and it will be public, daily and hard to spin.