Meta has found a way to make investors bear the financial risk of an AI empire while Mark Zuckerberg keeps the controls and the people whose data built the network carry the privacy harm. The Wall Street Journal reported that a new class of Meta stock is scheduled to take effect in January 2027 without affecting Zuckerberg’s voting control. Its purpose is plain: widen the capital base behind the AI buildout without changing who runs the company. That is the elegant part, and the betrayal.

Give the structure its due. Meta is not short of cash. It reported record third-quarter revenue of $64.6 billion, up 26 percent from a year earlier. AI infrastructure—chips, data centers, power contracts, specialized buildings—demands enormous sums before its returns become visible. The scale makes ordinary cash-flow planning look quaint. A separate stock lets Meta mobilize outside capital without treating every new dollar of infrastructure as something existing shareholders must approve on their own, while leaving Zuckerberg’s command of the strategy untouched. That is the strongest honest case for the arrangement.

The latest spending figures give investors reason to pause. Wall Street expected capital expenditures to accelerate from the preceding quarter’s $5.8 billion. Instead, they declined to $4.7 billion, though that remained more than twice what Meta spent a year earlier. This is not a spending curve ripping upward. Calling an AI buildout “massive” is easy; showing the money in capital expenditures is harder. The new stock makes a future spending plan a present-day financing story.

AI infrastructure has turned capital expenditure from a supporting line item into the central competitive fact. The company that can finance the longest can make waiting for its results look like a strategic choice rather than financial distress. Meta’s new stock is an acknowledgment of that reality: revenue gives it room, but the buildout may consume that room faster than ordinary cash flow can replenish it. The January 2027 effective date tells investors when the new structure takes effect. It does not tell them what return to expect.

The architecture is not peculiar to Meta. In July, Elon Musk called 2026 a “massive Capex year” for assets such as AI computing power. As Tesla beat third-quarter revenue forecasts and Musk floated a SpaceX merger, Tesla’s revenue was flat while Meta’s grew by 26 percent. Musk’s merger talk signals a larger corporate architecture; Meta’s new stock is a more controlled version of the same play—mobilizing capital while leaving Zuckerberg in charge. Neither gives investors a disclosed return.

I know this separation from the inside. I once traded agricultural futures in Chicago, where the farmer still bore the rain and soil while I watched the price move on a screen. The contract made risk look portable. It did not make the corn less real. Meta’s financing structure is a more sophisticated version of the same maneuver: risk is made to look portable across a wider capital base, while the corresponding authority remains concentrated at the top.

The trouble begins where capital separates from authority. Capital is fungible. Trust is not.

The shareholders and other investors financing Meta’s expansion share the economic consequences of its AI strategy. Zuckerberg retains the votes that determine the strategy itself. If the spending produces a dominant new computing platform, he captures the upside while preserving his control; if it produces expensive, obsolete infrastructure, the financial burden is distributed more broadly. New investors can fund the bet, but they cannot use their capital to displace the man who directs it. That separation is the whole trick.

It is corporate finance with teeth.

Control without disclosure is not stewardship.

Then there is the privacy judgment.

Hours before Meta reported its results, a New Mexico judge ordered the company to pay the state $25 billion for misleading users about its data-privacy practices. The state called it one of the largest verdicts in American history. The judgment shows what scale does to liability as well as infrastructure. Meta can separate capital from voting control; it cannot separate infrastructure from legal exposure. AI’s financial burden does not end with servers, power contracts, and compute.

A person’s private data is not ore. It is part of that person’s life, not feedstock for the next information economy. Human dignity does not become a line item because the machinery is expensive. Here the costs are legible enough: shareholders and other financiers carry the monetary risk; the people whose data and trust made the network possible carry the privacy harm; Zuckerberg retains the authority and remains positioned to capture the reward.

Catholic social teaching gives this argument its measure. Pius XI’s Quadragesimo Anno calls it a “grave evil” to assign to the higher power what lesser organizations can do. The sentence is not anti-technology, and it is not a demand that every data center be small. It is a warning against taking functions away from the institutions closest to the people who bear their consequences. Solidarity asks who pays. Subsidiarity asks where the decision belongs. Stewardship asks whether capital is being held in trust or merely put to work for those who control it.

Private property is legitimate, but it has a social function. So does corporate scale. Neither the market nor the state should receive immunity merely because the actors call their power freedom. Concentrated capital can build large systems efficiently; scale sometimes has genuine economies. The mistake is letting efficiency settle every moral question, as though a private government over necessary infrastructure becomes legitimate because its shareholders can price the chips.

Nor is the answer one ministry of algorithms in Washington. Those who distrust corporate concentration can become concentrated themselves; a public command over the same infrastructure would reproduce the disease in another coat. Conservative suspicion must fall on both forms of power. The proper answer is neither a sovereign cloud nor an empire with a founder’s voting fortress, but institutions that can build at scale without making every decision remote, unaccountable, and irreversible.

Meta’s financing plan does not solve the privacy problem. It solves a more immediate one: how to fund the buildout without ceding Zuckerberg’s voting control. The betrayal is one of stewardship. The company asks outside capital to bear the financial risk of its strategy, leaves the social consequences of its data practices in the wider world, and keeps the command of both. It asks capital to be communal when the losses arrive while keeping decision power private.

Disclosure is necessary, but it is not sufficient. The next disclosure must tell investors how much capital the AI buildout requires, on what timetable, and what return justifies the new equity. Better information can make an arrangement informed; it cannot make an arrangement sound. If the people providing the capital do not receive corresponding governance rights, disclosure merely teaches them more precisely what they are not permitted to change.

The opposite need not be a noble abstraction. It can be built.

Suppose the counter-model were an AI infrastructure cooperative owned by the people who build and maintain the systems and the institutions that depend upon them: workers, farm cooperatives, schools, hospitals, libraries, local businesses, and pension funds supplying patient capital. Every member would have one vote regardless of how much money it contributed. An elected board—not a founder with a permanent veto—would govern investment and pricing. Major infrastructure expansions and privacy rules would require more than a bare majority. Independent audits would publish what the cooperative knew, what it had erased, and who had benefited. Surplus would return to maintenance, service, and broader member ownership rather than flow upward as a private dividend.

Local nodes would retain authority over local uses under common standards for privacy, interoperability, and service. A pension fund could put patient capital into necessary infrastructure without buying command. A worker would have a vote in the institution that built the machine by which that worker’s labor is displaced. An institution whose patients, students, or customers supplied the data would have a voice in how that data was used. The bill and the authority would sit at the same table.

Adams County has already built one answer to that geometry: the member-owned rural electric cooperative in Friendship, which spread a necessary service across a dispersed county while keeping its governance in the hands of the people it served. An AI cooperative would ask the same discipline at a vastly greater scale. It would need patient rather than spectacular capital, working governance rather than a communications apparatus, privacy rules established before the data entered the model, and enough local knowledge to prevent the whole structure from hardening into another bureaucracy. It might be too slow. It might fail. Like every cooperative, it could degenerate into management by another name.

But that difficulty is the point. Meta can externalize the costs of its choices. A cooperative must keep the costs and the authority within the same institution. It is harder here because there is no single board that can settle the moral questions by decree, no founder’s vote that can make dissent disappear, and no distance between those who profit and those who carry the consequences. Meta’s structure makes power cheap. The counter-model makes power answerable.

Zuckerberg does not need to give up a vote to remain at the center of the AI era. He only needs a new piece of paper and an insatiable appetite for infrastructure. The question is whether the separation between those who bear the cost and those who retain the authority can remain this clean as the bill grows.

It cannot.

Meta’s new stock may buy time for the AI buildout. It cannot buy back trust; after a $25 billion privacy judgment, trust is the scarce input. The cooperative counter-model begins where trust stops being an extractive resource and becomes a responsibility attached to ownership. The tools, the vote, and the burden belong together.

That is harder than a new piece of paper. It is the construction in which the tools, the vote, and the burden belong together.