Who should own AI? Anthropic's fund and Musk's SpaceX shares revive the idea of 'capitalism for everyone'

WHAT HAPPENED. On October 18, Anthropic said it will commit 5% of its shares to a new Public Benefit Fund. Its stated goal is to cushion job losses caused by AI. The pledge needs shareholder approval. The board may also add 2 percentage points for each 50% rise in the company's valuation above a baseline it has not disclosed. According to CNBC, the fund will be a limited liability company wholly owned by Anthropic, not a tax-exempt charity. It also cannot be used for lobbying or political donations. A day later, Elon Musk replied on X: “I recommend giving the shares directly to the public.” He then promised that in 2027 he will give one of his SpaceX shares to every child in America. “This will be the first of many,” he wrote.

Our thesis is this. The underlying question is a good one, and it comes at the right time. If AI concentrates wealth in a handful of companies, the answer shouldn't be income alone, meaning a regular check. It should also be ownership: a slice of a growing pie. But as they stand today, both announcements are more gesture than system. Whether they help will depend on the fine print: who controls the money, when it can be sold and what it is really worth.

THE NUMBER THAT MATTERS. Let's run the numbers. In its last private round, in May, Anthropic was valued at $965 billion. At that price, 5% is worth about $48 billion. Financial press reports point to an IPO valuation of around $2 trillion, and the company has now publicly filed its offering document (the S-1). At that valuation, the fund would be worth roughly $100 billion. That is huge for a foundation and small for a country. Split among America's roughly 170 million workers, it would come to between $280 and $590 per person, paid once. Musk's offer is smaller still. The US has about 73 million minors. At SpaceX's October 16 close of $146.03, the gift would cost about $10.6 billion. Each child would get about $146, less than a sixth of the $1,000 the Treasury already deposits in the new “Trump Accounts” for children. A Bloomberg Opinion calculation cited by Al Jazeera says Musk would part with about 1.5% of the shares he owns outright. His voting power, around 82%, would fall by only about a tenth of a point. There is also a tax angle. Will McBride, chief economist at the Tax Foundation, estimates that giving the shares away instead of selling them could avoid about $2.6 billion in capital gains tax (the tax on the profit from selling an asset). One caveat matters here. Musk is not required to sell, so that money is not revenue the Treasury was certain to collect.

WHAT HISTORY TEACHES. Sharing out capital is not a new idea, and earlier experiments offer lessons. Since 1982, Alaska has paid every resident an annual dividend from its Permanent Fund, which is built on oil revenue. In 2026 it is $1,200: a $1,000 base plus a $200 energy relief payment. It works because it is universal, regular and rule-based. Its weakness is that lawmakers set the amount each year. Norway took a different route. Its sovereign wealth fund, an investment fund owned by the state, was worth about $2.39 trillion in June 2026, roughly $430,000 per resident. No Norwegian gets a check. Instead, a fiscal rule lets the budget draw about 3% of the fund each year. It is collective ownership, highly diversified and professionally run. The third precedent is a warning. In the 1980s, Thatcher's “popular capitalism” sold state companies to small savers, with campaigns such as British Gas's “Tell Sid” in 1986. The share of British adults who owned stock rose from 7% to more than 20% by the end of the decade. Then it fell back: many people sold quickly, and ownership concentrated again. Spain had its own version in the 1990s. Telefónica's 1997 share offering alone drew more than 1.2 million applications from individuals, according to SEPI, the Spanish state holding company. But thousands of scattered small shareholders had little say in how the companies were run. There are other models too. Connecticut's “baby bonds” set aside $3,200 for each baby born on Medicaid, the public health insurance for low-income families, and the state treasurer estimates the accounts could grow to between $11,000 and $24,000. And in 2021, OpenAI CEO Sam Altman proposed an “American Equity Fund.” Large companies would pay 2.5% of their value each year in shares, and landowners another 2.5% in cash. By his estimate, within a decade every adult would receive about $13,500 a year. The lesson is clear. The models that work are mandatory, recurring, diversified and rule-based. Today's voluntary pledges meet none of those four tests.

INCOME OR OWNERSHIP. The core debate is how to protect people if AI displaces work. One option is universal basic income: a regular, tax-funded payment to everyone. It protects people now, but it depends on each year's budget, and it gives nobody a share of the gains. The other is universal basic capital: giving every person a stake in the assets that generate wealth. Its big advantage is that if AI sends profits soaring, citizens benefit too. Its risk is the reverse: if the stock falls, citizens lose. SpaceX began trading in June at $135 and, according to Al Jazeera, is down 35% from its post-listing peak. Giving a child a single company's stock is a concentrated bet, exactly what any financial adviser warns against. Our view is that both are needed. Income is the tool for the short term, because job losses are real and they arrive before the gains. Ownership is the tool for the long term, because it is how the abundance AI may create gets shared.

THE FINE PRINT. Four issues separate image from substance. First, liquidity, meaning the ability to sell. Treasury rules published on September 30 bar families from refusing donated shares. They also generally cannot sell them for five years, or until the child turns 17. The Anthropic fund's shares, meanwhile, are not yet publicly traded, so their value exists only on paper. Second, dilution. If new shares are issued for the fund, every other shareholder's percentage shrinks, which is why the plan needs their approval. Third, governance. If the fund is a subsidiary of Anthropic itself, who decides how it spends its money, and how independently? The company says it will share more “in the coming months.” Fourth, political incentives. Ann Lipton, a law professor at the University of Colorado Boulder, argues in Al Jazeera that millions of shareholders who cannot sell will have an interest in SpaceX facing less regulation. The Treasury itself defended such donations on similar grounds: they could make recipients feel they have a stake in the company's fate. Al Jazeera also notes the timing: the pledge came two weeks before the November 3 midterm elections. It adds that Musk has given more than $276 million to Republican groups since July. One last detail shows how intertwined the industry is. According to Yahoo Finance's summary of Anthropic's filing, the company has agreed to spend up to $120.5 billion on computing capacity from SpaceXAI through 2029. The two rivals now competing in generosity are also customer and supplier.

OUR TAKE AND WHAT IT MEANS. We see the glass as half full, with caveats. The good news is that the question of who owns AI has entered public debate, and the people building the technology raised it. Anthropic had already committed $200 million to studying and easing AI's effects on jobs. Counting pledges from co-founders and employees, the company says about 20% of its equity will end up dedicated to public benefit. That is not trivial. But a fund the company controls and one share per child are no substitute for public policy. In the short term, someone who loses a job in administration, customer service or office coordination will need training, income and time, not $146 locked up for five years. In the long term, our bet stands. If AI helps eradicate diseases and makes almost everything cheaper, the wealth will exist. The question will be who owns it. We propose five tests for the next proposals: a diversified basket rather than a single company, rules rather than a donor's goodwill, independent governance, recurring contributions and reasonable liquidity. The tests will come soon. One is Anthropic's IPO, expected in November according to press reports. Another is OpenAI's next funding round; its nonprofit foundation has kept about 26% of the group since its 2025 restructuring. Europe, with its tradition of public funds and popular privatizations, should join this debate rather than watch from the sidelines. The goal is worth pursuing: letting AI's abundance free people to work on what they love. For that, capital for everyone has to be a system, not a tweet.