The State, AI and money: who will capture the windfall?

In April, we set two philosophies against each other: Musk's universal income and Piketty's capital endowment (our analysis). It was a debate of ideas. In June, it has become a negotiation. In Washington, the Trump administration is concretely discussing taking a share of OpenAI's capital; billionaires are proposing to tax tokens; and the same State that has just switched off Claude Fable 5 now wants to share its fruits. The question "who captures the value created by AI" has left philosophy and entered politics. And Europe, meanwhile, looks on.

1. Why now: the value of AI goes public

This debate is not exploding by chance this week. In a few days, the value of AI has moved from private to public: SpaceX carried out the biggest stock market debut in history (75 billion raised, making Elon Musk the first man worth 1 trillion dollars), while Anthropic filed its listing paperwork on 1 June and OpenAI on 8 June. Three of the most strategic assets on the planet are about to be listed.

Yet at the precise moment a form of wealth becomes public — tradable, measurable, visible — the question of how it is shared becomes political. As long as AI was a laboratory promise, "who benefits?" remained theoretical. As soon as it is counted in listed billions, the question becomes: why should those billions go to a few shareholders rather than to the community that supplied the data, the energy and the market? The IPO wave is not the backdrop of the debate: it is its trigger.

2. OpenAI's offer: a sovereign fund financed by AI

The idea comes from Sam Altman, who pitched it to the administration as early as 2025 before returning to it this week: a "Public Wealth Fund", a fund that would pay every American — including those who hold no shares — a share of the growth driven by AI. The mechanism is unprecedented: OpenAI would give the federal government between 1% and 5% of its capital, rather than sell it. No cost to the taxpayer: the company cedes shares, the State collects the dividends, the fund redistributes.

In its April 2026 public policy paper, OpenAI describes a fund that would invest in "diversified long-term assets", capturing the growth not only of the labs but of the entire economy adopting AI. The whole thing follows in the wake of Trump's February 2026 order calling for the creation of a national sovereign fund. A notable fact: the idea appeals on both left and right, which is rare in the United States.

3. The State as shareholder: redistribution or disguised nationalisation?

Trump has said it publicly: the State could hold pieces of OpenAI, and the same logic is being raised for xAI — even other giants. This is where enthusiasm must give way to lucidity. A State that becomes a shareholder of a lab that it also regulates creates a structural conflict of interest: the same actor sets the rules, controls the models, and collects the dividends. The boundary between public and private does not blur: it dissolves.

One can see social progress in it — the people finally get their share. One can also see the opposite: a State captured by the companies it is supposed to oversee. It all depends on who governs, and with whom. This is where the subject shifts from the financial to the political.

4. The other school: taxing tokens

Against the "State takes shares" model, another family of proposals is rising: taxing usage. Mark Cuban has costed a federal tax of less than $0.50 per million tokens, levied on AI providers: he estimates it at 10 billion dollars in the first year, with a potential of 30 to 100 times that over a decade. The advantage, in his view: it would push providers to optimise their tokenisation and energy consumption, while financing the external costs of AI.

In California, gubernatorial candidate Tom Steyer is proposing his own "token tax": a fraction of a cent per unit of data processed, to feed a state sovereign fund — infrastructure, clean energy, housing, and direct payments to Californians displaced by AI. The philosophical difference with OpenAI's fund is clear: the tax falls where AI is consumed, not where it is produced. It does not ask the labs for permission.

Sources: CNBC, TechCrunch, Gizmodo, BeInCrypto, Windfall Trust — May-June 2026.

5. Universal dividend (Musk) versus capital endowment (Piketty)

In practice, these mechanisms settle the debate of ideas we described in April (Musk vs Piketty). And the accident of the calendar makes it glaring: the man who has just become the first trillionaire in History is also the loudest spokesman for redistribution.

Musk's thesis: the abundance dividend. For Musk, AI, robotics and energy will produce such abundance that paid work will become optional. He is not talking about a basic income (UBI) but about "universal high income" — a high income for all, distributed as a dividend on the productivity of machines. He goes as far as advising people to stop saving for retirement. It is seductive; it is also the ideal argument for whoever owns the machines: if abundance falls from the sky, nobody asks who holds the tap.

Piketty's thesis: endowment, not alms. At the opposite pole, Piketty rejects the allowance paid from above: in Capital and Ideology (2019), he defends a universal capital endowment of €120,000 handed to every individual at 25 — roughly 60% of the average French household's wealth — financed by a progressive tax on wealth and inheritance. The idea: that everyone truly holds a share of the economy (a home, a project, training), and not merely a revocable cheque.

Piketty formulated this endowment before the AI wave, and does not make it a "data" argument. But — and this is an Otium reading here, not a quotation from the author — his logic extends naturally: if models are trained on our data, capture our attention and automate our work, then the community looks less like a beneficiary to be assisted than a co-owner who was never paid. Seen this way, a share returned to the citizen is no longer charity: it is the repayment of a debt.

The nuance is not semantic, it is about power. A dividend makes the citizen dependent on a payment that a government — or a billionaire — can reduce. A capital endowment gives them an asset they control. The American proposals of June (public fund, dividend) all lean towards the first model. That is not a detail: it is easier to offer a revocable income than to cede lasting power.

6. The conflict that tells the truth: Trump against Amodei

To understand what is at stake, one has to look at the fracture running through the sector. On one side, OpenAI has lobbied for a light federal framework and readily aligns with the administration. On the other, Anthropic and its boss Dario Amodei have become the most outspoken opponents of the deregulation Trump wants. Amodei published an essay calling for a federal agency able to block his own model launches — an "FAA for AI", with the legal power to prevent the release of a frontier model that failed safety tests.

This line goes much further than Trump's 2 June order, which merely encouraged labs to voluntarily share their models with government auditors. And the conflict has turned into a confrontation: on 12 June, the administration ordered Anthropic to cut foreign access to its latest models — which led to the worldwide suspension of Fable 5 and Mythos 5. The next day, David Sacks, an investor who has passed through the administration, publicly accused Anthropic of having been "reckless". Anthropic is now fighting on two fronts: against OpenAI, its rival; and against the White House.

This conflict is not a detail: it reveals two irreconcilable visions. One wants to own and deregulate (the State as shareholder, the free market); the other wants to supervise and brake (the safety agency, the red lines). The redistribution of the windfall will be decided in that space.

7. The real risk: techno-oligarchy

Here is the unspoken issue that hangs over everything else. Some of the tech bosses backed Trump — and today find themselves at the heart of the State apparatus. Elon Musk led the "Department of Government Efficiency" until May; Peter Thiel has seeded his associates throughout the administration, up to the vice-presidency; David Sacks was the AI and crypto "czar". When those same actors develop frontier models and advise the president on how to regulate them, the question is no longer only economic.

The scenario some fear: that "redistribution" serves first those who hold the models, the capital and the ear of power all at once. A sovereign fund fed by AI is a good idea if the State remains the referee; it is a bad one if it becomes the business partner of the labs it regulates. We are not claiming that the tipping point has occurred; we are saying that democratic safeguards are, here, the real subject — well before the size of the cheque.

8. And Europe? The planning trap

While America invents mechanisms — brutal, contestable, but real — to capture and redistribute the value of AI, Europe debates its sovereignty. The response rising on the left, carried notably by Raphaël Glucksmann, wants to make digital sovereignty "the great anti-capitalist battle" and names its adversaries — Musk, Altman, ByteDance — whose algorithms "enslave our minds". The proposed remedy: European planning, a large joint loan, the abolition of the veto between States.

The diagnosis is not wrong: Europe cannot content itself with consuming an AI it does not master, and the lag is documented — around 10% on most use cases, for want of mature skills and because of stricter regulation. But the remedy betrays a blind spot worth naming bluntly: you do not plan an innovation, you execute it. A five-year plan has never produced a frontier lab; it produces reports. What made Mistral was not a decree: it was a team that delivered fast. Planning doctrine confuses political intent (sovereignty) with its real engine (execution, venture capital, speed of delivery). America is arguing about how to share a windfall it has already created; Europe risks planning the sharing of a windfall it has not yet produced.

9. What it changes for you

For a French executive, this briefing is not an American spectacle. It carries three concrete lessons:

The value of AI is going to be taxed. Token tax, sovereign funds: whether you develop or consume AI, the real cost will tomorrow include a public dimension. Measuring your consumption by use today means anticipating that line.

Dependence becomes a sovereign risk. After Fable's suspension, the possibility of a State shareholder in your provider adds a variable: your tool depends on foreign political trade-offs. A fallback model is no longer a technical option, it is insurance.

Execution remains your only advantage. While the giants and the States fight over the rent, immediate value is created where AI is put to work: in your processes, your products, your teams. That ground cannot be planned; it is delivered.

10. FAQ

What is OpenAI's Public Wealth Fund?
A fund that would pay citizens a share of AI profits, seeded by a donation of 1 to 5% of OpenAI's capital to the federal government. The Trump administration is discussing it, and is considering similar stakes in other labs, starting with xAI.

What is a token tax?
A levy on the use of models, per million tokens. Mark Cuban proposes $0.50/million at federal level (an estimated $10bn/year); Tom Steyer a Californian version for a sovereign fund. The tax falls on consumption, not production.

Why is Europe behind?
About 10% behind on use cases, owing to a lack of mature skills and stricter regulation. The response through planning is debated: you create the conditions for innovation, you do not decree it.

AI has ceased to be a question of models and become a question of power: who owns it, who governs it, who collects its fruits. America cuts to the quick, at the risk of confusing the State with its champions. Europe debates method. And the real dividing line may not be between left and right, nor between Europe and the United States: it is between those who produce the value and those who merely organise its sharing.

Sources: CNBC, TechCrunch, Les Échos, SiliconANGLE, The Inquirer, Gizmodo — May-June 2026.