On the morning of May 22, the federal government decided two things about American technology. It would not require a single AI model to be reviewed before release. And it would take an equity stake in nine companies building quantum computers, writing IBM a check for a billion dollars in exchange for a piece of the company.

These look like opposite instincts. One is the state stepping back — declining to regulate the most consequential technology of the decade even at the level of a voluntary checkup. The other is the state stepping in — not lending, not granting, but buying, taking ownership in the firms building the machine that comes after the machine. Hands off here. Hands deep in the pocket there.

They are not opposite instincts. They are the same industrial policy, holding two different tools. And on May 22 you could watch it pick up one and put down the other in the span of a single news cycle.

The order that fell apart in a morning

The AI executive order Trump was set to sign that Thursday was already a husk. An earlier draft had imagined the government reviewing frontier models before release; by the time it reached his desk it no longer required any mandatory testing — only a "voluntary" framework the labs had largely been running already. Even that was too much. Hours before the ceremony, after last-minute calls from Elon Musk and Mark Zuckerberg and a final plea from AI czar David Sacks, the President pulled it. The fuller anatomy of that morning — why the same labs that had volunteered for the review killed the order that would have written it down — is its own reversal. What matters here is the residue: the order failed not because it asked too much — it asked almost nothing — but because, by mid-2026, asking anything at all was the wrong direction.

What the same government did with its other hand

While the signing ceremony was being torn down, the Commerce Department was announcing something that, in any other decade, would have been unthinkable. It would award $2 billion in grants to nine quantum computing companies — and take equity stakes in exchange. IBM was set to receive a billion dollars of that package. The market understood exactly what it was looking at.

closed up the day of the equity announcement
a stake from the same package
on a $1B federal check for a piece of the company

Read those two sentences together. The state that would not require an AI lab to show its work before shipping a model just became a part-owner of the companies trying to build the successor to the transistor. It declined to act as a regulator and volunteered to act as a shareholder, on the same day, about the same broad subject — the technologies that will define the next twenty years of power.

It would not tell the companies what to build. It would simply own the building.

The equity playbook was never about quantum

This was not improvisation. The quantum stakes are the latest entry in a playbook the administration has been writing for nine months, in public, one deal at a time.

It began with Intel. In August 2025, the administration was reported to be in talks to take a stake in the struggling chipmaker; days later it converted CHIPS Act grants into equity and the government owned 10% of Intel. The structure was telling. The deal was designed to deter Intel from selling its foundry — the state taking ownership specifically to lock in a strategic capability the market might otherwise have let go. Lutnick insisted Washington wouldn't sit on the board or touch governance. It just wanted to own.

Then the playbook compounded. In December 2025, the administration agreed to inject up to $150 million into xLight — a startup making lasers for EUV machines, with former Intel CEO Pat Gelsinger on its board — for an equity stake. And the quantum deal announced on May 22 had been visible since October 2025, when IonQ, Rigetti, D-Wave and others were reported to be in talks to give Commerce equity in exchange for funding. A Wall Street Journal reporter called it, dryly, a signal of "Washington's wider involvement in key parts of the economy."

By April 2026, the Intel bet looked like genius: the stake had quadrupled to roughly $36 billion. The government had not saved Intel by writing a rule. It had saved it — and made a fortune doing so — by buying it. The lesson the administration drew from Intel was not "this is a dangerous precedent," the warning some had raised at the time. The lesson was: this works, and the returns are extraordinary. Quantum was the next application of a method that had already paid off.

Two tools, one logic

The reason these two May 22 events belong in the same frame is that they answer the same question — what is the government's role in frontier technology? — and arrive at an answer that only sounds contradictory.

Regulation and equity are not opposite postures toward industry. They are opposite tools, and industry has a strong preference between them. A safety review constrains a company: it slows the release, exposes the model, imposes a cost the company would rather not pay. An equity stake subsidizes a company: it injects capital, signals federal backing, and — as Intel's shareholders learned — sends the stock up 30% in an afternoon. One tool makes the CEOs call the President to kill it. The other makes them stand next to him and smile.

So the same actors who killed the AI order would, in a different room, welcome a check. Musk and Zuckerberg lobbied against the rule. Quantum executives lobbied for the stake. The administration's industrial policy is not "hands off" and it is not "hands on." It is hands-off where the lever constrains, hands-in where the lever subsidizes. The state has discovered that it can shape an industry far more comfortably by owning a slice of it than by writing a paragraph about it — because ownership is something the industry will line up to receive, and rules are something it will spend a morning destroying.

There is an older version of this logic, and it did not come from a tech administration. State capitalism — the government as shareholder in strategic firms — has always preferred ownership to regulation for exactly this reason. A regulator imposes costs and earns resentment. A shareholder shares the upside and earns alignment. The administration that spent a year branding itself as the enemy of "AI red tape" has quietly rebuilt the relationship between Washington and Silicon Valley along the one axis that generates no resistance: capital, flowing in.

The shape of the new posture

The substitution is easiest to see at the level of specific companies. The federal government will not require OpenAI or Anthropic to submit a model for review. It will not impose a single binding constraint on the most powerful AI systems being trained right now. But it will write IBM a billion-dollar check for a piece of its quantum business, and it will book the gains when the stock moves.

This is not deregulation. Deregulation is the state withdrawing from the field. The state has not withdrawn — it is more entangled with American technology than it has been since the days of defense-funded semiconductors. It has simply changed which instrument it reaches for. It abandoned the regulator's clipboard and picked up the shareholder's certificate. The order it refused to sign would have given it a say in what gets built. The equity it bought gives it a stake in what gets sold.

On May 22, the government declined to review a single AI model before release, and bought a piece of nine companies building the machine that comes next. It will not tell them what to build. It will simply own the building — and clip the dividend.