Howard Lutnick is handing Nvidia, Amazon, Microsoft, OpenAI, and Anthropic the American state. The Commerce Department’s own program page states what a designation buys: expedited approvals for chip export licences, government support for a company’s work overseas, and access to federal financing. The program was announced about a year ago, and no company had been named. People familiar with the plans say more than two dozen names arrive within the week — among them Nvidia, Amazon, Microsoft, OpenAI and Anthropic — announced by the Commerce Secretary himself.

It is true that the strategic problem the Commerce Department says it is solving exists. To be fair, Chinese buyers have been reaching advanced American silicon through third countries; Congress spent the past week seizing on the Nvidia scandal to press for tighter export curbs; and Huawei’s march through 5G procurement across Europe and Latin America remains the standing demonstration of what happens when a rival’s stack becomes a country’s infrastructure. A country that exports advanced compute has a legitimate interest in where the compute goes. The trouble is that nothing in the “national champions” program addresses that interest. Expedited licences do not close a diversion route. Federal financing does not catch a smuggler. The list does not name the firms with the best record of keeping American technology out of Chinese hands; it names the firms best positioned to sell.

Read the program the way you would read any machine: by what it moves. The department’s own website lists three benefits, and each is a state power re-cut for private use. Expedited approvals for chip-export licences: advanced chips cannot lawfully leave the country without a Commerce licence, and “expedited” means the queue the rules impose on everyone else simply does not apply to you — expedited approval is what the border looks like when you know the border guard. Government support for work overseas: the trade missions and the embassies become the sales force for your product line. Access to federal financing: public money, on terms the public has not been shown. The fourth benefit goes unlisted, and it is the largest: designation itself — a published roster of the firms the state will work for, set against every firm it will not. A favourites list, published by the office that issues the licences, is not an incentive program. It is a moat, poured at public expense, in the office that controls the drawbridge.

An export licence is not a grant and not a subsidy. It is a permission slip — the legal authority to move a controlled item to a destination the control regime has already judged risky. Chips are controlled for a documented reason, and the reason keeps showing up in the record: this publication has traced the route that Nvidia silicon takes out through Southeast Asian data centers and into China. The licence requirement is not an incidental hoop. It is the whole apparatus of the control regime. What the designation proposes is that the wait grow shorter for the firms with the largest commercial interest in the wait being short. A rule tells everyone what happens to them; a list tells more than two dozen firms what happens to them and tells everybody else what happens to everybody else.

Set that against the rest of the month. The same department is moving fresh rules restricting chip diversion to China. The Treasury secretary has already faulted Commerce over Nvidia chip exports to China. The White House has been answering questions about Commerce’s Nvidia oversight. The agency being told, in three separate news cycles, that it has failed to police the pipeline is preparing to hand out fast passes on the pipeline.

It is worth being precise about what designation is not. It is not a finding that these companies are good at what they do — Nvidia’s chips power the most advanced models and are, by any measure, among the world’s most valuable commodities, and no document from Commerce is needed to establish that. It is not a security clearance either. Nothing on the program page says the vetting is harder than the licensing; the page says the licensing gets faster. Read literally, that is a merchandising plan for a federal agency.

Look at who is on the list, and read it as a map of the stack. Nvidia is the accelerator layer — the silicon the frontier models are trained on. Microsoft and Amazon are the cloud layer those models are trained on and served from; OpenAI and Anthropic rent that cloud to do both. More than two dozen firms will be designated in all, but Nvidia, Microsoft, and Amazon rank among the most valuable companies on earth, and OpenAI and Anthropic are the two most consequential private labs in the field. These are not five competitors in one market. They are successive layers of one market, held by a small set of firms, and the government is about to certify the stack as national policy. The designation does not pick winners; it certifies the firms that already won, and it performs the one act they could never manage alone: it makes their concentration official. Every startup outside the roster now competes against incumbents with fast-track licences, federal money, and a government sales force. Barriers to entry are things incumbents dig at their own expense over years. This one is being poured for them, at public expense, in an afternoon.

The cui-bono trace runs where it always runs. The designated firms receive expedited export approvals, government support for their overseas sales effort, and access to federal financing — three concrete benefits, drawn from the department’s own description. The costs land elsewhere: on the exporters left standing outside the list, on foreign buyers who now understand that the menu has been curated in Washington, and on the public purse, which is being asked to finance the winners of a contest the same public’s regulator is supposed to referee. The 2020 House Antitrust Subcommittee report spent 449 pages on what happens when the largest players sit at the referee’s table. The table is the policy.

Cory Doctorow’s account of how firms behave once the constraints come off is the useful instrument here, and he would want the credit: the four forces that historically held this kind of conduct in check are competition, regulation, self-help, and labour. The designation dismantles the first two by decree. Regulation is not being removed — it is being converted from a constraint into a menu, with the terms published in advance for the firms that can most afford to wait out everyone else. That is a more durable arrangement than deregulation, because it survives every future administration. A rule you can repeal; a list of favourites you inherit. The first force moves too: state backing tilts competition against the exporters left off the list. Self-help — the ad blocker, the rival client, the whole apparatus of competitive compatibility — has no answer to a licence queue. The fourth force, labour, has spent the past three years at these very firms learning what a layoff does to leverage.

Then there is the enforcement problem, which is quieter and worse. Lina Khan’s “Amazon’s Antitrust Paradox,” the 2017 essay that reoriented a generation of enforcers, argued that platform power has to be read structurally rather than off a price chart — that a firm can hollow out competition while discounting nothing. Whatever one makes of the record that followed — Judge Mehta’s 2024 liability ruling finding Google an illegal search monopolist, the Justice Department’s and FTC’s monopolization complaints against Google and Amazon, the FTC’s losses on Meta-Within and Microsoft-Activision — nobody litigating those cases was trying to make the defendants national champions. The next act supplies what the litigation withheld. You cannot credibly prosecute a monopolist you have crowned. The crown is, structurally, a promise not to prosecute: every future government lawyer staring across a courtroom at Nvidia or OpenAI now works for a state that has certified the defendant as the nation’s champion. Enforcement does not have to be repealed. It only has to be embarrassed.

The financing channel deserves its own sentence because it is the one benefit that reaches into the public purse directly. It is fair to say the buildout is real — Microsoft’s twenty-year deal to restart Three Mile Island Unit 1 and Meta’s twenty-year deal for the entire output of Constellation’s Clinton plant are the most legible evidence that data centres have become real-economy infrastructure with real-economy costs. Read the financing line next to the dockets it touches: one firm on the roster is already the defendant in a live Federal Trade Commission monopolization complaint, and two of the labs on it are in court over the copyrighted work their models were trained on. John Kenneth Galbraith called the interval between a fraud’s commission and its discovery the bezzle — the stretch when the embezzler has the money and the victim feels no loss — and it always inflates in a boom. Federal financing for designated champions is the state volunteering to keep the bezzle inflated: public money offered, at the peak of the boom, to firms with the deepest private capital pools in history standing behind them. This is not a public bank lending on published terms to any qualified applicant. It is a roster the department assembles in private.

“National champion” is an old word, and Canada holds the oldest file on it. The Canadian Pacific was the original: incorporated in 1881 with public money, land grants, and a charter that bound the prairies to its freight rates — the closest thing the country has had to a state church with rolling stock. What the railway produced over the following seventy years was not gratitude. It was the wheat pools, the Crow’s Nest freight-rate fight, an entire democratic inheritance grown in the resentment of a firm the nation was made to serve. To be fair to the old playbook, it at least charged for the crown: the railway had to build the line, and the grain-rate concessions of the 1897 Crow’s Nest Agreement were extracted from it, not gifted to it. The 2026 version inverts the exchange. The champions receive the licence, the financing, and the sales force. The obligations column is blank. My grandfather took that railway west in 1956, to a job in Selkirk, and he carried out of that economy a working man’s suspicion of anyone who described a corporation as the country’s.

Harold Innis built his staples economics on exactly this pattern: a political economy organizing itself around the movement of one dominant commodity, and what the organization costs the people who live along the route. The commodity has changed names since — fur, timber, wheat, freight — and the organizing has not. The champions roster is the same political economy with compute in the staple’s place. Tim Wu’s longer history of the information industries is a history of open networks closing into empires, usually with the assistance of the government that claimed to be supervising them, and the program’s stated purpose — packaging chips, models, and services so foreign buyers can purchase them together, with adoption steered against Chinese rivals — is that older move in its current form: chokepoint diplomacy. Doctorow’s chokepoint capitalism names the private version of the arrangement, an intermediary parked between producers and everyone else collecting rent on every transit; this is the same structure with a flag on it. The buying country accepts dependency at every layer — silicon, cloud, model — from sellers the same government has just promised never to discipline. Governments that spent a decade learning from Huawei’s 5G expansion what stack dependence costs them in leverage are being invited to sign the identical dependency, in American.

The timing is the tell, and it is not irony; it is the logic on display. The designation arrives in the same weeks the Treasury secretary faulted Commerce over Nvidia chips reaching China, after a year in which chips moved through a growing Southeast Asia data-center route under oversight that critics describe as a failure to crack down. The failure on the table is loose licensing discretion. The remedy on offer is faster licensing discretion for the favourites. The licence queue is the checkpoint — it is where diversion control physically lives. You do not answer a checkpoint failure by telling the heaviest traffic it no longer stops. And the program quietly converts its own stated rationale into the thing antitrust has spent three years trying to unwind: if foreign adoption is the goal, then in every negotiating room on earth “American AI” and “five firms” become the same phrase.

The corrective is dull, which is a point in its favour. Licence by published rule, identical for every applicant; abolish name-by-name designation and make chip export licence decisions under stated rules, recorded and reviewable; no federal financing for firms carrying unresolved monopolization complaints until the courts rule; a firewall between the office that licenses exports and the one that promotes them, because a regulator that sells cannot regulate. None of that requires anyone to admit anything. It requires only that a licensing power behave like a licensing power instead of a patronage office with a .gov address. And if Washington wants the world on American standards, it should fund what won the world the first time — open protocols and interoperability. The internet was a government project whose victory came from being nobody’s product.

The program’s website lists what the champions receive. It does not list what the country gets back, because the country is not a party to the bargain; it is the sales channel. The announcement is expected within the week, and the people familiar with the plans say the date could still slip — the only variable left in a program whose recipients were chosen before the program had a public face. The department has not announced a policy. It has announced a guest list. A champion is supposed to be the one who fights for you; read the roster as written and the service runs the other way. A national champion, it turns out, is a firm the nation works for — and read it as what it is, a work order, and the rest follows: the nation is the contractor, the champions are the crew, and everyone not on the list is the job.