Economy · September 29, 2026

The Permit Is the Product

This is not a new play.

A century and a half ago, the railroads were the commanding technology of American life. They also fought one another to exhaustion. Prices collapsed in one season and soared in the next. Secret discounts went to the biggest customers. Small lines were swallowed. The public saw a monopoly. The railroad men saw a fight they could not police among themselves.

So they went to Washington.

They were not simply dragged into federal regulation. They helped ask for a national referee. State rules looked angry. A federal commission looked like order they could live with. Congress created the Interstate Commerce Commission. Later laws made it stronger. The commission did not only protect farmers from the railroads. It helped the railroads protect themselves from new competitors.

That was the wall: not only steel and land, but a government blessing. Once you need official permission to enter a business, the people who already have it start to look like the only serious players. The map froze. When trucks and highways arrived, they did not grow up inside the licensed system. They grew up around it.

Nobody needed a secret plot. The company that has already built the network prefers rules that make the next network expensive.

Now translate the ledger.

A 1900 dollar does not look like a 2026 dollar. The number on the old page is smaller. The buying power is not. The old railroad books are converted into today's money so a reader can compare one fortune with another without doing the arithmetic in his head.

The roads did not need a fairy tale. They needed the price war to stop. Regulation did not invent the railroad fortune. It protected the one already built. For two generations the government did what the private pools could not: it helped hold the price. New competitors had to ask permission to exist. The companies that already owned the map collected the rent.

By 1900 the railroads reported about $11.5 billion in gross capitalization. A 1900 dollar is worth roughly forty of today's dollars. That capital stock is on the order of $400 billion to $460 billion in 2026 money — not one company, the paper value of the industry.

In 1900 the roads took in about $1.5 billion in operating revenue. In today's dollars that is roughly $60 billion a year, collected in a smaller country, on a network no rival was free to duplicate. Just before the First World War, net operating revenue ran near $800 million a year — about $25 billion to $30 billion now. After the fixed claims, net income was still on the order of $10 billion to $12 billion a year in 2026 dollars, industry-wide.

The families are easier to see. Cornelius Vanderbilt died in 1877 with about $100 million. His son William Henry roughly doubled it, to more than $200 million by 1885 — about $7 billion to $8 billion today. Measured as a share of the American economy then, it looks larger still. In 1918, Forbes put W. K. Vanderbilt at $100 million and the Harriman railroad estate at $80 million: about $2.4 billion and $1.9 billion now.

Two cautions. After Washington gained the power to cap rates, official returns often sat in the mid-single digits. Trucks arrived. The commission that had stopped a price war became a cage. Deregulation in 1980 was the admission that the cage had outlived the feast. A price conversion makes the comparison readable. It does not make the old men small. A fortune that large, in an economy that small, was command of the main track. The public was told it was buying fairness. The established roads were buying sleep.

Follow the incentives.

Today's leading AI companies are the railroads of a new kind of track. Compute, data, and talent are one defense. A government license is a better one. They can afford inspectors, delays, and a permanent staff in Washington. A small new company cannot. A free system released on the internet does not come with a legal department. A locked system sold by the month does.

That is not theory. The largest closed labs have gone to Washington asking for tests, licenses, and a single national standard they are already staffed to survive. Nearly two hundred smaller firms have warned that fencing off free systems would not make the country safer. It would leave them paying rent to the same few landlords.

That is why the public language is safety and the practical demand is a gate.

The chiefs of the biggest closed systems have begun to agree: test the most powerful systems before they go out, create one official body to set the standard, restrict what is judged too dangerous, and let Washington lead rather than fifty states. They argue about the name on the door. They agree there should be a door.

The strongest safety argument is not foolish. A tool that can write computer attacks, help design weapons, or run itself without a person in the loop is not a pocket calculator. If the danger is real, looking before you leap is adult behavior.

The strongest competitive argument is not foolish either. The same firms that can pass the test will help write the test. 'Regulate only the most powerful systems' sounds careful. In practice it can mean that only the already-rich may legally be powerful. Everyone else rents intelligence from the few who already passed.

Both things can be true at once. A real risk can still be an excellent excuse.

Who wants what is not mysterious.

The established company wants a rule it can obey and a rival who cannot. It wants one federal rule so it does not have to win fifty statehouses. It wants the China threat close enough to justify locks on the door, and far enough that it can still sell into nearby markets through side doors.

Watch the behavior anyway. Some American firms have sold access through overseas offices tied to Chinese companies the Pentagon already treats as a problem. They ask Washington to restrict the free systems that would let a startup compete without paying rent, while they remain the landlord. That is not treason. It is not patriotism. It is interest.

Beijing is useful three ways at once. It justifies a license a small American firm cannot carry. It raises the cost of the open marketplace. And it supplies a villain so the public does not notice the most important general-purpose tool of the century collecting in a handful of domestic boardrooms. If the rule is clumsy, the advantage flips. A United States that turns its best systems on and off by political mood teaches the world a lesson Beijing has been selling for years: American supply can disappear on Friday. Chinese supply will still be there on Monday.

Examine the strongest arguments.

One side says only the government can prevent disaster. If they are right about the worst case, delay has a moral claim.

The other side says the builders are asking the government to charge admission to a business they already occupy. If they are right about the cost of compliance, that is the point of the rule.

The uncomfortable question is not which tribe is correct.

It is this: if the law is written to govern only the most powerful systems, who will be allowed to build the next one?

The men asking for the law already built this one. They will help define what 'powerful' means. They will afford the audit. The newcomer will be told that safety requires patience, capital, and a relationship with the commission. That is how a living industry becomes a utility. Utilities can be well run. They are not where the next railroad is born.

Why it should concern an ordinary citizen.

A country that puts hiring, medicine, credit, schooling, and the news behind a government license is not merely regulating software. It is deciding who may participate in ordinary life.

Concentrated private power is an old American problem. Joined to a federal licensing power, it is a different one. The first can be competed against. The second can make competition illegal — or legal only for people who already won.

If a handful of firms write a railroad rule for thinking machines, they will not need to win every price fight in the open market. They will need the fight to end. Converted into today's dollars, the old fortune is large enough that no serious person should pretend the petitioners are asking only for public virtue.

The Founders did not anticipate thinking machines. They anticipated men who acquire power and then discover reasons why the next man should wait.

People seek advantage. People protect advantage once they have it. Industries that cannot suppress rivals in the market will try to suppress them in the statute book. Foreign threats make the statute book easier to pass. The public is invited to debate the end of the world so it does not inspect the application form.