The question is how much power can be concentrated in the hands of a few people before the state decides that this concentration has become a danger in itself.
Dario. Demis. Elon. Mark. Sam. Five names that no longer need a surname to identify themselves. Five of the most powerful figures in American technology, who in a few years have gone from leading technology companies to a much more unusual position: that of people who are directly influencing the way the economy, politics and society of the future may look. At a time when Artificial Intelligence is developing at a speed that its creators themselves admit may be difficult to control, the question is not who will win the AI race.
The question is how much power can be concentrated in the hands of a few people before the state decides that this concentration has become a danger in itself.
American history has a clear precedent. More than a century earlier, another group of entrepreneurs was leading a technological revolution that was changing America at an extraordinary pace. Railroads were connecting the country, steel was building cities, and oil was transforming the economy. At the forefront of this transformation were John D. Rockefeller, Andrew Carnegie, and Cornelius Vanderbilt. They became enormously wealthy by building companies that dominated the new industrial economy. But with their wealth came something else: power.
And so the parallel with the AI era today comes. Rockefeller was not just a successful businessman. Standard Oil became so dominant in the oil market that its control began to be seen by the government as a threat to competition. At the height of his economic power, Rockefeller’s wealth, measured in relation to the size of the American economy, reached an incredible size, almost 1.5% of it.
JP Morgan went even further in another direction. In the financial crisis of 1907, he personally intervened to organize the rescue of the American financial system. In practice, a private banker was performing a function that today we would associate with the central bank and the state. This is what makes the comparison with today’s Silicon Valley interesting. Because even AI bosses don’t just control very large companies. They control or direct technology that could become the infrastructure of a large part of the future economy. Elon Musk has xAI and a technological ecosystem that extends far beyond AI. Sam Altman runs OpenAI. Dario Amodei Anthropic. Mark Zuckerberg controls Meta and is investing aggressively in Artificial Intelligence. Demis Hassabis leads Google’s DeepMind efforts.
These companies are racing to build the most powerful AI models. But the paradox is that even as they race against each other, some of them are beginning to agree on something very unusual: the race may be moving too fast.
Amodei has called for the industry to slow the pace at which it improves its most advanced models. Altman has publicly supported him. Musk has, too. Such a merger is unusual among rivals fighting for billions of dollars in investment, top talent and technological edge. But this rare consensus suggests something more important than a technical debate over the safety of the models. It suggests that the people closest to the technology are themselves starting to worry about its pace. If a system becomes more capable faster than human institutions can understand, test and control its behavior, the question is no longer simply which product will win the market. The problem becomes political.
Who controls the technology? Who sets its boundaries? Who decides what risks are acceptable? And above all, who controls the people who control the AI? This is the question often missing from the public debate on Artificial Intelligence. We talk a lot about what AI can do. We talk about productivity, automation, superintelligence, jobs, and competition with China. But we talk less about an equally fundamental issue: the concentration of capacity to build this technology. The most advanced models require colossal capital, data centers, chips, energy, talent, and infrastructure that cannot be easily built by an ordinary company. This makes the AI race, in a way, also a race for the concentration of capital.
And that’s exactly what America has experienced before. During what is known as the “Gilded Age,” the concentration of wealth and control over new industries created what American history would later call the era of the “robber barons.”
The term was negative not because these people did not bring development. Quite the contrary. Rockefeller made the oil industry more efficient. Carnegie built a steel industry that helped physically transform America. Vanderbilt’s railroads changed the way people and goods moved. The problem was that the benefits of technological progress came with an extraordinary concentration of economic power. And when economic power becomes large enough, it begins to translate into political power. This is also the concern that is emerging today about AI. If a few companies control the most powerful models, they can gain not only economic advantage, but also influence over technological standards, information, security, public policy, and the very way society interacts with technology.
This is not to say that Altman, Musk, Amodei, Zuckerberg or Hassabis are the new Rockefellers. The differences are vast.
Rockefeller controlled a significant portion of a physical industry. AI is a much more distributed technology, with universities, companies, open-source resources, and governments all having a say in its development. But in one respect the comparison is even more troubling. Oil, steel, and railroads were strategic industries. AI could be the technology that impacts all industries at once. This doesn’t just change the scale of the problem. It amplifies it!
A monopoly on oil gives you power over one sector of the economy. A decisive advantage in AI can give you an advantage over countless sectors: finance, defense, healthcare, communications, manufacturing, scientific research, and information. So the question “is AI more powerful than oil?” is less interesting than the other question: Could control over AI become a new form of control over the economy?
That’s where government comes in. The Trump administration has so far been inclined to favor competition among private companies, arguing that America can’t afford to slow itself down while China tries to win the technology race. The dilemma is real…
If the United States imposes too many restrictions, there is a risk that American companies will lose ground to China. If it imposes almost no restrictions, there is another risk that technology will develop faster than the state’s capacity to regulate it. Thus comes the great paradox. To win the race with China, America may need to give its AI companies more freedom. But this very freedom could create the concentration of power that will later force the state to intervene. Rockefeller’s story shows that this is not a theoretical fear. In 1911, the Supreme Court ordered the breakup of Standard Oil into 34 companies, after deeming its structure to be in violation of antitrust law. Two years later, Congress created the Federal Reserve, institutionalizing a function that in the crisis of 1907 had been performed to an extraordinary degree by J.P. Morgan.
In both cases, the state was trying to restore a balance that the concentration of private power had disrupted. This could be the path of AI as well. Not necessarily by breaking up OpenAI, Anthropic, or other companies. Not necessarily by banning powerful models.
But with rules that make it impossible for a handful of companies to single-handedly set the boundaries of the technology that society as a whole will use. So Amodei’s proposal for independent evaluators, stronger testing, and coordination between companies is not just a technical issue. It touches on the age-old problem of American capitalism: how to let innovation run its course without allowing innovation to create a concentration of power that then threatens the system itself. This is the dilemma that America experienced during the Industrial Revolution. And it may be experiencing it again. At the beginning of the 20th century, the American state concluded that the free market was not enough to guarantee competition. It intervened, broke up monopolies, and built new regulatory institutions.
Today, at the dawn of another technological revolution, history is not repeating itself exactly. But the structure of the problem is strikingly similar. A transformative technology. A handful of extremely wealthy entrepreneurs. Brutal competition for dominance. A state that needs them to maintain strategic advantage. And, above all, an increasingly difficult question to avoid: What happens when the people who build the most powerful technology of the time become so powerful themselves that the state begins to need to control them?
Rockefeller learned the answer. Standard Oil was too big to be left unchecked. Perhaps the question of the 21st century will be whether AI can become too powerful to be left alone in the hands of those who build it.
And if that happens, the next battle won’t just be between OpenAI, Anthropic, Google, Meta, and xAI. It will be between them and the state. Just like a century ago, the battle won’t be about whether technology should exist. It will be about who has the right to decide how much power it can have.