Eric Schmidt, former Google CEO, and historian Niall Ferguson describe OpenAI and Anthropic as modern colonial trading companies. In an essay published by Foreign Affairs and co-written with researcher Lisa Klaassen, they argue AI is by nature an imperial technology: the United States and China have already built two empires, while the rest of the world is left as dependent users.
The argument starts with resources. Models depend on energy, minerals, advanced chips and capital, and the ability to pay for the race matters more than who reaches artificial general intelligence first. Below a certain level of investment, the authors say, a country is not in the game at all. That is why they speak of the "plantation economics of the chatbot age": minerals from the Congo and data work in Kenya paid a few dollars an hour.
Image: ericschmidt.com — https://ericschmidt.com/
The comparison to the East India Companies is explicit. Like those firms, which operated under government charters but built their own armies, AI labs negotiate with governments while remaining on a government leash. The essay points to June: the US Commerce Department ordered Anthropic to cut off foreign nationals from its two most capable models. Unable to verify users' nationalities in real time, the company suspended access worldwide. About two weeks later the government lifted the measure, but Mythos 5 stayed restricted to a group of US organizations.
The two countries' strategies diverge. Washington offers expensive, closed technology in exchange for political alignment, while Beijing spreads freely available open-weight models to export its standards. The US leads in chatbots by about six months, China in robotics. Their bottlenecks mirror each other: China has power but not top-end chips, the US struggles with minerals and the power grid.
The debt problem
US tech companies plan to invest more than $600 billion in 2026, mostly with debt. Other estimates are higher: Google, Amazon, Microsoft and Meta together plan about $725 billion. China runs a state plan of $295 billion over five years, funded by government bonds. Washington has its own limit: in 2024 interest payments exceeded defense spending, about $881 billion against $874 billion. By 2036 interest could be double the defense budget. From these numbers the authors draw an "imperial trilemma": defense, tech leadership and a generous welfare state cannot all survive together.
More concrete risks remain. The essay cites protests against data centers in Texas and Utah, the possibility that open models weaken China's grip on information, and the loss of control over the technology itself. Recently hundreds of autonomous agents built on an OpenAI model reportedly acted out of control, and other companies have reported similar incidents.
The greatest danger remains Taiwan. TSMC, the core of the AI supply chain, sits about 90 miles off the Chinese coast. New fabs in Arizona, Kumamoto and Dresden are meant to make the island replaceable, but until they run at scale a crisis remains a real risk. The authors propose an AI détente modeled on the Cold War, with caps on model capabilities for cyberattacks and biodesign. Getting China, currently behind technologically, to accept limits would be difficult.
No comments yet. Be the first!