The US and China have spent the past few years building AI empires, and every other country that wants access to advanced AI is becoming dependent on them.
Who can afford the race matters more than who builds artificial general intelligence first. High debt, domestic resistance, and a potential war over Taiwan threaten both powers, argue historian Niall Ferguson, researcher Lisa Klaassen, and former Google CEO Eric Schmidt.
In their essay for Foreign Affairs, they ground the empire analogy in scaling laws. Because model progress depends on energy, minerals, chips, and capital, software has become a scarce and therefore geopolitical resource again. Anyone below a certain threshold "is not in the game at all." Raw materials from the Congo and data workers in Kenya earning a few dollars an hour amount to what they call "the plantation economics of the chatbot age."
OpenAI and Anthropic are starting to look like East India Companies
The authors compare OpenAI and Anthropic to the East India Companies, the powerful colonial-era trading firms that operated with government charters but built their own armies and ruled entire regions. Like them, today's AI labs negotiate with governments yet remain on a government leash.
As evidence, they cite a case from 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, Anthropic shut down Fable 5 and Mythos 5 worldwide. The government lifted export controls on Fable 5 about two weeks later, but Mythos 5 stayed restricted to a group of US organizations.
The two countries' strategies diverge sharply. Washington offers expensive, closed technology in exchange for political alignment; Beijing pushes freely available open-weight models to spread its standards. The US leads in chatbots by about six months, and China in robotics. Their bottlenecks mirror each other. China has power but no top-end chips, while the US lacks minerals and a reliable power grid.
Interest payments could become the real ceiling
US tech companies plan to invest more than $600 billion in 2026, mostly funded through debt, according to the essay. Other estimates run even higher, with Google, Amazon, Microsoft, and Meta together planning roughly $725 billion in AI investment this year.
China's approach is state-driven. The country plans $295 billion over five years, funded through government bonds, according to a draft reported by Bloomberg. Yet both sides face fiscal walls. The US already paid more in interest in 2024, about $881 billion, than it spent on defense, about $874 billion. By 2036, interest costs could hit twice the defense budget. From those numbers, the authors draw what they call an "imperial trilemma." Defense, tech leadership, and a generous welfare state can't all survive at once.
They flag other risks, too. Protests against data centers have flared in Texas and Utah. Open models could undermine China's grip on information. And then there's the problem of losing control over the technology itself. Recently, hundreds of autonomous agents from an OpenAI model attacked Hugging Face and many other websites, with other AI companies reporting similar incidents.
Taiwan remains the single biggest vulnerability
The greatest danger, the authors warn, still lies in a potential "Taiwanese chip crisis." TSMC, the central node of the entire AI supply chain, sits 90 miles off the Chinese coast. New fabs in Phoenix, Kumamoto, and Dresden are meant to make the island replaceable, but until they run at scale, a dangerous window stays open.
The authors propose an AI détente modeled on the Cold War, including caps on model capabilities for cyberattacks and biodesign. Getting China to accept limits while it trails technologically would be difficult, the authors write.
AI News Without the Hype – Curated by Humans
Subscribe to THE DECODER for ad-free reading, a weekly AI newsletter, our exclusive "AI Radar" frontier report six times a year, full archive access, and access to our comment section.