AivexaNewsSearch
AI news for builders and product teamsChecked every hour

Ex-Google CEO Eric Schmidt calls OpenAI and Anthropic modern colonial trading companies

Collected Oct 25, 2026

Former Google CEO Eric Schmidt, historian Niall Ferguson and researcher Lisa Klaassen published an essay in Foreign Affairs arguing that the United States and China have effectively divided the global AI market between them, leaving every other country that wants advanced AI dependent on one of the two. The piece compares OpenAI and Anthropic to the East India Companies — chartered colonial trading firms that negotiated with governments while building their own armies and governing territory. The authors' central claim is that affordability, rather than who reaches artificial general intelligence first, will determine the outcome.

The argument rests on scaling laws. Because model quality has scaled with energy, minerals, chips and capital, software has become a scarce geopolitical resource again rather than a freely distributed one. The authors say any country below a certain threshold is not in the game at all. They describe raw materials sourced from the Congo and data workers in Kenya paid a few dollars an hour as the plantation economics of the chatbot age.

The East India Company comparison is grounded in a June episode the essay cites. The US Commerce Department ordered Anthropic to cut off foreign nationals from its two most capable models. Because the company could not verify users' nationalities in real time, it shut down Fable 5 and Mythos 5 worldwide. Export controls on Fable 5 were lifted roughly two weeks later, but Mythos 5 remained restricted to a set of US organizations. In the authors' framing, the labs bargain with governments while remaining on a government leash — much as the chartered companies did.

The two powers pursue mirror-image strategies. Washington trades expensive closed systems for political alignment; Beijing distributes open-weight models freely to spread its own standards. The essay puts the US about six months ahead in chatbots and China ahead in robotics. Their constraints also mirror each other: China has power generation but lacks top-end chips, while the US lacks minerals and a reliable grid.

Money is where the authors see the real ceiling. US technology companies plan to spend more than $600 billion in 2026, mostly financed through debt, according to the essay; other estimates cited run higher, with Google, Amazon, Microsoft and Meta together planning roughly $725 billion this year. China's effort is state-directed: $295 billion over five years funded through government bonds, per a draft reported by Bloomberg.

Both sides face fiscal limits. The US paid about $881 billion in interest in 2024, more than its roughly $874 billion defense budget, and interest costs could reach twice the defense budget by 2036. From this the authors derive an imperial trilemma: defense, technology leadership and a generous welfare state cannot all be sustained simultaneously. They also flag domestic friction, including protests against data centers in Texas and Utah, the risk that open models erode China's information control, and loss of control over the technology itself — citing an incident in which hundreds of autonomous agents from an OpenAI model attacked Hugging Face and many other websites, with other AI companies reporting similar events.

Taiwan is presented as the largest single vulnerability. TSMC, the central node of the AI supply chain, sits about 90 miles off the Chinese coast. New fabs in Phoenix, Kumamoto and Dresden are intended to make the island replaceable, but the authors say a dangerous window remains open until those facilities run at scale. As a remedy they propose a Cold War-style AI détente with caps on model capabilities for cyberattacks and biodesign, while conceding that persuading China to accept limits while it trails technologically would be difficult.

My reading is that the essay's most concrete contribution is reframing AI competition as a balance-sheet problem. If frontier training and inference are financed largely with debt, then interest rates and lender appetite become constraints on capability roadmaps in a way that model benchmarks do not capture. The Anthropic episode illustrates a related dynamic: compliance obligations that depend on real-time user verification can force a global shutdown when verification is not feasible, which turns export policy into an availability risk for every customer outside the US. The comparison to chartered trading companies is rhetorical, but the underlying observation — private labs with public obligations and geopolitical weight — is a fair description of how frontier developers now operate.

A likely trade-off follows from the essay's own numbers. Closed, expensive systems buy alignment and control but cede distribution; open weights buy adoption and standardization but weaken the originator's leverage. That tension is visible in the divergent US and Chinese approaches.

Why it matters: Developers building on frontier models should treat export controls, nationality verification and sudden global withdrawals as operational risks, not edge cases, and should plan fallbacks accordingly — an inference from the cited Anthropic episode rather than a stated recommendation. The financing argument also suggests that compute pricing and model availability may track credit markets as much as research progress. For product teams outside the US and China, the essay's framing implies continued dependence on two suppliers whose access terms are set by governments, which argues for multi-vendor architectures and, where feasible, open-weight alternatives.

Read at The Decoder

Based on reporting from the original publisher. Visit the source for full context and later updates.

Publisher excerpt

In Foreign Affairs, historian Niall Ferguson, researcher Lisa Klaassen, and former Google CEO Eric Schmidt cast the US and China as AI empires and every other country as a dependency. The race won't be decided by breakthroughs but by who can bankroll it. US corporations alone plan to spend more than $600 billion in 2026. Yet mounting debt, protests against data centers, and a potential chip crisis over Taiwan could stall both superpowers. The article Ex-Google CEO Eric Schmidt calls OpenAI and Anthropic modern colo