Eric Schmidt says AI race is becoming a modern empire battle
An essay by Niall Ferguson, Lisa Klaassen and Eric Schmidt argues that AI power now depends on chips, energy and capital as much as model quality.
The global AI race is becoming less about who reaches artificial general intelligence first and more about who can keep paying for the machinery behind it. A new essay by historian Niall Ferguson, researcher Lisa Klaassen and former Google chief executive Eric Schmidt argues that the United States and China are building rival AI empires, while the rest of the world is becoming dependent on both.1
The argument is deliberately bigger than a model leaderboard. The authors say progress now depends on an enormous mix of electricity, minerals, advanced chips and investment capital. Countries below a certain threshold, they write, are effectively not in the game. That makes AI infrastructure a geopolitical resource rather than just another software platform.
Two competing models
According to the essay, Washington is exporting expensive, closed systems in exchange for political alignment, while Beijing is using open-weight models to spread Chinese technical standards. The approaches are different, but both rely on supply chains that are difficult to reproduce elsewhere. China has power and industrial capacity but lacks access to the most advanced chips; the US has leading chip access but faces constraints around minerals, electricity and grid expansion.1
The Decoder compares OpenAI and Anthropic to the East India Companies: private organisations operating with government support and influence, but still tied to state power.2 That comparison is provocative, yet it captures a real tension. AI labs increasingly need public infrastructure, export rules and political protection, while governments are relying on companies to deliver capabilities they cannot build alone.
The bill comes due
The financial case may be more important than the imperial metaphor. The essay points to hundreds of billions of dollars in planned US technology investment, much of it funded through debt, alongside China’s state-backed spending plans.1 Those commitments assume that demand, energy supply and investor confidence will keep expanding together. They also make the AI race vulnerable to higher interest costs, public opposition to data centres and a downturn that makes infrastructure spending harder to defend.
The most serious risk is still Taiwan. TSMC remains a central supplier of the advanced chips used throughout the AI industry, and new factories in the United States, Japan and Europe will take time to reach scale.1 A crisis around the island would therefore affect far more than one company or one national market; it could interrupt the physical layer on which both AI strategies depend.
Our opinion
The useful part of this argument is its refusal to treat AI as weightless software. Every chatbot promise eventually meets a transformer, a power contract, a chip shipment and a balance sheet. The East India Company comparison is eye-catching, but the practical warning is simpler: countries that depend on somebody else’s energy, chips and capital do not fully control their AI future.
A Cold War-style détente over model capabilities would be difficult, particularly while Washington and Beijing still see technological leadership as strategic leverage. But the essay is right that competition without supply-chain resilience is brittle. The next decisive AI breakthrough may not be a smarter model. It may be the country that can build enough power, chips and public legitimacy to keep its models running.