Empires of AI

Foreign Affairs | Niall Ferguson, Lisa Klaassen, Eric Schmidt

China and the United States are establishing unprecedented technological empires by monopolising the compute, physical infrastructure, advanced semiconductors, and frontier artificial intelligence models that dictate global knowledge and power projection. Access to frontier artificial intelligence now forces non-aligned nations into structural submission to these two rival metropoles. This divergence highlights two competing interpretations of imperial expansion: while Washington's authority stems from private capital markets, commercial labs, and chip design, Beijing's power relies on state-directed planning, cheap energy, and digital infrastructure embedded across the Belt and Road Initiative.

Both powers convert scarce raw inputs into centralized technological chokepoints across global supply chains. Resource limits constrain ambition. Historical precedent, such as the United Kingdom's spiralling naval expenditure following the launch of HMS *Dreadnought* in 1906, indicates that maintaining such capital-intensive technology can drain an imperial center's treasury, leaving both superpowers vulnerable to fiscal exhaustion long before either secures artificial general intelligence.

Comment

The capital intensity of frontier foundation models transforms artificial intelligence from a software sector into a heavy industrial race bounded by power generation and semiconductor fabrication. China's 2017 State Council AI Development Plan explicitly treats compute as sovereign infrastructure, leveraging state-backed energy grids to support training clusters. Conversely, American developers like Anthropic depend on private capital and foreign sovereign investment to secure high-density data centers. This structural requirement elevates gigawatt-scale electrical capacity and advanced extreme ultraviolet lithography above algorithmic innovation as the primary determinants of national power.

This expenditure trajectory mirrors the Anglo-German naval arms race following the introduction of HMS *Dreadnought* in 1906. By rendering all previous capital ships obsolete, Britain's technological leap forced the Admiralty to double naval appropriations between 1900 and 1914, creating direct friction with domestic spending commitments under the Old Age Pension Act of 1908. In similar fashion, the compute race risks absorbing sovereign capital reserves in Washington and Beijing, where gigawatt data center subsidies threaten to outstrip the fiscal capacity of both imperial metropoles.

Strategic Question for Discussion

If the capital demands of frontier AI infrastructure mirror the fiscal pressures triggered by HMS *Dreadnought* in 1906, which factor will ultimately constrain superpower reach first — domestic energy generation limits or national debt service costs?

The available evidence indicates that energy generation limits present the more immediate operational bottleneck for both Washington and Beijing. While debt markets can absorb prolonged deficit spending, expanding high-density electrical grids and nuclear capacity requires multi-year lead times that cannot be compressed by capital injection alone. Consequently, my assessment is that regional grid capacity will dictate the immediate ceiling of frontier compute expansion before fiscal exhaustion sets in.

Share your assessment in the comments below.

Read the original article