Historian Niall Ferguson and former Google CEO Eric Schmidt argue that the United States and China have evolved into new types of empires defined not by territory, but by control over the artificial intelligence stack. The core thesis suggests that by dominating large language models, advanced chips, physical infrastructure, and capital, Beijing and Washington are setting the terms for all future scientific and intellectual progress. For developers and infrastructure builders, this means the current tech landscape is less a market competition and more a geopolitical struggle for control over the 'operating system' of global innovation.
The Return of Scarcity in Software
The authors argue that AI has shattered the previous era of software economics, where mass production occurred at near-zero marginal cost atop abundant silicon. Now, progress is strictly bound by scaling laws that require massive amounts of energy, critical minerals, and compute. This reintroduces imperial-style scarcity, where control over finite strategic resourcesβlike the rare-earth elements monopolized by China or the advanced semiconductors led by the U.S.βbecomes a matter of geopolitics rather than just supply chain management. The 'center-periphery' structure is now visible in the tech stack itself, with raw data and labor extracted from the periphery and refined in the metropoles of Silicon Valley and Shenzhen.
Chokepoints and Corporate-State Hybrids
Modern AI empires rely on dense lattices of chokepoints, much like the British Empire relied on coaling stations and the Suez Canal. Today, these chokepoints are chips, models, and cloud infrastructure. Frontier labs like OpenAI and Anthropic function similarly to the Dutch and British East India Companies, negotiating directly with heads of state for access to land, power, and sovereign capital. The U.S. model is characterized by private capital and improvised export controls, while Chinaβs approach is a deliberate act of state, mobilizing ministries and state banks to achieve global leadership by 2030. This hybrid corporate-state dynamic means that technical decisions are increasingly driven by political alignment rather than pure engineering efficiency.
Watts Versus Transistors: The Infrastructure Gap
The article highlights a critical asymmetry in infrastructure: the U.S. has the transistors but lacks the watts, while China has the energy and minerals but struggles with leading-edge chip manufacturing. The U.S. grid is aging and struggling to support the insatiable needs of new data centers, whereas China offers abundant electricity and subsidized fabrication. For builders, this creates a binary choice in deployment. The U.S. offers an expensive, closed ecosystem with strict export controls, while China offers cheap, open-weight models and hardware that can be freely downloaded and modified. This divergence forces global governments to pick a side, creating a fragmented infrastructure landscape that complicates cross-border development.
Key Takeaways
- AI control is now a geopolitical imperative, with the U.S. and China acting as 'AI superpowers' that make other nations beholden to them for access to advanced machine cognition.
- The 'East India Company' model has returned, with frontier AI labs negotiating directly with governments for resources and regulatory accommodation, blurring the line between corporate and state power.
- Infrastructure bottlenecks are now the primary constraint on AI scaling, with the U.S. facing energy grid limitations and China facing semiconductor manufacturing challenges.
- The global tech stack is bifurcating into a 'Pax Silica' led by the U.S. (closed, premium, security-focused) and a Chinese alternative (open-weight, low-cost, state-supported).
The Bottom Line
If you're building AI infrastructure today, you're not just optimizing for latency or costβyou're navigating a new imperial order where energy availability and chip access are political weapons. The era of neutral, global software distribution is over; welcome to the age of AI feudalism.