The Prophecy That Forgot to Check the Weather

Kai-Fu Lee wrote this book in the afterglow of AlphaGo's victory over Ke Jie, and you can feel the heat of that moment on every page. He was right about the broad strokes: AI would reshape labor markets, China would not be a bystander, and the contest between Beijing and Washington would become the defining axis of technological power in the 2020s. These were not trivial calls in 2018, when much of the American commentariat still treated Chinese tech as a copycat industry running on stolen IP and cheap labor. Lee saw clearly that China's advantages — vast data reserves, a relentless entrepreneurial culture, and a government willing to function as both investor and customer — would produce a distinct AI ecosystem, not a subordinate one. The "Book City" chapter on local government guidance funds and city brain projects reads now less like futurism and more like a field report filed just before the deluge. Alibaba's City Brain did scale. The guidance funds did multiply. The ecosystem he described has, in many respects, arrived.

What Lee could not see — or chose not to examine — was the degree to which geopolitics would throttle the very competition he was narrating. The U.S. chip export controls that began in earnest in 2022 and tightened through 2025 fundamentally altered the hardware substrate of Chinese AI development. Lee's framework assumed a world where data and algorithms were the primary battlegrounds and compute was a commodity. That assumption aged poorly. The semiconductor chokepoint — TSMC, ASML, NVIDIA — turned out to be the fulcrum, not the footnote. His book also preceded the large language model revolution that would rewrite the rules of what "AI" even meant to most people. Lee's AI was narrow, task-specific, optimized: face recognition, loan approvals, autonomous delivery. GPT-3 arrived two years after publication; by 2024, foundation models had made his taxonomy of AI waves feel like a map drawn before the discovery of a new continent. He was charting the rivers while the ocean was forming offshore.

There is a deeper blind spot, and it is ideological. Lee's treatment of China's surveillance infrastructure is notably gentle. He acknowledges the data advantages that flow from a population with fewer privacy protections, but he frames this almost entirely as a competitive asset rather than a political problem. In 2026, after years of reporting on the uses of AI-enabled surveillance in Xinjiang, after the social credit system's uneven but real expansion, and after the chilling effects on speech documented by researchers inside and outside China, this framing feels not just dated but complicit in a certain kind of techno-optimism that mistakes scale for virtue. The book's emotional climax — Lee's cancer diagnosis and his subsequent call for a more humane AI future — is genuine and moving, but it operates at the level of individual ethics. It does not confront the structural question of what happens when a state with no independent judiciary wields the tools he celebrates.

The book occupies an interesting seat in the corpus. It inherited the anxiety of Martin Ford's *Rise of the Robots* and the geopolitical framing that would later be sharpened by Chris Miller's *Chip War*. It gave the English-speaking world a credible insider account of Chinese tech ambition at a moment when that account was badly needed. Its successors have been more precise — more attuned to the hardware layer, to alignment risk, to the strange emergent capabilities of large models — but Lee's book remains the document that forced a particular conversation into the open. It said: this is a two-player game, and one player has advantages you have not bothered to understand. That was useful. It remains useful, even as the game has turned out to be less about data volume and more about compute access, talent migration, and the willingness of governments to weaponize supply chains.

If Lee were to revise this book today — and it needs revising — the question it would have to answer is one it never thought to ask: what happens to the AI superpower thesis when the superpowers discover they would rather deny each other capabilities than develop their own?