Book: AI Superpowers: China, Silicon Valley, and the New World Order Author: Kai-Fu Lee In one line: AI has moved from breakthroughs to execution - which hands power to whoever has the most data and the fiercest builders, threatens to displace millions of jobs, and leaves human compassion as the one thing machines cannot supply.
1 · Age of implementation
Deep learning was the last great discovery; the game now rewards execution - abundant data, tenacious founders, and supportive policy - not the next lab breakthrough. Applying a proven technique to floods of data beats chasing a new one. That levels the field between the US and China.
2 · Data is the new oil
Modern AI runs on huge labelled datasets, and more data usually beats a cleverer algorithm. China’s scale, mobile-first payments, and blended online-offline life make it, in Lee’s phrase, “the Saudi Arabia of data” - the fuel that powers the whole system.
3 · Compassion is the moat
AI optimises and predicts; it does not care. After a cancer diagnosis Lee reframes the future of work as compassion plus computation - pairing machine precision with the human warmth only people can give, and the one job that never runs out.
AI has crossed from an “age of discovery” - where a handful of elite researchers chased fundamental breakthroughs - into an “age of implementation,” where ordinary engineers apply proven deep-learning techniques to floods of data. Lee’s central claim is that this shift changes who wins. When the science is largely settled, raw research advantage matters less than data, capital, tuning, and relentless execution. That is precisely the terrain where China excels.
So the world now has two AI superpowers, the US and China, pulling ahead of everyone else. China’s rise rests on three engines: an ocean of data from nearly a billion mobile users, “gladiator” entrepreneurs hardened in a brutally competitive market, and a government that pours money and infrastructure into an explicit build-it culture. Silicon Valley still leads on elite research talent and moonshot ambition, but Lee argues execution advantages compound faster than research ones.
Beneath the superpower race sits the story Lee cares about most: the coming shock to jobs and inequality as AI automates routine cognitive and physical work. His own cancer diagnosis convinced him the real answer is not technical but human - a society that rewards care, creativity, and community over pure productivity.
The engine of the book is one historical shift plus four waves of deployment. First the shift, then the waves rolling from the screen into the physical world.
The discovery-to-implementation shift. The breakthrough era peaked when deep learning matured around 2012-2013. Since then progress comes from applying it, not reinventing it - which rewards data, engineering grit, and market speed. This is why Lee thinks China caught up so fast: it is an implementation contest, and China plays it hardest.
Wave 1 - Internet AI. Algorithms recommend content and ads by learning from clicks. Toutiao (now ByteDance) built a news app with no human editors, ranking stories by what each user actually reads. Every tap is labelled training data, so the product sharpens itself with use.
Wave 2 - Business AI. Companies mine decades of structured records - loans, insurance, medical, legal - to find patterns humans miss. Lee’s example: Smart Finance issues micro-loans judged by thousands of data points a bank would ignore. Here China lags the US, whose firms hold cleaner historical data.
Wave 3 - Perception AI. Sensors, cameras, and microphones digitise the physical world, blurring the line between online and offline. Think smart speakers, cashier-less stores, and face-scan payments. China’s willingness to instrument the real world (and looser privacy norms) gives it an edge here.
Wave 4 - Autonomous AI. Machines that sense, decide, and act - self-driving cars, warehouse robots, autonomous drones and tractors. This wave arrives last and slowest because the physical and safety stakes are highest, but it reshapes the physical economy most deeply.
Two superpowers, four drivers
AI dominance rests on abundant data, hungry entrepreneurs, skilled scientists, and supportive policy. China leads on three; the US still leads on elite research. Why it matters: in an implementation age, three-of-four beats one-of-four.
Gladiator entrepreneurs
China’s copy-then-innovate arena - the O2O (online-to-offline) wars, price battles, outright cloning - forged founders who are fast, thick-skinned, and ruthless. Cloning was the entry fee; survival demanded real innovation on top. It produced operators, not just inventors.
Copycat to innovator
Early Chinese startups copied Western products, but Lee argues they then leapfrogged. WeChat fused messaging, payments, and services into a single super-app with no Western equal - proof the “copycat” label expired years ago.
Data is the new oil
Deep learning is the enabler, but it is hungry: performance tracks the volume and richness of labelled data. China’s mobile-payment ubiquity generates real-world data (where, when, what you buy) that Western tech, built on cards and browsers, simply does not see.
Toutiao and the feedback loop
An AI news feed with no editors beat established portals by optimising purely for engagement. The lesson: build products where usage generates the data that improves the next version - a loop that compounds into a moat rivals cannot copy.
The great decoupling
The real threat is not robot overlords but jobs and inequality. AI automates routine cognitive and physical tasks, potentially displacing a large share of work within 15 years and concentrating wealth among those who own the systems.
Compassion plus computation
Machines optimise; they do not love. Lee’s answer is to pair AI analysis with human care - the oncologist who delivers a diagnosis, the nurse, the teacher - and to redesign work so the human half is valued, not eliminated.
Lee opens with China’s “Sputnik moment” (AlphaGo beating Ke Jie) and the discovery-to-implementation thesis, then tells how China’s copycat era hardened its entrepreneurs. The middle lays out the four AI waves and why data and policy tilt the race. The final third turns dark then hopeful: the job-displacement warning, his cancer diagnosis and the change of heart it triggered, and his blueprint for human-AI coexistence built around a care economy and a social investment stipend.
Audit your work for routine. List the parts of your job that are repetitive analysis or prediction - those are most exposed. Start shifting your time toward the creative and human parts now.
Place your role on the 2x2. Ask how creative and how social your work is. Move toward the creative-plus-compassionate corner, which stays safe longest.
Get close to data. Whatever you build, design it so real usage generates data that makes the next version better - Toutiao’s loop is the model.
Pair AI with a human. For high-stakes calls, use AI for the analysis and a person for the care and accountability, rather than swapping one for the other.
Read opportunities by wave. Ask which of the four waves an idea rides; perception and autonomous AI move slower but reshape the physical economy hardest.
Take the inequality warning seriously. Whether you build, invest, or govern, plan for displacement - reskilling, safety nets, and care-economy roles - not just efficiency gains.
Lead with compassion. Lee’s closing lesson: let AI handle optimisation and spend your freed time on the connection and care that machines cannot fake.
A few things to hold lightly. The book’s timelines (self-driving ubiquity, displacement percentages) are confident forecasts, not certainties, and the field has moved fast since 2018. Lee is unusually optimistic about China’s model and light on its downsides - surveillance, censorship, and the human cost of the data he praises. Critics note the US-versus-China framing can flatten a messier, more global picture, and that the hopeful care-economy blueprint is sketched more than costed. Read it as a sharp thesis and a moral appeal, not a settled prediction.
“China is the Saudi Arabia of data.”
We have left the age of discovery and entered the age of implementation.
The real risk of AI is not killer robots but the quiet displacement of jobs and the inequality that follows.
AI is a tool for analysis and optimisation; it is up to humans to supply the love.
The question is not what AI can do, but what we choose to do with the time and attention it frees.