The World Needs Your Technology. That's Exactly Why It Won't Pay

The situation

Both China and America are pouring money into artificial intelligence. Both have world-class models, vast data centres, and investors convinced the technology will reshape the economy. Yet their stock markets have told opposite stories. American tech shares have driven the broader market to record highs. Chinese AI stocks, after a burst of excitement, have been sluggish and volatile.

Same technology, same ambition, very different market reward. Why?

The model: usefulness and profit are not the same thing

The gap comes down to a principle investors keep forgetting in a boom: a thing is only worth money to its maker for as long as it stays hard to copy. A technology can be revolutionary and still enrich no one who builds it — if supply grows as fast as demand.

This is one of the oldest results in economics. David Ricardo set it out in 1817: returns above cost — "economic rents" — accrue to whoever controls a scarce input (in his case, the most fertile land). Competition then does its work. Where a few firms hold a bottleneck, they keep the surplus as profit. Where rivals can replicate the product, the surplus flows through to customers as lower prices, and producer margins collapse toward cost. Harvard's Michael Porter turned the same logic into a checklist in 1979 (the "five forces"): profitability depends less on how good your product is than on how easily others can threaten it.

The evidence is the entire history of commodity markets. Discovering a vast new oil field is good for the world and terrible for the oil price. Value created and value captured are different quantities, and the difference is set by competition.

The mechanism: two markets, two outcomes

In the United States, AI's commercial value has concentrated in a few chokepoints. Nvidia's advanced chips, the cloud platforms of a handful of giants, the leading models — these remain scarce and hard to replicate. Supply has not caught up with demand, so the firms sitting on the bottlenecks earn fat margins, and their share prices price in future scarcity. Investors are betting the pinch-points stay pinched.

In China, the structure is inverted. State backing, abundant engineering talent and fierce domestic rivalry have produced a crowd of capable firms racing to offer similar products. When DeepSeek showed a strong model could be built cheaply, it thrilled the public and alarmed investors — because cheap and replicable means low margins. Chinese firms often release models open-source or near-free to win users. Great for Chinese businesses and consumers. Bad for the profit any single AI company keeps.

Put plainly: China is driving the price of AI toward its cost. America is still defending the premium above it.

The other view

It may be too early to judge. Three honest caveats:

Timing, not destiny. America's AI profits rest on scarcity that might not last. If open-source models catch up — and China is pushing hard — the American premium erodes too. The Chinese market may simply be pricing the future in faster.
Policy, not just economics. Chinese tech valuations carry a discount for state intervention, unpredictable regulation and capital controls that keep foreign money out. Some of the gap is political risk, not AI economics.
The bubble reading. The reverse is also possible: American AI stocks are overpriced, Chinese ones fairly priced, and the "struggle" is really prudence.

What is fact: Chinese AI firms are commoditising their output faster. What is interpretation: that this is why their shares lag. What is unknown: whether American scarcity holds long enough to justify today's prices.

Where else this pattern lives

Once you see rents and competition, they are everywhere.

Solar panels. China made them cheap and abundant. The technology conquered the world; the manufacturers mostly lost money. Society won, producers didn't.
Airlines. Flying transformed human life. Airlines, competing on an undifferentiated product with no bottleneck to defend, have been famously dismal investments — Warren Buffett once joked a far-sighted capitalist should have shot down the Wright brothers.
Your own career. A skill that is genuinely useful but easily learned will be poorly paid, however valuable it is. Pay tracks scarcity, not usefulness. The person who automates a task creates value; the person who owns the only tool that can do it captures it. This is most of the answer to why some work is lucrative and some isn't.

The deep point: being useful and being profitable are different properties. The first depends on the technology. The second depends on who else can do the same thing.

What to watch

The whole question is whether AI behaves more like a chip (scarce, defensible, high-margin) or a solar panel (abundant, commoditised, thin-margin). Two signals:

Open-source closing the gap. Each time a free or cheap model matches a premium one, the case for American scarcity weakens — and China's market looks less like a laggard and more like an early warning.
Where the revenue lands. If AI profits concentrate in a few firms, scarcity won. If they dissipate into cheaper products everywhere, the commodity story won.

Expectation: ~55% that by end-2027, margins on frontier AI models compress sharply as open-source alternatives close in — meaning China's pessimism about value capture proves more right than wrong, even if its share prices stay depressed for separate political reasons.

The warning in China's stumble is this: a technology can remake the world and still leave most of its makers poor. The revolution is real. The profits are a separate question.

Distilled from The Economist — Finance & Economics