The most consequential economic question of the decade isn’t whether AI will create wealth. It’s who gets to keep it. And on that question, the world’s two dominant AI blocs—the United States and Asia—are both improvising, and neither is improvising particularly well.
Start with the numbers everyone cites but few can fully verify. A tiny cluster of companies sits at the center of the AI boom: chip designers, cloud providers, and a handful of model builders. Their valuations have swelled to levels that would have seemed absurd three years ago. Nvidia briefly became the most valuable company on Earth. Microsoft, Alphabet, Amazon, and Meta have collectively added trillions in market capitalization on AI enthusiasm. Whatever the exact figures, the directional story is unmistakable: the upside is pooling at the top of the stack, among firms that own compute, data, or distribution.
That concentration isn’t inherently sinister. Capital-intensive industries tend to consolidate. But it creates a political problem that democracies are only beginning to confront: if the productivity gains from AI flow mostly to shareholders and a narrow band of highly skilled workers, the broader public may end up with higher prices, disrupted labor markets, and little of the dividend.
The American approach has largely been to let the market run and hope the benefits diffuse. Washington’s instinct is to avoid slowing down the frontier labs, on the theory that leadership in AI is a geopolitical imperative. Redistribution, to the extent it’s discussed at all, arrives through the tax code or through retraining programs that critics say are underfunded and poorly targeted. There’s talk of sovereign wealth-style funds or “data dividends,” but these remain think-piece fodder rather than policy.
Asia presents a different picture, but not necessarily a better one. Japan, South Korea, and China have all published ambitious AI strategies, and each has a distinct social contract shaping how gains might be shared. Japan’s approach leans on corporate paternalism and lifetime-employment norms that AI could quietly erode. South Korea is betting on semiconductor dominance while worrying aloud about youth unemployment. China’s model channels AI gains through state-owned enterprises and industrial policy, which can direct investment but also concentrates decision-making in ways that are hard to audit.
The SCMP opinion piece that prompted this analysis makes a fair point: the debate over AI wealth sharing has become a proxy for deeper anxieties about fairness, power, and national competitiveness. But it stops short of offering a workable mechanism, because no one has one.
Part of the difficulty is measurement. Economists still argue about how to count AI’s contribution to productivity. If we can’t agree on how much wealth AI is creating, we can hardly agree on how to divide it. Another part is timing. The most disruptive effects may not arrive for years, which makes it easy for policymakers to defer hard choices. And a third is jurisdictional: AI profits are global and lightly taxed, while redistribution happens at the national level. That mismatch is a recipe for arbitrage.
There are glimmers of experimentation. Some governments are exploring public compute infrastructure, treating AI capacity the way earlier eras treated roads and electricity. Others are studying portable benefits and wage insurance to cushion workers displaced by automation. A few are floating the idea of taxing automated labor, though the practical and philosophical objections are substantial. None of these has scaled into a coherent framework.
Meanwhile, the private sector isn’t waiting. Companies are racing to embed AI into products and workflows, capturing efficiency gains that show up in earnings reports long before they show up in household incomes. That lag is the crux of the problem. The wealth is being created now; the sharing mechanisms, if they ever arrive, will come later.
A useful comparison is the early internet. The dot-com boom concentrated enormous wealth, then a crash redistributed much of it—brutally. The survivors built platforms that eventually spread economic activity widely, though the gains were uneven. AI may follow a similar arc, or it may not. The difference is that AI touches cognitive work, which is a much larger share of the economy than e-commerce ever was.
What’s clear is that neither the American faith in markets nor the Asian faith in industrial policy has produced a satisfying answer. Both models generate wealth impressively. Neither has figured out how to route it broadly without kneecapping the innovation engine. That gap between capability and distribution is where the next round of political fights will be fought—and where the tech industry’s social license will be tested.
The companies building AI would be wise to engage with that question rather than assume it will resolve itself. History suggests the public eventually notices when the gains don’t reach them, and the correction is rarely gentle.