The valuations of artificial intelligence (AI) companies have turned strange. OpenAI was worth roughly $850 billion this past March; Anthropic about $965 billion in May. Neither has ever posted a profitable year. For numbers like these to make sense, there has to be a market large enough to justify them — and only one market is that big: the global labor market. Behind the polite language of “productivity tools” sits a business model that, in the end, replaces human work and the wage bill that comes with it. That makes the direction of this technology less a story about a new product than a question about people’s livelihoods.
This is where a trap appears. When a company swaps employees for AI, costs fall and the share price climbs. In February, Jack Dorsey’s fintech firm Block cut its workforce from more than 10,000 to under 6,000 — about 40 percent — saying AI had “changed what it means to build and run a company,” and the stock jumped more than 22 percent. Block was not alone; several large tech firms trimmed headcount around the same time under the banner of “AI efficiency.” But the people laid off are also someone else’s customers. If every firm makes the same move, the pool of people who can buy anything shrinks. Each company grows leaner while demand across the whole economy erodes. A paper released this year in the Wharton School’s research series, “The AI Layoff Trap,” argues that because the spending lost to any one firm’s layoffs is spread thinly across the entire market, cutting jobs is rational for each firm yet ruinous for all of them together. Some see this leading to what they call a “dead economy” — one where factories hum and money circulates, but ordinary people are left outside it.
None of this is settled, though. Re-run the same model with different assumptions, and a follow-up analysis finds it produces stability rather than collapse. Daron Acemoglu, who won the 2024 Nobel Prize in economics, estimates that only about 5 percent of tasks are currently cost-effective to automate with AI, and that its productivity boost over the next decade will land under 1 percent. History cuts both ways as well: machines have destroyed jobs and created new ones at the same time. Farming, which once employed most workers, has shrunk to around 2 percent without the economy falling apart, and many of today’s occupations did not exist a generation ago. The catch is that those adjustments took decades. The camp warning of mass unemployment and the camp calling AI overhyped are evenly matched.
Korea sits in a different position. This is not a country with workers to spare; with the world’s lowest birth rate, it is a country running short of them. Seen that way, AI may be closer to filling the gap left by a vanishing workforce than to stealing jobs. Even so, one question remains: who captures the wealth AI creates? If the gains flow to the few who own the machines and the capital, the distribution problem stands on its own — whether jobs grow or shrink.
Whether AI pushes people out or shores up a thinning workforce, and who ends up sharing in the gains — that is the heart of the argument now underway. Whichever way the answer falls, the results will surface first in our workplaces, and before long.