When AI makes the economy more productive, does more of that extra income flow to machine owners or to workers?

This explores who captures the gains when AI makes the economy more productive: does more of the income go to the owners of machines and software, or to workers? Only one note in the collection answers this directly. Anthropic's scenario modeling finds that as AI-driven growth speeds up, the labor share falls and the capital share rises Does AI growth inevitably shift wealth away from workers?. The detail that surprises people is that average wages can still go up while knowledge workers' wages stall or drop. A growing economy and a shrinking slice for workers can both be true at once. In that model, the split is decided by two things: who owns the AI, and how hard it is for workers to move into new occupations. Faster technology alone doesn't decide it.

Firm-level evidence suggests one reason the gains may pile up with capital. Firms with more AI exposure replace freelance workers with AI tools faster and more cheaply than less-exposed firms Do firms substitute labor for AI at different rates?. That points to returns to scale: firms that already have AI capability get better at substituting for labor, which favors large owners of AI infrastructure. The opposite pressure comes from how AI exposure is spread across a job. When AI touches only a few of a job's tasks, workers can shift to the tasks it doesn't cover, and the net effect on employment stays small Does concentrated AI exposure enable workers to adapt and reallocate?. A survey of executives also finds labor moving into new roles rather than shrinking overall Do AI productivity gains feel larger than they actually measure?. So whether labor's share erodes depends partly on how AI cuts across a job, task by task.

There's also a measurement problem: nobody yet knows how large the gains are. Survey-based estimates put generative AI's current aggregate productivity gain at about 1.1% What is generative AI's actual aggregate productivity gain today?. Brynjolfsson reads 2025's combination of slower job growth and GDP growth as a 2.7% productivity surge Is AI productivity finally showing up in economic data?. Notice that his signal, more output with fewer new jobs, is the pattern you'd expect if capital were taking a larger share. His earlier J-curve work explains why the numbers are confusing in the meantime. Firms spend heavily on training, process redesign and software that statistics don't count well, so productivity looks weak at first and then looks inflated later Why does AI productivity appear weak when investment is high?. Those uncounted investments are a form of capital, so the measured split between capital and labor shifts depending on where we are on the curve.

The last piece is that a worker's personal gain doesn't automatically become a raise. McKinsey finds that 80% of respondents report personal productivity gains from AI, but only 37% see their company's profits improve Why do individual AI gains not translate to business results?. Many workers also hide their AI use because of workplace stigma Why do workers hide productivity gains from AI use?. A gain nobody can see can't be bargained over in a pay negotiation. Workers who report the biggest AI speedups are among the most worried about losing their jobs Does AI productivity gain always ease job displacement fears?, which suggests they already sense the gains may not come back to them. The collection is thin on direct estimates of capital and labor shares beyond the Anthropic scenarios. Its stronger contribution is showing the mechanisms that steer the split: who owns the AI, how exposure spreads across tasks, and whether gains are visible enough to bargain over.