Two numbers have been sitting on my desk for a month, and they refuse to reconcile.
The first: roughly sixteen thousand. That is how many American jobs, net, are disappearing into AI every month — about twenty-five thousand erased, nine thousand created back. Not a forecast. A measurement, pulled from actual payroll records by a Goldman Sachs economist named Elsie Peng, published in a client note on April 6.
The second: six hundred thousand. That is the number of industrial jobs Japan cannot fill. Not “will struggle to fill.” Cannot. “No one’s raising their hand and signing up for it,” is how Ally Warson, a partner at the venture firm UP.Partners, put it.
Set those two facts side by side and the instinct is almost irresistible. One country is shedding workers it doesn’t need. Another is desperate for workers it can’t find. Surely this is a plumbing problem. Surely the surplus on one side flows to the scarcity on the other, and the system clears.
It does not. And the reason it does not is, I think, the most important macro story of the decade — and the one almost everyone is getting backwards.
The comfortable arithmetic
Here is the story the market has quietly told itself, and you can hear it in the way central bankers and strategists talk now.
Yes, AI will displace jobs. But the rich world is aging. Working-age populations are shrinking — by about a million a year in Europe, by nearly six million in the United States by 2032, by more than two hundred million in China by 2050. So the displacement and the shortage offset. AI takes jobs; demography needs them taken. The two forces cancel. Unemployment stays low. Nothing breaks.
It is a soothing piece of arithmetic. It is also wrong, and it is wrong for a reason an economist…