Kevin Warsh has given us three futures for artificial intelligence, and the framing is the tell. Boom, displacement, bust. Each is presented like a weather system rolling in from the Pacific, something the economy will undergo and the Fed will observe. Nobody is shown choosing anything.
But look at who pays in each scenario. In the first, growth is strong and employment holds up, and the question left unasked is who captures the gains. In the second, machines displace workers and unemployment rises. In the third, the trillion-dollar bet fails, valuations fall, and the layoffs arrive anyway, without even the consolation of productivity. In the best case, the owners of the technology win big. In the other two, working people absorb the damage. That asymmetry isn't a forecast. It is a distribution of risk, and a few companies and their investors wrote it.
The bust scenario deserves the most attention, because the country is already acting like a gambler who can't leave the table. AI-related investment has been propping up a striking share of recent growth, and the data-center buildout is straining power grids and, in some places, raising household electricity bills. If this spending doesn't pay off, we will have wagered a good part of the economy on one story. Americans who never bought a share of anything will be told the correction is the price of progress.
Then comes Warsh's line that America can't be a superpower without "super intelligence." That is not economics. It is a permission slip. Wrap a capital-spending cycle in the language of national destiny and every question about it starts to sound unpatriotic: who owns it, who profits, who gets protected when it fails, why ratepayers are subsidizing server farms. Industrial policy gets a flag draped over it, and the public gets the bill.
It is also worth remembering who is speaking. Warsh has argued that AI-driven productivity is disinflationary, a case that conveniently supports lower interest rates, which is what the president who picked him wants. Maybe he's right. But a central banker who treats the optimistic scenario as a policy premise has a stake in that scenario coming true. The Fed is supposed to be humble about forecasts, not recruited to them.
My own side shouldn't feel smug. Democrats have spent years saying "retraining" whenever technology threatens a job, as if a six-week coding course were a plan for a displaced paralegal or call-center worker. Retraining is what you say when you don't want to talk about power. The more serious question is whether workers will have any voice in how AI is deployed, and the answer today is mostly no. Only a sliver of the private workforce is unionized, and management decides what gets automated.
So here is what the Left should demand. If the public is going to underwrite this boom through cheap capital, tax incentives, permitting, and electricity, then the public should own a piece of it: equity stakes, enforceable community-benefit agreements, and rules that make data centers pay their own energy costs. Workers need a seat in decisions about automation, with advance notice and bargaining rights, not a pamphlet after the pink slip. And if displacement comes, wage insurance and a real safety net must exist before the layoffs do, not after.
Warsh's three scenarios are useful in one respect. They show that the technology isn't the main variable. Policy is. Whether AI produces shared prosperity, mass insecurity, or a hangover that falls on those who never got the upside depends on who writes the rules while everyone is still marveling at the demos. That should be settled in public, not by default.
How it may affect me
Your exposure runs through three channels: your job, if it involves routine writing, analysis, or customer service; your 401(k), which likely holds heavy AI-stock concentration through index funds; and your utility bill, where data-center demand can push up rates. Check your retirement allocation and watch state utility-commission proceedings.


