Confronted with a trillion dollars a year flooding into artificial intelligence, a former Biden economic official has a fix: let Washington decide where the money goes instead. The boom is not what threatens the economy. That instinct is.
Jennifer Harris lays out the case in the New York Times this week. The AI buildout, she warns, is "hollowing out" the economy, bidding away the capital, land, and talent that would otherwise go to housing, factories, and start-ups. She has the facts right and the culprit exactly wrong.
What she calls a drain is the most ordinary thing a market does: capital moving toward whoever expects to earn the most with it. That is the market working.
The problem isn't crowding out. It's who does the crowding
Start with the word she leans on, because it has a precise meaning. Crowding out happens when one big claimant takes so much of the available capital that it bids up the cost and squeezes others out, and a trillion-dollar investment wave can certainly do that. Harris is right that AI is competing for finite savings, chips, land, and crews. So concede it plainly: private capital crowds out other private capital all the time. That is just bidding, and bidding is how a market decides what gets built.
The dangerous kind is the other one: government claiming the resources for political ends. And Harris's own source gives it away. The economist she quotes to indict AI, RSM's Joseph Brusuelas, warns of "the combination of government borrowing and A.I. investment" squeezing out other sectors, naming the government in the same breath. The difference is everything: AI has to outbid the rest of us and pay what the market asks, while government borrows and takes without bidding at all. Government overriding the price rather than paying it is what crowding out has always meant.
What looks like starvation is the market doing its job
Strip the verbs and look at what is happening. Capital is moving toward where its owners expect the highest return. That is the mechanism by which an economy reorganizes itself around something new, not a sign the machinery is broken.
Markets can get this wrong at scale, not just one firm at a time. A whole sector can herd into the same bet, and a good deal of this buildout is debt-financed and circular; if it is a bubble, a chunk of it is misallocation, not genius. But notice who clears that mistake. The railroad mania Harris cites ended in panic and bankruptcies, and left the track that opened the West; the dot-com bust vaporized fortunes, and left the fiber that carried the next decade.
None of that is painless. A bust this large takes real capital, real lenders, and real years down with it. But preventing it would require spotting the bubble in advance, the one thing neither a market nor a planner can reliably do. So the real choice is not damage versus safety; it is a bust that liquidates the error later, or a subsidy that freezes it in place and calls the freeze stability.
The strongest version of her worry, which she gestures at but does not quite make, is sharper still. By one Harvard economist's tally, data centers accounted for nearly all of America's economic growth in the first half of last year, and without the boom, interest rates and power prices would likely be lower. Those higher rates and bills land on people who never placed a bet on AI: the young family priced out of a mortgage, the manufacturer paying more to plug in. Grant all of it, without flinching.
It is still the cost of building something large, and it is paid through prices, which ration the scarce thing and summon more of it. A planner staring at the same shortage has no better information than the market does, and a worse incentive, because the capital he commands answers to a constituency, not a return. The cure for scarce transformers is more transformers, pulled out by their rising price, not a loan officer ruling that a transformer belongs in an approved industry rather than a data center.
There is one place her instinct has a real foundation. Private and social returns can part ways, basic research the clearest case, where the gains spill to everyone and no single firm can capture them. But that opening has to be argued, not assumed. It takes more than pointing at capital flowing somewhere you would rather it went, and Harris shows the flow without ever showing the gap.
The wreckage she points to is the last industrial policy
Consider the example she leads with. Manufacturing construction fell last year even as data-center construction climbed nearly thirty percent, and she reads it as AI strangling the factory floor. Some of that drop is higher interest rates, and some is data centers genuinely outbidding factories for the same crews and transformers, the priced competition described above.
But the largest piece is plainest. The slump is concentrated in computer and electronic plants, the exact category Washington inflated with hundreds of billions under the CHIPS Act, its semiconductor-subsidy law, and the boom it bought has simply crested. The biggest single force pulling the line down is not AI starving the factory floor. It is the last industrial policy wearing off.
The half-finished plant is not the case against the market. It is the case against her own remedy.
Her cure is the disease
Look at what she actually proposes. It is a wish list as old as economics: discounted loans steered to "vital sectors," an industrial policy that subsidizes the industries officials favor, and a curb on the buybacks that hand cash back to shareholders.
Grant her the strongest version of the buyback point, that some managers do buy back stock when they should be investing. But a curb does not make those managers invest any better. It only traps the cash with the very people she just called poor stewards of it, when a buyback would hand it back to shareholders free to move it to the next firm, including the starved start-ups she worries about. That is a problem for boardrooms and pay packages, not a reason to let Washington decide which industries keep the cash.
Each of her fixes does the same thing. It takes the decision about where capital goes away from the people whose money it is, and hands it to the government. That handoff is the one move in this whole story that genuinely starves an economy: not a rival industry bidding for chips, but a government commandeering capital, talent, and materials toward politically chosen ends and calling it foresight. You cannot plan your way to the allocation a market is already making, and the most recent proof is the half-built plant she just pointed at.
The railroad and the internet taught the same lesson, twice, in booms smaller and slower than this one. The gains were real, they came later than promised, and the busts everyone feared were how the capital found its best use. No planner produced that result then. None will now.
The hollowing starts the day Washington picks the winners.
