AI investments are driving the American economy. That’s a dangerous tightrope. By Matteo Wong Over the past several days, as Donald Trump has flown across the country to rally Republican voters in contested midterm races, he has returned again and again to a version of the same point: “I had the greatest economy in the history of our country in my first term,” as the president told crowds in Michigan. “This term is going to blow it away.”To the extent Trump can claim the economy as a strength, what he’s mostly talking about is the biggest infrastructure build in American history.

The five largest AI-data-center companies, including Amazon and Google, are on track to spend $900 billion on AI in 2026, roughly as much as they did in the previous three years combined. Just two AI chipmakers accounted for one-third of growth in corporate profits this past quarter, while as of September, AI stocks accounted for nearly half of growth in the S&P 500 in 2026. Depending on how you count, AI investments might already make up more than 8 percent of the nation’s GDP.

Then there’s the rest of the economy, which is, if not struggling, at least confusing. AI has been pushing up the stock market for years, but the scale of financial expenditures is beginning to frighten investors, prompting a recent tumble. Gas and grocery prices are rising, while wages are largely flat or falling. There are tariffs and a miserable housing market. Unemployment is low, but at the same time, many Americans aren’t looking for work and job growth is slow. Consumer spending is strong, but that’s due in no small part to AI doubling the value of people’s stock portfolios since roughly the launch of ChatGPT.

For years we’ve spoken of AI as a “boom,” describing explosive growth in one industry. It is more correct to speak of the entire market as the AI economy.

Meanwhile, there’s real anxiety around the products these AI companies are selling. AI-powered hacks have brought about a cybersecurity apocalypse. OpenAI and Anthropic, the two companies whose futures the U.S. economy rests on, are losing control of their own agents and warning that these bots could end the world. The AI industry appears to be heading in two different directions—one financial, the other technological. Combined, their effect is to induce a sort of dependency: to bring about so much rapid change that AI becomes impossible for us to quit.

The AI build-out was always big, but of late it has become gargantuan. Silicon Valley is committed to pushing as much money, human talent, electricity, concrete, and land toward AI as possible. The scope of the investment is staggering, estimated to be a bigger portion of GDP than the railroad boom and interstate highway system were, combined. Anthropic’s public offering next month could be the largest ever, perhaps valuing the company at $2 trillion. According to a recent Securities and Exchange Commission filing, Anthropic already has up to $807 billion in future contractual obligations for AI infrastructure to pay and expects “expenditures to increase significantly.”

For reference, U.S. defense spending was $954 billion in 2025. The biggest cloud companies, similarly, have nearly $3 trillion in contractual obligations already on the books, and will soon surpass $1 trillion in annual spending. Earnings reports last month showed that many of these companies—Meta, Google, Amazon—are already losing money because of their AI investments, which is stunning given that they all generate hundreds of billions of dollars a year in revenue.

The role the AI industry has claimed for itself is in essence that of a central bank, managing the economy and issuing money to keep it afloat. The tremendous costs of data centers require the economy to run hot. AI firms need their customers to have money on hand in order to sustain meteoric revenue growth. But, according to at least one analysis, the AI industry’s cash-flow projections dwarf the far more modest ones of its biggest customers, raising the question of who’s going to fund AI’s continued development. Similarly, a robust stock market is necessary to attract more investors.

And more wealth all around makes it easier for Meta, Google, and the like to issue debt to fund more data centers; this summer, foreign investors purchased more bonds from U.S. companies than from the federal government. So those same firms keep putting money into the flywheel: Nvidia, for instance, is dumping tens of billions of dollars into data centers used by distressed cloud companies that will, in turn, spend that money to buy Nvidia chips. It’s very possible the U.S. economy would be in a recession, or at least stagnant, without the AI boom. But it’s almost certain that there would be no AI build-out if the economy were to flail.

Another way to think of the AI industry’s role in the broader economy is as a parasite, starving the rest of the organism. There are historic precedents for this: In 1959, for instance, oil companies in the Netherlands discovered enormous natural-gas reserves in the North Sea. The Dutch natural-gas industry grew rapidly—as did the nation’s GDP—but at the expense of other industries, including manufacturing, which withered. Today, in the U.S., profit margins are soaring at the biggest tech companies but flat for the rest of the S&P 500.

Data-center construction is booming, but all other private construction is in free fall. The only source of private investment contributing to GDP growth is tech, according to an analysis by Pacific Investment Management Company. Meanwhile, the boom in demand for chips, natural gas, and construction materials driven by data-center build-outs is helping drive up inflation and make everyday life still more unaffordable. Silicon Valley is thriving, as are many broad macroeconomic indicators, but if you scratch beneath the surface, you’d find the finances of many Americans floundering.

And what’s it all for? A large part of the reason for the step change in AI spending this year has been the rise of agents—programs, such as Claude Code and Muse, that write software and take complex actions on users’ behalf. Over the past few months, these products have caused chaos. Agents are being used by hospitals to charge more for health care and gaining unauthorized access to bank accounts. Hackers are breaching those hospitals, as well as governments, utilities, and financial instutions—recently pulling off perhaps the biggest cyberheist ever, draining billions from the crypto exchange Binance—and threatening the stability of the internet as we know it.

That’s not to mention the ways the AI boom has, for years now, left classrooms a mess, driven up carbon emissions, and polluted the internet with slop. The strongest part of the American economy is also seemingly the most wasteful and destructive.

The societal benefits of AI are largely either vague or delayed. A wave of AI-discovered drugs has yet to hit pharmacies. The technology has yet to solve climate change, as top executives including OpenAI CEO Sam Altman and Nvidia CEO Jensen Huang have claimed it will do. Economy-wide data do not show clear signs of AI-driven productivity gains.

The AI industry is conscious of the dissonance between their economic and technological ambitions and the mess their products are making. This fall, in an interview with The New York Times’ Ezra Klein, Huang likened the cost of developing the technology to surgery: “They’ve got to cut you open to save you. They’ve got to inflict an enormous amount of pain and suffering on you so that they can save you. And so I think AI’s kind of like that.”In the meantime, thank Nvidia for your 401(k) doing very well.

A year ago, there was a sense that the AI boom had put the American economy in a sort of “benevolent hostage situation.”Now the industry seems to be holding not just the American economy but our entire society hostage: Nvidia, Google, and the like had better succeed financially, or there go your savings. Anthropic and OpenAI had better succeed in their battle against AI-powered hacks, or there goes the internet. The technology had better be used to devise new carbon-free technologies, personal tutors for every kid, and widely available mental-health resources.

There’s a perverse way in which even strident AI critics may be forced to cheer on some version of this industry: The only thing more destructive than the AI boom might be a crash.