One unifying dynamic in America today: Americans really, really hate artificial intelligence. What began as a form of NIMBYism — no one wants a data center in their neighborhood — has broadened into severe doubts about the technology as a whole. For those of us who remember the giddy optimism that greeted the tech boom of the 1990s, the widespread negativity toward AI is astonishing. Here are the results of a new Wall Street Journal poll:

Only 28 percent of Americans believe that AI technology should move full speed ahead. Everyone else wants it slowed down or stopped, temporarily or permanently.

Of course, Donald Trump remains a huge booster. He believes that the public’s antipathy towards AI is just a marketing problem — and that he can solve that problem by forcing everyone to call it “super intelligence”. But the Journal poll also says that only 7 percent of Americans trust Trump’s leadership in regulating AI. Given Trump’s abysmal approval ratings, his advocacy of AI — err, super intelligence — probably taints the technology even further in the eyes of Americans.

Historically, Americans have been technology enthusiasts. As a nation, we’re not Luddites. So what explains Americans’ dislike of AI?

Obviously, there are near-daily prominent warnings that AI will destroy the human race or, failing that, destroy everyone’s job. And these warnings aren’t coming from left-wing cranks — they’re coming from the AI industry itself.

And we should be afraid – very afraid. The power of AI models to hack into supposedly secure systems — and to do so on their own, without prompting and against their owners’ intentions, and to lie to the humans about it — is truly shocking. Moreover, there is surely an element of financially motivated hype: “Our technology is so powerful that it can destroy the world, so you’d better invest in our IPO.” And there is more than a little bit of karma here: the race by the AI giants to reach the holy grail of AGI – artificial general intelligence – was surely a strategy to make vast riches from Wall Street, in contrast to the much more modest (and probably more successful) Chinese strategy of building modest, cost-effective models.

There is also, however, a more prosaic reasons for the public’s dislike of AI: This is a technology of, by and for oligarchs, with hardly any of the benefits trickling down to regular Americans.

Or to put it a different way, never before in history have corporations spent so much money — playing a major role in soaring interest rates — to create so few jobs.

Data centers themselves are basically banks of blinking servers that employ hardly any workers. Here’s data from the Bureau of Labor Statistics on a category that includes data centers as well as a number of other activities. Employment in that category has declined since November 2022, often cited as the start of the AI boom because it’s the month ChatGPT was released to the public:

Source: Bureau of Labor Statistics/Created with Datawrapper

Still, won’t the AI boom at least temporarily create a lot of construction jobs? It will create some, and Jared Bernstein has a good post on how to ensure that these are good jobs. But all indications are that AI investment will create far fewer jobs, even in construction, than expected.

Bear in mind that we are talking about a truly huge investment boom. In Sunday’s primer I cited estimates by Stijn van Nieuwerburgh suggesting that we’re entering by far the biggest capital-expenditure boom in U.S. history, dwarfing even the huge railroad boom that followed the Civil War:

These are projections rather than spending that has already happened. But the AI investment boom is already well underway. Here’s IT investment as a percentage of GDP, which really took off in late 2024 and is already well above its level during the 1990s tech boom:

Source: US Bureau of Economic Analysis via St. Louis Federal Reserve (FRED)

Given this spending surge, one should expect a sharp rise in nonresidential construction spending — basically construction for businesses rather than housing.

But that’s not what we actually see. Nonresidential construction rose rapidly during the Biden years, largely thanks to that administration’s promotion of clean energy. But nonresidential construction has basically flatlined under Trump, despite the immense AI investment boom:

How is this possible? Because even the physical construction of a data center involves relatively little construction. According to Van Nieuwerburgh, when a multi-billion-dollar AI training campus is built,

About one-third [of the investment spending] reflects the facility and power infrastructure, while roughly two-thirds consists of compute hardware and related IT systems.

So not a lot of conventional construction. And most of the “compute hardware” etc. is imported!

Overall, then, the deep unpopularity of AI isn’t hard to explain. Here we have a technology that is enriching a small number of people, but whose creators warn may destroy millions of jobs if not the world. And even the usual argument for pro-oligarch policies — job creation! — falls flat because hardly any jobs are being created even in the short run.

Will AI’s unpopularity matter? Indicators are that it will. So many things are going wrong for Republicans right now that it’s hard to single out any one issue, but Chris Caldwell has an interesting recent article arguing that the deep unpopularity of data centers may be the tipping point that turns Ohio blue this year, possibly deciding control of the Senate.

Thus there are good reasons, even beyond possible apocalypse, for the public backlash against AI. And that backlash is so strong, and has emerged so quickly, that it may have important political consequences.

Isn’t that super?

Paul Krugman is a Nobel Prize-winning economist and former professor at MIT and Princeton who now teaches at the City University of New York's Graduate Center. From 2000 to 2024, he wrote a column for The New York Times. Please consider subscribing to his Substack.