We are witnessing one of the largest peacetime mobilizations of capital in modern American history. Topping $1 trillion annually by next year, the artificial intelligence buildout is expected to rival or surpass previous technological booms at their peaks \u2014 rail, electrification and the internet revolution.<\/p>\n
Many economists believe that at a time of rising inflation, a weakening job market and global unrest, this boom is keeping the U.S. economy afloat. \u201cA recession tied to the balloon of A.I.,\u201d is how PJ Vogt, a popular podcaster, describes the perspective. Look more closely, however, and the picture changes. A.I. is vacuuming up so much of our land, talent, semiconductor chips, building materials \u2014 and, above all, so much of our money, that it is beginning to crowd out the rest of the economy.<\/p>\n
In other words, A.I. isn\u2019t merely compensating for the weakness in the rest of the economy. It is, at least in part, causing it.<\/p>\n
Jason Thomas, research head at the investment firm Carlyle, noted in a January<\/a> report that data center investment may be swelling to the point that it could consume virtually all the private money available for new, non-housing investments. Researchers, economists<\/a> and other market analysts<\/a> are ringing the same alarm. They are particularly worried that the deluge of investment, much of which is plowed into data centers, is beginning to starve the rest of the economy of the money it needs (to say nothing of the talent and physical materials).<\/p>\n The money flowing to A.I. is bypassing some of our country\u2019s highest priorities.<\/p>\n Start with housing. New homes that could ease the affordability crisis aren\u2019t getting built, as land slated for houses is sold instead to data center developers. Consider Prince William County in Northern Virginia, a region that suffers from an estimated shortage of more than 75,000 homes. A residential developer who had purchased land there for just over $50 million with plans to build homes ultimately sold a portion of the land<\/a> to Amazon for $700 million.<\/p>\n Land costs are higher \u2014 at times more than 17 times as high as they were even three years ago in pockets near Dallas. Scott Finfer, a local residential land developer, told The Wall Street Journal<\/a> that, for home builders, \u201cThere\u2019s no possible way you can make those numbers work.\u201d The financiers who help decide how much new housing gets built and where it goes are shifting their money away from homes and into the data center rush.<\/p>\n The push to reinvigorate American manufacturing, which has been championed by both major political parties, is also in peril<\/a>. Joseph Brusuelas, chief economist at the accounting firm RSM, worries that<\/a> \u201cthe combination of government borrowing and A.I. investment may be crowding out other industrial sectors from obtaining capital and critical materials.\u201d<\/p>\n Investment in manufacturing construction declined sharply last year, while spending on data center construction rose nearly 30 percent year over year by the end of the year. Projects to build<\/a> alternative energy sources and electric vehicle charging<\/a> infrastructure are also stalling out as builders and supply chains stampede to the more profitable business of data center construction.<\/p>\n Then there is venture capital, investments that help determine which industries will drive our economy in the future. A.I. firms captured nearly two-thirds of all global venture capital investment in 2025, up from roughly 30 percent in 2022.<\/p>\n A prominent venture investor lamented to me how an A.I. idea scribbled on a napkin can bloom into a billion-dollar valuation, while hot companies in other hot sectors \u2014 companies actually in operation \u2014 can\u2019t get the funding they need to grow. As a handful of A.I. superstars explode, the \u2018\u2018middle class of start-ups is hollowing out,\u201d Roy Bahat, head of venture firm Bloomberg Beta, told The Washington Post. Funding for start-ups without the luster of an A.I. phenom has plummeted to a decade low, according to Silicon Valley Bank<\/a>.<\/p>\n This dynamic may even be driving inflation. Red-hot demand for scarce semiconductor chips, in particular, is jacking up the cost of consumer goods that rely on them, from cars to laptops to phones. On Thursday, Apple raised prices<\/a> of some Macs and iPads by $200 or more: its chief executive, Tim Cook, called such hikes \u201cunavoidable\u201d given surging memory and storage costs.<\/p>\n This capital starvation is even bad for the A.I. industry itself, as so much of its future is dependent on selling its wares to other businesses. Companies may be too cash-strapped to switch out old machinery and train the staff they need to best utilize the technology. Instead, firms may just rely on A.I. mainly as a cost-cutting tool (that is, replacing workers).<\/p>\n All the investment funneling into A.I. bears similarities to the early years of railroad expansion and the internet.<\/p>\n The railroad buildout that began in the 1820s absorbed a yearly average of 2 percent of America\u2019s gross domestic product during the 1850s<\/a>. But years later, when the railroad boom didn\u2019t deliver the financial benefits investors promised, the economy sank into a depression. Roughly 18,000 businesses vanished within two years. By 1876, unemployment had reached about 14 percent.<\/p>\n The internet and dot-com booms tell a similar story. The level of investment was comparable to today\u2019s A.I. investment. But, like rail, the internet\u2019s initial productivity benefits were far more muted<\/a> than popularly understood, and were largely exhausted by 2004.<\/p>\n The recession of the early 2000s stemmed in part from what the former Goldman Sachs analyst Jeff Currie called the \u201crevenge of the old economy.\u201d In a 2023 interview, Mr. Currie said<\/a>, \u201cThe term was coined in February 2002, and was meant to capture the fact that over the previous decade, the dot-com boom had stolen so much capital from the old economy,\u201d that it choked off the investment it needed to grow.<\/p>\n The lesson isn\u2019t that the rail or internet buildouts weren\u2019t worthy investments \u2014 or that A.I. isn\u2019t worthy today. The railroad eventually helped settle the West and create Los Angeles; the internet birthed entire new industries like digital communications, cloud computing and telemedicine and transformed everyday life for many of us. The lesson is that until these benefits manifest themselves, these technological booms can entail formidable opportunity costs as investment-hungry portions of the economy go without. Often, these costs, if not correctly managed, can lead to recession.<\/p>\n What is clear in hindsight is that the economic benefits of a new technology don\u2019t just happen. They are determined by the myriad choices we make about how to use it. Right now, just as was the case with rail or the internet, the economic gains from the technology are initially lagging<\/a>. <\/p>\n What, then, to do? Historically, we have largely just sat on our hands, suffered the consequences of booms and waited for the benefits to show up. But we can do better than that. We can take action to address the growing pains that accompany technology booms. Fortunately, we can utilize the policy tools we already have, the ones we deploy to dig ourselves out of recessions or to temper speculative bubbles.<\/p>\n The federal government and the Federal Reserve should see that vital sectors \u2014 such as housing, entrepreneurship, energy infrastructure and critical supply chains \u2014 don\u2019t starve as we continue building A.I. This could take the form of discounted lending to boost investment to these areas. Shortages in almost all these critical sectors are known inflation culprits. With inflation reigniting, the deflationary potential of investing<\/a> in these branches is reason enough to do this.<\/p>\n Next, policymakers could develop an honest to goodness industrial policy for A.I. \u2014 subsidies, tax incentives, regulations and different ownership structures \u2014 that prioritizes deploying the technology for the kind of economy we want to live in: inexpensive, with clean energy, biomedical breakthroughs and advanced manufacturing.<\/p>\n Finally, introduce reforms to the myriad rules our corporations must follow, to focus them more on developing new products and services and less on the kind of financial engineering that simply gooses a company\u2019s stock price. One could be a curb on the explosion of companies using their extra cash to purchase their shares on the market, which research<\/a> suggests suppresses innovation. Greater experimentation and more patient shareholders could accelerate the process of turning a new technology into a tangible economic benefit.<\/p>\n The economist Robert Solow\u2019s famous quip \u2014 that the computer age showed up everywhere but in the productivity statistics \u2014 can be viewed as a point about patience. We might, though, reinterpret it as an invitation to seize the wheel. We could avoid another period of price spikes and recession while we await the uncertain payoffs. Instead, we can see that A.I. really transforms everyday life for the better and also ensures that the ride there proves a little smoother than the one we appear to be in for.<\/p>\n Jennifer M. Harris directs the Economy and Society Initiative at the William and Flora Hewlett Foundation. She previously served as an economic official in the Biden White House.<\/p>\n The Times is committed to publishing a diversity of letters<\/a> to the editor. We\u2019d like to hear what you think about this or any of our articles. Here are some tips<\/a>. And here\u2019s our email: letters@nytimes.com<\/a>.<\/p>\n Follow the New York Times Opinion section on Facebook<\/a>, Instagram<\/a>, TikTok<\/a>, Bluesky<\/a>, WhatsApp<\/a> and Threads<\/a>.<\/p>\n