Remember the housing bubble in 2008? When it burst, it caused the national housing market to collapse. That resulted in an $8 trillion loss in household wealth as home prices collapsed by 30%. That, in turn, had other knock-on effects, including huge stock market losses and a doubling of the unemployment rate to 10%. By the end of that economic calamity, it had cost the nation a total of $20 trillion in economic losses. Called the Great Recession, it was the largest economic collapse since the Great Depression in 1929.
When financial bubbles burst, they usually cause economic catastrophe. The housing market collapse is thankfully in the rearview mirror, but in recent months, there is increasing chatter about another bubble emerging—this one caused by the U.S. economy becoming so dependent on a few behemoth technology companies like Nvidia, Amazon, Google, Meta, and Microsoft who control much of the markets.
These companies are sinking unprecedented amounts of money into the development of artificial intelligence (AI) and the vast infrastructure needed to further this latest technological advancement. Certainly, AI is a promising new technology that will provide efficiencies and innovations throughout the economy and society. And to realize that potential, a lot of money will need to be invested in different applications and services.
But at this point, money is being thrown around like drunken sailors on a shoreside binge. And for the last ten years, that wild spending has become the prime catalyst for investment and the growth of the economy, creating jobs and destroying others. What happens if that investment is based on a bubble and the spending suddenly collapses?
Is AI innovation turning into a bubble?
The big Silicon Valley companies are investing heavily, but it’s not just in the AI technology itself. They are also mega-investing in gargantuan data centers and server farms that are the backbone of this development. Google, Amazon, Meta/Facebook, and Microsoft are on course to collectively spend around $400 billion on AI this year alone. Morgan Stanley analysts estimate that big tech companies will invest about $3 trillion on AI infrastructure through 2028.
To avoid burning up their own cash, these companies are instead taking on large amounts of debt to cover about half of the needed investment. A Goldman Sachs assessment found that key tech firms have taken on $121 billion in debt over the past year, a more than 300% uptick from the industry's typical debt load. Silicon Valley is taking on all this new debt with the assumption that massive new revenues from the invention of new AI-based products and services will cover the tab. But there is reason for doubt.
For example, OpenAI claims that it is planning to spend $1.4 trillion over the next eight to ten years on AI data centers and infrastructure, but its current annual revenue is no more than $20 billion. Most experts are in agreement that the current pace of investment in AI infrastructure far exceeds any foreseeable returns. The numbers just don’t add up.
In the meantime, not just the level of debt but the type of debt and financing that these companies are taking on is causing concern. It goes by odd names like “circular funding” and "special purpose vehicles,” which sound reminiscent of the shaky financial practices used during the housing bubble.
For example, recently, Nvidia pumped $100 billion into industry leader OpenAI to bankroll the building of more data centers. OpenAI is then supposed to purchase Nvidia chips made in those facilities. In other words, Nvidia is essentially subsidizing one of its biggest customers, giving OpenAI money to buy Nvidia chips, artificially inflating and propping up the price as well as actual demand for Nvidia chips. Meta also has a similar $27 billion private debt deal with Nvidia. It sounds like a Ponzi scheme, and the last time we saw this kind of circular funding was during the dot-com bubble in 2000-2002, when the giant energy company Enron collapsed catastrophically in 2001.
By other measures, such as the S&P 500 price to earnings ratio (P/E ratio), today’s stock prices are so inflated that they are even higher than the dot-com bubble's peak. Like an investment casino, a huge amount of money has poured into the AI sector in a very short period of time, to the point where even the CEO of Google, Sundar Pichai, says there are “irrational elements” in the investment patterns right now. Pichai says if the market crashes, the damage will be widespread; even highly capitalized Google will not be immune.
Or will AI investment lead to a transformative boom?
So yes, a new bubble certainly may be emerging, but there are some intriguing differences between this AI bubble and previous bubbles. For example, a handful of the AI companies, especially marquee names like Google, Microsoft, Nvidia, Amazon, and Meta, are hugely profitable with strong cash flows. So they might have enough of a financial cushion to withstand a large downturn. Also, the massive infrastructure build-out is producing real capital infrastructure in a way that the dot-com bubble or the housing bubble never did. Data centers and advanced computer chips are being laid down much like how railroad tracks were constructed all across the country in the 1870s through 1890s. To some, this suggests a long-term boom rather than a short-term bubble.
President Donald Trump’s AI czar (and venture capitalist) David Sacks says, "I don't think this is the beginning of a bust cycle. I think that we're in a boom. We're in an investment super-cycle."
Also, there are signs that businesses and consumers are starting to adopt real-world uses of AI in the workplace and industry, with a perhaps realistic hope that this will generate new efficiencies and eventually new jobs (though the jury on that is still out). JPMorgan Chase executive Mary Callahan Erdoes echoes this, saying, "We are on the precipice of a major, major revolution in a way that companies operate."
While this perspective also has a ring of truth, it’s hard not to notice that this "revolution" remains mostly speculative, and that those who are most enthusiastic have the most to gain from continued casino-level spending on AI.
Environmental impacts
Even if the AI bubble doesn’t pop, another concern has gained increasing attention: environmental impacts. That’s because their gigantic data centers and server farms consume huge amounts of electricity and water.
As massive, round-the-clock consumers of large amounts of electricity, these data centers are already contributing to upward pressure on wholesale utility prices. Construction Review reports that there are six mammoth data centers currently under construction needing to be fed by over one gigawatt (GW) of power—an amount sufficient to power 750,000 homes. In Frankfurt, Germany, where the government and private sector have invested heavily in AI, data centers have become the leading source of electricity consumption in the city, eating up 40 percent of the city’s total power demand. The local energy supply is being pushed to its limits.
Goldman Sachs has estimated that building the necessary energy infrastructure for AI data centers will require $1.4 trillion in investment by 2030. But as the Wall Street Journal has reported, “If the AI hype is overblown or the tech industry doesn’t ultimately need as much electricity as projected, other customers would get stuck with the infrastructure costs.”
Will the AI bubble collapse, and if so, how deep will the damage be?
During the dot-com crash from 2000-2002, the internet clearly transformed into a promising new technology, but telecom companies over-invested in transmission facilities for internet traffic. When the dot-com bubble finally popped, technology stocks dropped 80 percent, and half a million people lost their jobs as the unemployment rate zoomed to 7% (and 10% in the tech sector). Twenty-three telecom companies went bankrupt, including the collapse of the telecom giant WorldCom, which, at the time, was the single largest bankruptcy in U.S. history.
So bubble collapses can have catastrophic and widespread consequences, much like a cyclone ripping ashore. But in the longer term, the new technologies still result in fundamental transformation. The global internet networks got built, despite a worldwide financial collapse, and so too will AI get built. Even if the AI bubble bursts, the technology will advance, and winning and losing companies will emerge.
The great economist John Maynard Keynes once wrote, “When the capital development of a country becomes a by-product of the activities of a casino, the job is likely to be ill-done.” If the AI bubble creates a casino economy, with wild swings and layoffs and a stock market crash, but we also get new AI-based products and services in health care, energy, transportation, and more, will it all have been worth it?
We are in the middle of an experiment about how we define “progress.” It’s anyone’s bet how it will turn out.
Steven Hill was policy director for the Center for Humane Technology, co-founder of FairVote and political reform director at New America. You can reach him on X @StevenHill1776.
Why does the Trump family always get a pass?