Artificial Intelligence is a powerful and transformative technology. But the magnitude of AI spending and growing debt is dangerous to the overall health of the markets and the economy.
AI is a transformational technology. No question.
In a short amount of time, AI has gained widespread acceptance and use among individuals and businesses alike. AI is transforming the way we access information and the way we work.
This transformation is being fueled by the rapid adoption of AI models developed by OpenAI and Anthropic. And simultaneously, this growth is facilitated by the massive investment in the AI infrastructure required to run these models at scale. These capital expenditures have created jobs, spurred demand for raw materials like copper, and added support to the economy and markets at a time when consumer confidence is at an all-time low.
The Big Spend
The big players in the AI build-out, that is, the “hyperscalers” (Amazon, Google, Microsoft, and Meta) will spend roughly $730 billion combined on AI capital expenditures this year. And that is on top of $410 billion that was spent last year, and $226 billion the year before.
This level of capital expenditures is equivalent to roughly 2.3% of US GDP. At the peak of the telecom sector’s spending in the year 2000, a period of time that would later be known as the “dot-com bubble,” spending on telecom and fiber infrastructure rose to roughly 1.2% of GDP.
In fact, the best historical comparison to today’s AI spending might be the great railroad expansion of the late 1800’s. Like AI spending today, capital spending on railroad infrastructure also averaged 2.4% of GDP.
And how did that end?
Not well–at least in the short term. There was a stock market panic in 1873 that closed the NYSE for 10 days. And during the subsequent two years, new construction dropped by 80% and roughly 18,000 businesses failed.
Longer-term, railroads fulfilled the promise to connect a new and sprawling country and enabled the rapid growth of the 20th century.
Put a different way, the investment in the railroads was completely appropriate because it was a transformative technology that enabled economic growth on a scale previously impossible. However, implementing such a huge investment through a fabric of independent, profit-seeking enterprises financed with debt was fraught with risk.
The current AI build-out could prove similar; a transformative technology, yet the unprecedented size and pace of capital flowing into its expansion may unsettle markets over the near and intermediate term — especially if AI business prospects dim.
Concerns Growing
There are always nuances and historical differences that are important to consider when comparing today’s environment to past economic crises. First off, the railroads were financed predominantly by borrowing. Railroads issued bonds to pay for the new railroad lines. And some railroads inevitably became overextended. Then, when earnings failed to keep pace with the interest payments due on the bonds, railroads failed.
Today’s AI spend has a bit more runway. To start with, most of the AI spending has been funded through ongoing operations of the very profitable hyperscalers. However, this year the magnitude of the spending has eclipsed the hyperscalers free cash flow. As a result, debt has started to build. And companies such as Meta and Google have announced plans to issue more stock as another source of funding in the AI arms race.
A warning that borrowing could be reaching a dangerous point came on Oct. 13 when the International Monetary Fund issued a report on financial stability. So far, soaring AI investment has been “an important source of support for the global economy and financial markets,” it said. That’s exactly why an AI stall would be dangerous. The report went on to warn of “rising leverage in the AI ecosystem,” concluding that “the consequences of an AI correction could be global.”
Some AI-related stocks and bonds sold off sharply the very week the I.M.F. issued its warning. Shares of CoreWeave, a U.S. operator of AI data centers, fell 25 percent that week and an additional amount the following week, leaving them down more than half from their 52-week high.
It is impossible to know if a stumble in a highly-leveraged AI company like CoreWeave is analogous to the subprime stumble that preceded the much larger financial crisis of 2008. But it absolutely raises concerns.
Anthropic IPO
Anthropic, the developer of the Claude AI chatbot, is a five-year-old company that is planning to issue stock to the public this year in what could be the largest IPO ever.
Anthropic’s revenue was $16.3 billion in the first half of the year, and its net operating loss was $700 million. Granted, the company is growing very fast and still in its early stages, so one might not expect it to be profitable yet. But remarkably, the company has committed up to $807 billion in contracts for data centers, chips and computing power, many of which cannot be canceled, over the next 10 years, according to a tally by The New York Times.
Just to restate these jaw-dropping statistics… Anthropic has annualized sales of $30-$40 billion, which have resulted in losses (not profits) of around $1.4 billion, and they’ve committed to spend over $800 billion in new infrastructure investment over the next ten years. One might ask how this is even possible, won’t Anthropic simply run out of money?
The numbers underscored how much Anthropic is betting that its future business will justify its outlays. Even its public offering, which could raise more than $100 billion, may not provide the company with enough money to fund its intense infrastructure needs, as it warned that it might need to seek additional financing sooner than anticipated.
Two Big Differences
In comparing today’s AI expansion to the dot-com bubble, or other past bubbles, one notices two big differences. First, AI infrastructure is being expanded so quickly due to high current demand. That is very different from the late 1990s, when telecom infrastructure was being expanded to meet expected future demand. Today’s demand for AI is real and growing. Data centers are being constructed to accommodate today’s demand. Data centers are not being constructed to meet some expectation of future demand that may or may not come to pass.
Secondly, stock market bubbles almost always have one thing in common: unreasonable investor euphoria. This was certainly true during the dot-com era, the housing bubble that ended in 2007, the railroad mania, and the so-called “Roaring Twenties” that ended with the stock market crash of 1929.
Today, one might describe the average investor as being complacent, but not euphoric. High investor complacency often precedes market volatility, or market corrections. These periods tend to restore investor caution and remind investors that market risk is a real thing. And in fact, historically speaking, one could certainly argue that the stock market is due for such a pullback.
Investment Markets
Markets continue to benefit from strong corporate earnings and AI spending. If the S&P 500 simply remains at its current level through 2026, it will record its fourth consecutive year of double-digit returns, with two of those years producing returns well over 20%. The last time the S&P 500 recorded four consecutive years of double-digit returns was 1995-1999. This was the run-up to the dot-com bubble, which ended in 2000.
Unlike the dot-com era, today’s stock market returns are built on the sizable profits produced by big tech companies. And luckily, these earnings are not likely to disappear anytime soon. However, the concentration of big tech within the S&P 500 has grown to a historical extreme.
AI-related companies account for nearly half of the S&P 500’s total value. And the strength in AI has masked weakness in market stalwarts like Home Depot, McDonald’s, and Nike.
Some experts believe that an AI crash could send the overall market down 40 percent and a subset of the market down 70 percent. This may sound alarmist, but it’s less than what happened when the dot-com bubble burst in 2000. While a stock market meltdown isn’t as systemically dangerous as a wave of defaults, a big drop would make people feel poorer.
Bank lending to private equity and credit rose 26 percent this year, according to the I.M.F. report. “History teaches us to worry when any form of credit, especially a new one, grows this fast,” said Mr. Hanson, the Harvard Business School economist. Lending by private credit firms is not inherently risky, he said, but “You’re just worried more about unknown unknowns.”
Higher interest rates are hurting, too. AI stock prices are based on profits expected years from now, and the higher the rate, the less those distant profits are worth today. AI doesn’t have to fail for high rates to do damage. It just has to succeed more slowly than its lenders and shareholders are counting on — making it one more reason that there might be more peril in U.S. markets than meets the eye.
There’s a truism in finance circles that crises take longer, and then happen faster, than you think. There are meaningful parallels between the AI economy of today and that of past boom-and-bust cycles. However, there are also meaningful differences. Therefore, it is likely that the best strategy for investors at this juncture is to stick to their long-term plan.