The financial logic of the current AI boom is under scrutiny, with a top analyst arguing that massive capital spending by cloud giants is on a path to destroy shareholder value, setting the stage for a wave of initial public offerings designed to transfer risk to retail and institutional investors.
"These IPOs are essentially a large-scale transfer of investment risk from current holders to retail investors, pension funds, and others willing to buy the story before the hype truly subsides," Joachim Klement, a managing director at Panmure Liberum, wrote in a recent analysis.
U.S. corporate spending on IT equipment and software has swelled to nearly $1.5 trillion in 2025, with tech investment accounting for 93% of the nation's GDP growth over the past year. Klement's analysis projects that capital expenditures by hyperscalers like Microsoft and Google are set to grow 20% annually, outpacing expected revenue growth of 15% and implying negative returns on new AI data center investments.
The core issue is that for these investments to generate a 10% return, cloud providers would need to create an additional $2 trillion to $5 trillion in annual revenue, a staggering figure compared to their current combined total of about $1.5 trillion. This sets up a potential market crash in 2027 or 2028 that could mirror the dot-com bust, which saw tech stocks lose over half their value in the first year.
A Bubble Bigger Than the Dot-Com Era?
The scale of today's AI enthusiasm dwarfs the tech, media, and telecom (TMT) bubble of the late 1990s, according to the report. At the peak of the TMT boom, annual U.S. corporate IT spending was about $466 billion, or $829 billion adjusted for inflation—less than half of today's nearly $1.5 trillion figure.
More critically, the economy's reliance on this spending is unprecedented. While tech investment drove about 60% of U.S. GDP growth at the height of the dot-com bubble, it accounts for an estimated 93% of it over the last four quarters. This concentration of economic drivers in a single sector creates significant downside risk. If tech investment contracts by even a modest four to six percent, the U.S. economy could quickly face a recession, Klement argues.
"These numbers indicate that if the hyperscalers continue on their current trajectory, the AI boom will turn into one of the largest-scale destructions of shareholder value in history," Klement said.
IPOs as a Risk Transfer Mechanism
Against this backdrop, the rush of AI companies toward the public markets is seen as a strategic move to cash in on peak sentiment. Foundational model developers like OpenAI and Anthropic are reportedly preparing for listings later this year. Their IPOs would join a market that has already shown a fervent appetite for AI-related offerings.
Chipmaker Cerebras Systems (CBRS) saw its shares surge 68% on its first day of trading, raising over $5.5 billion in its offering. The IPO pipeline includes Elon Musk's SpaceX, which confidentially filed last month and could seek to raise between $70 billion and $75 billion.
Klement's analysis frames these impending IPOs as a window of opportunity for early venture capital and private equity investors to exit their positions at high valuations. The risk of negative returns from unsustainable capital spending is effectively passed on to public market participants, including individual retail investors and pension funds that will be the ultimate bagholders if the financial logic does not prove out.
For investors, the analysis serves as a stark warning. While the AI narrative continues to propel markets to new highs, the underlying financial models for the companies building the infrastructure show signs of strain. The report suggests that while a pullback is unlikely in 2026, the "impossible math" will eventually assert itself, with a high probability of a reckoning in 2027 or 2028.
This article is for informational purposes only and does not constitute investment advice.
