The scale of AI infrastructure spending has reached mind-boggling proportions, with tech giants and investors pouring unprecedented sums into the digital backbone that powers everything from ChatGPT to humanoid robots that’ll set you back $20K. But when the man who started the entire AI bull run is looking skittish, it may be time to plan the exit.

House of chips: Global markets have pulled back in the past week as investors increasingly question the stability of an unprecedented AI-driven bull market. But just how much is at risk if the bubble pops? Morgan Stanley estimates global data center expenditure will hit nearly $3T between now and 2028, with traditional tech companies covering just half that bill.

• Big Tech is expected to spend over $750B on AI capex in the next two years, with MicrosoftMSFT
  putting more than $60B into neocloud data centers — double last month’s pledge.
• Despite all this investment, MIT research showed that 95% of organizations got zero return on investment from their generative AI pilots, despite spending $30B-$40B across 300+ initiatives.

Wall Street’s Overexposure Sparks an Exit

One of the biggest worries is that the spending boom is increasingly funded by complex debt that could spill over into the broader economy if the AI bubble bursts. JP Morgan says AI-linked companies now make up 14% of its investment-grade index, overtaking US banks. Morgan Stanley estimates private credit — higher-risk shadow lenders — could supply over half the $1.5T needed for data center construction through 2028, raising fears of systemic stress if projects fail. This debt-driven exuberance is stirring uncomfortable flashbacks to previous market manias.

• Yale’s Jeffrey Sonnenfeld warns the maze of cross-investments — from OpenAI’s 10% stake in AMD to Nvidia’s $100B bet on OpenAI — resembles the “cable cowboy” days when nobody knew who owed whom.
• Deutsche Bank is looking to hedge its data center exposure by exploring AI-stock shorts and synthetic risk transfers to guard against potential defaults — a cautious approach in a potentially booming market.

Tensions are showing: In a recent interview, when asked how OpenAI could justify $1.4T in spending with only $13B in revenue, Sam Altman replied, “Enough.” And industry insiders are publicly voicing their concerns. Goldman Sachs CEO David Solomon warned investors should expect “a lot of capital that was deployed that [doesn’t] deliver returns,” and the Uptime Institute’s Andy Lawrence says many planned data centers “will never be built, or will be built and populated only partially.” Yale experts warn the bubble could burst through concentrated interdependencies, government conflict exposing AI limits, or quantum and chip breakthroughs that make today’s data centers obsolete — any of which could spark a 2008-style financial crisis.