Behind the celebratory headlines of $5tn tech valuations and $100bn quarters, a troubling narrative is emerging about the $3 trillion AI datacenter boom. Analysts are pointing to a trifecta of risks—risky debt, rapid depreciation, and doubtful returns—that could deflate the AI exuberance.
The most immediate concern is debt. A $1.5 trillion funding gap for new datacenters is being filled by private credit, a “shadow banking” sector. Mark Zuckerberg’s Meta has already taken $29bn from this market. Experts warn that lenders, desperate to invest in AI, are ignoring the “very quickly depreciating assets” they are funding. Harris Kupperman, a hedge fund founder, warned datacenters will depreciate twice as fast as the revenue they generate.
This “speculative” investment is flying in the face of doubtful returns. An MIT study found 95% of businesses piloting generative AI are getting zero return on their investment. This data undermines the “lofty revenue expectations” that generative AI will be a $1tn market by 2028, which is the core justification for the $3tn datacenter spend.
The Uptime Institute, which inspects datacenters, confirmed that many announced projects are pure hype. Its executive director said many are “speculative” and “will never be built,” or will only be “populated partially.” Alibaba’s chair, Joe Tsai, has also warned of a “bubble” in projects raising money without customer commitments.
While tech giants like Microsoft and Google are building on a foundation of massive cash flow, the wider boom is being propped up by “unproven” projects. If these debt-fueled, speculative builds fail, the resulting defaults could trigger “structural risk to the overall global economy.”
