A stark paradox is at the heart of the $3 trillion AI datacenter boom. While tech giants are spending “incredible” sums on infrastructure, a recent MIT study found that 95% of organizations are getting “zero return” from their investments in generative AI pilots. This “rattled” investor faith and threatens the entire premise of the boom.
The $3tn spend, including $750bn in two years from “hyperscalers” (Amazon, Meta, Google, Microsoft), is predicated on “lofty revenue expectations.” Morgan Stanley projects the generative AI market will soar from $45bn to $1tn by 2028. This is the assumption that has driven Nvidia to a $5tn valuation.
But the MIT data suggests this business “takeup” is failing. If 95% of businesses aren’t seeing a return, they will not become the paying customers the industry is counting on to justify the $3tn expense. This questions the very foundation of the “exuberance.”
This disconnect makes the “speculative” side of the boom even more dangerous. A $1.5tn “funding gap” is being filled by “private credit,” financing projects “without their own customers.” These projects are betting on a revenue stream that the MIT study suggests may not exist.
The industry is hoping the 800 million weekly users of ChatGPT are a sign of unstoppable demand. But the “zero return” figure from corporate pilots suggests the $3tn boom is being built on consumer hype, not a sustainable business-to-business revolution.