The $3 trillion AI datacenter spending spree is being met with a healthy dose of skepticism from industry watchdogs, who warn that many of the “fanfare” announcements are “speculative” and may “never be built.”
Andy Lawrence of the Uptime Institute, which inspects and rates datacenters, cautions against believing every headline. He stated that many of the announced projects “will be built and populated only partially, or gradually, over a decade.” This suggests the $3tn figure may be inflated by hype.
This warning is echoed by industry leaders. Joe Tsai, chair of Alibaba, said in March he was seeing “the beginning of some kind of bubble,” pointing directly to projects “raising funds for construction without commitments from potential customers.” This is the “speculative” side of the market that worries economists.
These “speculative assets” are being funded by a $1.5tn wave of private credit, which is being deployed “without properly assessing the risks.” This is happening while an MIT study shows 95% of businesses are getting zero return on their generative AI pilots, raising doubts about who will eventually use and pay for these datacenters.
While tech giants like Microsoft and Google are undeniably building massive, “healthy” infrastructure projects, they are part of a larger, “exuberant” market. The risk is that the “speculative” half of the boom will collapse, taking billions in debt with it and potentially damaging the “healthy” half in the process.
