A skeptical look at AI investment
Daron Acemoglu, Jim Covello, Alison Nathan
While Goldman Sachs projects generative AI could boost U.S. GDP by 6.1% over a decade, MIT's Darren Acemoglu and Goldman's Jim Covello warn that high infrastructure costs and technical limitations may restrict automation to less than 5% of tasks. Driven by fear of missing out in a competitive "arms race," major technology firms are currently pouring capital into GPU-heavy infrastructure despite the absence of a cost-effective "killer application," sparking concerns that the sector mirrors the overbuild of the 2000 internet bubble. Analysts emphasize that the next 12 to 18 months serve as a critical inflection point where the emergence of tangible business use cases will determine whether this trillion-dollar investment cycle results in sustained productivity gains or a sharp correction in valuations.