Jim Covello has been saying the quiet part out loud about AI investing for two years now. The difference is that in 2024, the market mostly shrugged. In June 2026, with over $1 trillion in planned AI-related capital expenditures on the books and profitability still largely theoretical, the shrugging has gotten a lot harder to justify.
Covello, Goldman Sachs’ head of Global Equity Research, used a recent podcast appearance to lay out what he sees as a widening disconnect between what companies are spending on AI and what they’re getting back. His assessment was blunt: the economics of AI are “more questionable today than two years ago.”
The trillion-dollar question nobody wants to answer
Goldman Sachs has noted over $1 trillion in planned AI-related capex. And yet, as Covello points out, the buyers and developers pouring money into AI infrastructure have yet to establish clear profitability from those investments.
Covello distilled his concern into a single line:
“At some point, you’ve got to make money.”
A skeptic with receipts
Covello has been one of Wall Street’s most prominent AI skeptics since at least 2024, when he co-authored a widely referenced Goldman Sachs report that asked whether the trillions being invested in AI infrastructure would actually generate matching economic value.
That July 2024 report estimated it would take 18 months to two years for companies to see returns on their AI investments. We’re now past that window. The broad-based profitability that was supposed to validate the spending binge has not materialized at scale.
The gap between AI spending and AI profitability hasn’t closed. By Covello’s reading, it has grown. That’s a problem when you’re staring down a wave of AI-related IPOs from companies like OpenAI and Anthropic, both of which are expected to test public markets in the near term.