U.S. data centers alone may require $5 trillion to $6 trillion of investment over the next five years, and according to Jim Zelter, president of Apollo Global Management, this staggering figure is exactly why investors need to put the brakes on their AI enthusiasm.
But I am saying Goldman Sachs‘On the Exchange Podcast Thursday, Zelter drew on three decades of market cycles to draw a clear distinction between transformative technology and profitable technology. He referred to the cell phone era as a parallel: utility was never in doubt, but economic returns to capital owners were much less predictable.
“Just because companies need capital doesn’t mean they’re all great investments,” he said. The comment cuts to the core of the debate increasingly dominating Wall Street: whether the AI buildout is creating value or simply burning through it.
One of Zelter’s sharpest comments concerns the structural change that AI is imposing on the technology sector. Historically, software and platform businesses were particularly celebrated because they required little physical infrastructure. That dynamic is rapidly reversing.
“There is a big capital expenditure cycle going on that is turning asset-light businesses into asset-heavy ones,” Zelter said. For investors accustomed to valuing technology over margin expansion, this shift demands a fundamental rethinking of how risk is priced.
Zelter is not alone in his doubts. Oaktree Capital Management co-founder Howard Marks said in December that many market participants are approaching AI with a “lottery-ticket mentality.” Veteran economist Steve Hanke went further, telling Business Insider in February that AI is overhyped and potentially dangerous.
Adding to these concerns, a KPMG US CEO survey showed that three-quarters of CEOs at major companies believe that generic AI has been overhyped over the past year. Yet, nearly 80% of these CEOs told KPMG that they will allocate at least 5% of their budgets to AI investments in 2026.
In the case of Apollo, this influx of capital has opened up an opportunity for financing. However, Zelter highlights that this is an opportunity that demands discipline.
He pointed out that the risks associated with investments similar to equities cannot be considered fixed income risks and financiers should ensure strong downside protection before funding AI infrastructure firms.
