OpenAI CEO Concedes That Questions About AI’s Return on Investment Are Legitimate
Sam Altman, CEO of OpenAI, acknowledged Monday what critics and investors have been saying for months: the AI industry is burning through enormous sums of money without clear proof that the returns will justify the costs. Speaking in a CNBC interview, Altman called the concern “the most fair criticism right now of AI.”
“You hear companies saying, I am spending a ton of money on AI. And I know some great stuff is happening, but I know there’s a ton of waste,” Altman said. He added that businesses are increasingly asking how long they must wait before AI spending shows up in revenue — and when costs will come under control.
A Spending Boom With Thin Evidence of Payoff
The admission lands at a precarious moment for the industry. An April report from The Wall Street Journal revealed that OpenAI itself missed key revenue and user growth targets last year — a notable stumble for the company at the center of the AI investment frenzy.
The broader picture is equally troubling. Data from cloud optimization platform Cast AI, drawn from an analysis of 23,000 clusters across thousands of companies, found that average GPU utilization sits at just 5% — meaning roughly 95% of provisioned graphics-processing capacity is sitting idle.
Cast AI cofounder and president Laurent Gil said companies are hoarding scarce AI chips out of fear of missing out, not because they have immediate operational needs. The result is a growing stockpile of expensive, underused computing infrastructure.
Researchers Warn of Historic Capital Misallocation
Gary Marcus, a longtime AI researcher and professor emeritus at New York University, has been more blunt in his assessment. He described some companies’ AI capital expenditure plans as the “greatest capital misallocation in history.”
In a post on X, Marcus noted that Amazon, Google, Microsoft, and Meta are collectively spending more on AI infrastructure each month than the entire Manhattan Project cost — without demonstrating returns that match the scale of those investments.
The pattern reflects a dynamic common in speculative technology booms: infrastructure spending races far ahead of actual productive use, driven by competitive anxiety rather than demonstrated demand. Whether the AI industry can close that gap — and how much economic damage accumulates in the meantime — remains an open and consequential question.