The Spending Is Real, the Productivity Is Not
"There's still no sign of AI boosting overall US productivity," per Apollo.
The assessment comes from Apollo Global Management's economics team, which has tracked the gap between the enormous capital spending on AI infrastructure and the official productivity statistics, where the payoff is supposed to eventually appear.
The spending side of the ledger is not in doubt. Anthropic alone disclosed $895 billion in future compute commitments this week, and the hyperscalers have been raising capital expenditure guidance quarter after quarter.
US labor productivity, meanwhile, continues to grow at roughly its long-run pace of around 2 percent a year, with no visible step-change coinciding with the mass adoption of generative AI tools since 2023.
Why Economists Keep Looking for It
Productivity is the variable that decides whether the AI buildout ends as a boom or a bubble. If AI lifts output per worker materially, the debt and equity financing behind the data center buildout gets repaid out of real economic growth. If it does not, the spending is a very large bet on future revenue that has not yet arrived.
The historical comparison bulls reach for is the 1990s, when computers were everywhere except, for years, in the productivity statistics, until the payoff arrived late in the decade. Apollo's point is narrower and current: as of the latest data, the arrival has not happened.
The comment lands in the middle of the worst week for AI stocks since April, with the Nasdaq 100 down more than 5 percent in a session and traders already asking whether the buildout's financing layer is cracking, as a mystery seller auctions $500 million of data center loans.
Options and Market Impact
The productivity question is ultimately a question about the returns on the largest capex cycle in market history, and it prices through the builders: Nvidia, Microsoft, Alphabet and Meta.
Nvidia's repricing this week has been the sharpest expression of the doubt: the stock fell more than 11 percent from its October record levels as forward multiples compressed across the AI complex.
Apollo's framing gives bears a macro citation to set against the micro earnings story, which is why comments like this one travel fast on a down week. Watch whether upcoming productivity releases and earnings commentary from the hyperscalers answer it with numbers.
Bottom Line
Per Apollo, there is still no sign of AI in overall US productivity, despite spending commitments now measured in the hundreds of billions. Until that series moves, the burden of proof in the AI trade sits with the builders, and this week's selloff shows how quickly the market can start asking.
Track the flow as the story develops, and sign up for Unusual Whales to follow market tide, options flow and GEX in real time.