Economy 5 min read By Bethany Hadley
Goldman Sachs and Apollo economists warn AI payoff will take longer than expected
Two leading economists say AI spending is outrunning its economic returns, with productivity gains still absent outside the largest technology firms, even as corporate leaders remain bullish on the long-term payoff.
Two of the market's most closely watched economists, both at firms financing the artificial intelligence buildout, have publicly questioned whether AI spending will soon pay off in broad economic growth. Within 48 hours, Goldman Sachs chief economist Jan Hatzius and Apollo chief economist Torsten Slok published warnings that the investment is running ahead of its returns. Neither doubts where the technology is heading. Their concern is the pace at which the wider economy will catch up.
Hatzius estimates that the growth boost from the AI buildout will peak in late 2026 or early 2027. The initial spending surge is still rising, but it is losing momentum as AI capital expenditure growth as a share of GDP slows, the boost from last year's tax depreciation incentives fades, and the stock-market wealth effect cools. Goldman still believes in the long-term payoff, but it expects near-term labour-market disruption to weigh on growth before productivity lifts it.
The productivity gains that were supposed to cushion that slowdown are not yet visible in the data. Slok showed that profit margins and revenue per worker are rising among the Magnificent Seven, while remaining flat across the other 493 companies in the S&P 500. Outside the technology sector, he noted, many mainstream companies are beginning to view token spending as a cost centre that weighs on margins rather than a source of growth. He argued that the fate of the AI boom will be decided in the S&P 493 rather than among the Magnificent Seven, and that for now those 493 companies are barely moving.
The view from corporate leadership is considerably more confident. Goldman Sachs chief executive David Solomon said in January that the pace of capital investment would continue. Apollo chief Marc Rowan has described the AI transition as certainly the biggest technology cycle of his career, one in which almost every job will be enhanced or replaced. In his April letter to JPMorgan shareholders, Jamie Dimon predicted that AI adoption would unfold far faster than earlier technological transformations such as electricity or the internet.
That urgency is already prompting questions about labour and society. BlackRock's Larry Fink has asked what happens to everyone else if AI does to white-collar workers what globalisation did to blue-collar workers, and Rowan expects some amount of political upheaval. Executives and politicians are quietly weighing whether the social contract between corporations, government and communities can hold under that strain.
The authors of the research argue that both camps are partly right. Chief executives are right about the destination, while economists are right about the distance. Short-term capital, raised on an impatient timetable, is financing a transition whose payoff runs over several years. Productivity gains will arrive gradually and will take a year or more to show up in corporate earnings and macroeconomic data.
Candour among executives points the same way. In a poll of 200 chief executives at the Yale CEO Summit in June 2025, 40 per cent said AI hype had led to overinvestment, and more than 40 per cent admitted that AI investments had been allocated inefficiently. Filling newly built computing capacity for inference, the stage at which models are put to work for customers, will take longer than the building itself, requiring patient financing to match slower uptake by businesses and households.
Deployment is also proving harder than training. Putting AI into live customer relationships exposes proprietary data and brand reputation to a technology that still fails at an unacceptable rate. The risk grows as applications shift towards agentic systems that carry out multi-step tasks autonomously, where small errors compound at each stage. In a Morning Consult poll conducted for the 2026 Yale CEO Summit, 82 per cent of American adults said human customer-service representatives had become too hard to reach, and 44 per cent said a company's use of AI in customer service made them less likely to stay loyal.
2
