The AI gold rush has produced plenty of shovels but precious little gold — at least for the companies buying them. That is the blunt conclusion from Apollo Global Management's economics team, which argues that outside the handful of firms building and selling AI infrastructure, American corporations have seen "zero financial payoff" from their artificial intelligence investments.
The assessment lands at a particularly awkward moment. Businesses have poured hundreds of billions into AI tools, consultants, and internal initiatives since ChatGPT's late-2022 debut. Nvidia has become one of the world's most valuable companies on the back of that spending. Yet Apollo's analysis suggests the buyers, rather than the sellers, are still waiting for returns.
The productivity paradox, redux
Economists have a name for this phenomenon: the productivity paradox. In the 1980s, Nobel laureate Robert Solow quipped that computers were visible everywhere except in the productivity statistics. Four decades later, AI may be following the same script. Corporate earnings calls overflow with AI mentions, but aggregate productivity gains remain stubbornly modest.
Apollo's argument is not that AI lacks transformative potential — it is that the benefits have concentrated almost entirely among a small cohort of technology vendors. Microsoft, Google, Amazon, and Nvidia have captured the upside. Everyone else has captured the expense reports.
Why the lag persists
Several factors explain the disconnect. Implementation costs have proven higher than anticipated. Many firms purchased AI tools without clear use cases, chasing competitive anxiety rather than concrete efficiency gains. Integration with legacy systems remains painful. And the most valuable applications — those that genuinely automate high-cost labor — often require organizational changes that move slower than software updates.
There is also a measurement problem. Productivity improvements from better customer service chatbots or faster document summarization may be real but diffuse, showing up as marginal time savings rather than headcount reductions or revenue jumps. CFOs struggle to quantify "my employees waste fewer hours" on a balance sheet.
The bull case survives, barely
AI optimists counter that transformative technologies always exhibit adoption lags. Electricity took decades to reshape manufacturing after its invention. The internet's productivity dividend arrived years after the dot-com crash. By this logic, Apollo's snapshot captures the trough of disillusionment, not the final verdict.
That may prove correct. But the argument offers cold comfort to executives who justified AI budgets with near-term ROI projections now quietly shelved. The gap between boardroom expectations and measurable outcomes has widened into a credibility problem.
Our take
Apollo's analysis is less a condemnation of AI than an indictment of corporate hype cycles. Businesses bought what they were sold — the promise of effortless transformation — without doing the harder work of identifying where AI actually fits their operations. The technology will eventually deliver. The question is whether patience outlasts the next budget review.
