Goldman Sachs analysts dropped a bombshell this week, claiming AI had zero impact on US economic growth in 2025 despite companies pouring $410 billion into AI investments. This stark assessment comes after months of the bank issuing carefully-worded warnings about AI over-investment, marking a dramatic escalation in their skepticism about the technology's near-term economic payoff.
The disconnect reveals two fundamental problems with AI's economic promise. First, much of that $410 billion flows overseas—when US companies buy chips from Taiwan or cloud services from international providers, those dollars boost foreign economies, not domestic GDP. Second, even when AI does make individual workers more productive, those gains remain trapped within company walls rather than cascading through supply chains or generating measurable economic growth.
Goldman isn't alone in this reassessment. Dario Perkins at TS Lombard told the Financial Times there's "no evidence that AI deployment is either boosting productivity or damaging US employment," attributing recent productivity gains to cyclical forces rather than automation. Former New York Fed regulator Brian Peters called AI's "near-term economic payoff" debatable despite extraordinary capabilities and unprecedented capital deployment. Even researchers at the National Bureau of Economic Research identified a "productivity paradox" where perceived gains vastly exceed measured ones.
For developers and AI builders, this matters because it signals a potential reckoning ahead. If the productivity gains we've all been promised don't materialize in economic data, expect tougher questions about ROI and more pressure to demonstrate concrete business value rather than theoretical efficiency improvements.