Federal Reserve Study Finds AI Investment Soars While Productivity Gains Remain Marginal

Research by the Federal Reserve Bank of St. Louis, examining roughly 490,000 earnings‑call transcripts from 5,198 U.S. firms between 2000 and 2025, shows that AI discussion in productivity commentary rose from virtually zero before ChatGPT to about 15 % by the end of 2025. Around 95 % of AI‑related productivity remarks are forward‑looking, and 95 % of those are optimistic, yet the aggregate productivity data still shows no measurable increase.

The study estimates that AI contributed only about 1.1 % to total productivity growth by the end of 2024. At the same time, investment in AI in the United States is projected to approach $600 billion in 2026—roughly 2 % of U.S. GDP—driven by spending on data centers, hardware imports from Taiwan, and a sustained equity rally, according to Goldman Sachs and Oxford Economics.

Authors Serdar Ozkan and Aakash Kalyani caution that the abundance created by AI may be eroding the value of the output it makes cheaper, potentially offsetting any real gains in the productivity statistics.