US AI Economy Grows 2600% Per Year but Invisible in GDP Statistics

Economists from the University of Virginia, Anthropic, and the Bank of Canada have published a study showing that the AI economy in the United States is growing at an unprecedented rate: nominal computing expenditures rose from $37 billion in 2023 to $90 billion in 2024 and $219 billion in 2025. US computing capacity is increasing by more than 200% per year, and qualitatively adjusted AI output grew by approximately 2290% in 2024 and 2271% in 2025. However, this growth is barely reflected in standard GDP statistics because prices per unit of AI quality are falling almost as fast as output rises. The authors note that AI is the first technology capable of replacing human labor at scale and propose three solutions: develop "satellite accounts for AI" for statistical agencies, improve data collection on primary sources (e.g., the distribution of computing resources between training and inference), and include AI production capacity indicators in medium-term economic forecasts. Without this, governments risk underestimating the likelihood of a shock to the tax base and failing to develop responsive measures, such as reforming the tax system or creating sovereign wealth funds.