The Bank for International Settlements published a working paper finding that large scale artificial intelligence automation could ultimately constrain economic growth by shifting output from consumption to investment. In the calibrated model, a shrinking consumer market reduces the rents that fund innovation, weakening incentives for further AI development even when the technology has substantial productive potential. Under the paper’s transformative AI scenario, annual growth reaches 5.5% after two decades and continues to rise. Under the empirically plausible demand bottleneck scenario, growth accelerates briefly before falling permanently below the pre-AI rate of 2%, although capital deepening may mask the underlying slowdown in realized output growth for decades. The wide range of outcomes creates substantial uncertainty around the natural rate of interest. The model also finds that research subsidies could offset the growth shortfall, while redistribution between households would not avert the bottleneck because innovation depends on consumption’s share of total output.
Bank for International Settlements working paper identifies consumer demand bottleneck as a plausible constraint on AI growth
A Bank for International Settlements working paper finds that AI automation could shrink consumption’s share of output and weaken the consumer rents that support further innovation. Its calibrated demand bottleneck scenario produces a brief acceleration before growth falls below the pre-AI rate of 2%, while the most optimistic scenario reaches 5.5% after two decades. The divergent paths imply substantial uncertainty for the natural rate of interest.