No Signs of AI in the Productivity Data

The piece explains the difference between total factor productivity (TFP) and labor productivity and why that distinction matters for judging whether artificial intelligence is delivering genuine technological progress. TFP captures how much more output an economy produces from given labor and capital and therefore signals improvements in how inputs are combined; labor productivity (output per hour) can rise simply because workers get more or better equipment or the workforce becomes more skilled. Because AI investments - data centers, chips, model training - represent heavy capital deepening, a near-term rise in output per hour is expected even if firms have not become fundamentally more efficient.

Examining utilization-adjusted TFP alongside output-per-hour growth shows little sign that AI has raised TFP: the TFP series sits slightly below zero with no acceleration since the recent AI capex cycle began, while output per hour runs near 2.5%, above its post-2005 average. That pattern - strong labor productivity with flat TFP - is consistent with capital deepening rather than a broad technology shock. TFP has historically swung between about −3% and +4% over decades and prior general-purpose technologies have taken a decade or more to register in aggregate numbers, so the productivity payoff from AI remains a forecast rather than an observed fact.