Bank of America Report Compares AI Productivity Impact to Electricity and Internet

Bank of America’s global economics team released a report on Thursday stating that artificial intelligence differs from earlier general-purpose technologies because it can accelerate the invention process itself. The bank wrote that electricity automated physical processes and the internet moved data faster, yet neither increased the speed of inventing new things.

The report placed current measured productivity gains from AI at 0.1 percent per year, a figure the bank described as small relative to global growth of 3.5 percent. Bank of America calculated that AI can currently transform about 20 percent of workplace tasks, but only 23 percent of those tasks are economically viable to automate given today’s prices, producing an estimated 0.66 percent theoretical ceiling before frictions reduce the realized gain.

Bank of America cited work by economist Philippe Aghion and co-authors published in 2024 that modeled cumulative productivity gains over the next decade at roughly ten times current estimates. The bank argued that falling inference costs, which it said halve every three months, will expand the share of tasks that become both technically and economically feasible to automate.

The report added that capital deepening and non-linear compounding effects could further magnify aggregate productivity gains. Bank of America projected that these dynamics could add up to one percentage point to global growth over ten years.

and Measured Forecasts Economist Tyler

Cowen stated at the Sana AI Summit on Thursday that AI would contribute between 0.5 and 1 percentage point to U.S. growth, citing institutional constraints in sectors that represent 40 to 50 percent of GDP. Panmure Liberum strategist Joachim Klement separately argued eight days earlier that current hyperscaler capital expenditure plans exceed levels needed to justify projected returns.