AI turned up in three central bank texts in two days, and each gave it a different job

Victoria Shinder29 September 2026, 07:482 min read

Between 28 and 29 September, artificial intelligence appeared in three central bank texts on three continents, and in each it played a different role in the argument.

At the Fed, a source of inflation first. Governor Lisa Cook said AI appears to be adding inflationary pressure and delaying the return to the 2% target. She separated sector-specific price surges in chips, computers and software, which she said monetary policy should not fight, from broader pressure through construction labour, energy and wealth effects. She expects productivity to bring modest disinflation within a few years, but not in time to offset pressure later this year, and she warned that job losses driven by a skills mismatch would leave the Fed with limited tools.

At the ECB, a credit and investment story. At the European Parliament, President Christine Lagarde said firms are set to put about 10% of their total investment into AI in 2026 and that AI-related borrowing is already about a quarter of credit growth to firms. She presented AI as a long-term productivity opportunity whose size depends on adoption, noting that only 7% of euro area firms reported significant use by late 2025, and flagged the risk of a sharp reassessment of highly valued, increasingly indebted AI companies.

At the RBA, an imported price pressure. In its decision to raise the cash rate to 4.60%, the Reserve Bank of Australia listed AI-related demand driving rapid growth in global prices for technology goods among the upside risks now materialising. It also said growth among Australia's trading partners had been stronger than expected because the boost from AI investment outweighed the effects of the Middle East conflict.

What connects them

All three treat AI as something already in the data rather than a forecast. They differ on where it shows up: in prices for the Fed and the RBA, in credit and investment for the ECB. The shared caution is on timing. Productivity gains, the part that would help central banks, are described in every case as later and uncertain, while the demand and price effects are described as present. For anyone reading rate decisions, that asymmetry is the practical point: for now AI is an argument for tighter policy, not looser.