RBA's Hunter Says AI Has Not Yet Lifted Productivity but Is Adding to Inflation

Speaking at the UBS Australasia Conference in Sydney, in a speech titled "What Happens to the Economy as 'Intelligence' Gets Cheaper?", Hunter said "there is limited evidence that AI has materially lifted aggregate productivity." The largest gains from transformative technologies "often take longer to emerge than initially expected", she said. She expects a productivity J-curve, in which measured productivity dips while companies reorganise and then rises, as it did when factories were redesigned around electricity.

The assessment carries weight because it comes from a central bank that has been tightening. The RBA has raised its cash rate five times this year, to 4.85%, an 18-year high, and Reuters reported that a boom in domestic data centre investment is one reason. Hunter said the cash rate sits above the range of central estimates of the neutral rate, and that demand for AI investment may push the neutral rate higher.

Surveys, Job Ads and Young Workers

Hunter cited an RBA survey of firms in which about 35% reported no productivity gains from AI, about 30% minor gains, 23% moderate gains and around 11% significant gains. Published estimates of AI's effect on annual productivity growth over the next decade range from "over 2 per cent" to 0.1 per cent, she said.

On jobs, about three-quarters of firms in the bank's liaison programme reported no change in headcount because of AI. Around 12% reported slower hiring or unfilled roles, about 8% had cut headcount and about 6% had added staff. Growth in SEEK job advertisements since late 2022 has been broadly similar across occupations with low, medium and high exposure to AI, though slightly lower in the most exposed group.

The clearest early signal is among young people. Employment growth for 15- to 24-year-olds has been weaker than for older workers in the occupations most exposed to AI and somewhat stronger in the least exposed, although Hunter said employment in that age group is volatile and the differences are modest. Citing Jobs and Skills Australia, she said about 4% of workers are highly exposed to automation by AI, compared with 31% who are highly exposed to AI augmenting their work, and that "exposure to AI does not necessarily mean job loss."

Data Centres, Prices and Asset Risk

Realised data centre investment in Australia has risen about tenfold from a very low base over three years, Hunter said. Much of that spending goes on imported equipment, which limits its net contribution to GDP. Semiconductor supply constraints are feeding into some consumer electronics prices, which carry small weights in the consumer price index, and data centre construction is adding to pressure on already stretched trades. She expects supply to catch up and said "AI's long-run effect on inflation is likely to be broadly neutral."

She also pointed to risks for financial stability. US AI-related share prices have roughly quadrupled since the end of 2022, and RBA staff estimate that about 5% of Australian households' financial wealth is held in AI-related equities, directly or through superannuation. With market exposure to the investment boom growing and risk premia low by historical standards, she said, equities and corporate debt could face "sharp repricing and possibly disorderly adjustments" if returns on AI investment disappoint. Rapid advances in AI could also worsen cyber threats before defences catch up, she said.

The RBA treats the size, timing and policy implications of AI as a research priority. Hunter said the bank is studying the technology's effects on productivity, labour markets, inflation, financial stability and monetary policy, while weighing near-term price pressure and possible wealth-driven spending against the extra capacity that productivity gains would eventually bring.