RBA’s Hunter warns AI spending is outpacing its demonstrated productivity gains

AI is helping people finish some tasks faster, but that has not yet translated into clear gains across the economy, Reserve Bank of Australia Assistant Governor Sarah Hunter said on November 9. Speaking in Sydney, she warned that investment is already adding demand, while disappointing returns could expose financial markets to sharp falls.

Faster tasks are not the same as a more productive economy

In her speech to the UBS Australasia Conference, Hunter distinguished between improvements in particular jobs and aggregate productivity: how much an economy produces from its inputs. Lawyers can review documents more quickly, customer service workers can tailor responses faster, and some software businesses are using agents to generate batches of code.

Turning those capabilities into wider gains takes investment and changes to how businesses operate. Hunter described a productivity J-curve: an early period of weak or falling measured productivity while firms experiment and reorganize, followed by improvement as the technology becomes embedded. Managing new operational risks also takes resources.

Her historical example was electricity. The largest gains did not come simply from swapping steam engines for electric motors. They came from redesigning factories and production processes around the new power source. That distinction helps explain why widespread adoption alone need not immediately show up as stronger productivity.

The eventual scale remains highly uncertain. Hunter cited estimates of AI’s contribution to annual productivity growth over the next decade ranging from 0.1% to more than 2%. She also pointed to a possible further benefit: cheaper research, design and experimentation could help businesses develop products and scientists test more hypotheses.

Construction spending arrives before the efficiency dividend

The demand side is less hypothetical. Building data centres and utility networks is already generating spending, alongside purchases of servers and advanced semiconductors. Hunter said this investment supports activity in major Australian trading partners and boosts hardware exports from technology-intensive economies, including Taiwan and South Korea.

Australia does not retain all of that spending as domestic output. Much goes toward imported machinery and equipment. Still, Hunter said the RBA was seeing modest indirect benefits for construction, engineering and infrastructure, particularly energy generation and transmission. AI infrastructure is also becoming an increasingly important source of demand for commodities.

That sequence creates an inflation problem when demand grows faster than available supply. Hunter said the initial price pressure should ease as data-centre capacity and related sectors expand. Broader productivity gains, if they arrive, would let economies produce more with the same inputs, providing another route for that pressure to dissipate.

Interest rates could face a separate, longer-running influence. Hunter said a global rush to invest, potentially alongside lower saving, could increase competition for available funds. That could raise the neutral interest rate—the rate that balances demand for investment with the supply of savings. She presented this as a possible channel, not a precise forecast.

Workers and investors face different adjustment risks

On jobs, Hunter’s assessment was cautious rather than alarmist. Reuters reported that she saw limited evidence of large employment effects so far. But she identified tentative signs of weaker job advertising and employment outcomes in more AI-exposed occupations, alongside stronger growth in less-exposed areas, particularly for younger workers.

The distribution of gains matters for spending, too. Hunter said financial rewards have so far been concentrated among relatively few firms and their owners or shareholders. Workers facing declining demand for their roles could lose income, so the consumption response depends partly on how widely AI’s benefits are shared.

For markets, the risk is that expectations outrun returns. Hunter warned that exposure to the AI investment boom is increasing while compensation for financial risk is low by historical standards. If returns disappoint, equities and corporate debt could face sharp repricing, potentially turning an investment disappointment into a disorderly financial adjustment.