RBA outlines AI’s productivity potential and investment costs

Hunter said AI could lift productivity and transform workplaces, but the RBA said there was limited evidence that AI had materially increased aggregate productivity so far.[2][1] The RBA described an initial period in which measured productivity may be weak or fall as firms devote time and resources to investment.[1] Lawyers reported that they could review and summarise large volumes of documents more quickly.[1] Hunter said estimates of AI’s effect on annual productivity growth over the next decade range from 0.1 per cent to more than 2 per cent.[1] Those estimates are a wide range, not a settled forecast.

RBA research estimated that Australian data-centre investment had grown around tenfold over the previous three years.[3] The RBA expected a further $115 billion in real investment between mid-2026 and the end of 2028, excluding supporting infrastructure.[3] It estimated that this forecast investment would equal almost 2 per cent of GDP each year in gross terms and assumed around 70 per cent would be imported.[3] Separately, the RBA said data-centre construction alone could equal 2 per cent of GDP for the next two years.[2] The scale makes investment assumptions worth tracking alongside expected productivity benefits.

Hunter said there was little evidence AI had affected overall employment, while noting signs of lower hiring among younger workers in highly AI-exposed occupations.[2] RBA liaison contacts said AI had not affected most firms’ headcounts and was not expected to do so over the next year.[2] The RBA also said data-centre construction was competing with the construction sector for trade workers.[2] For workforce planning, distinguish these specific signals from evidence of broad job losses: the research describes limited, uneven effects rather than a general employment shift.

The RBA said strong electricity demand from data centres could put upward pressure on wholesale electricity prices for a time; AEMO estimated data centres accounted for around 3 per cent of NEM operational electricity consumption.[3] RBA research estimated that 5.3 per cent of Australian household financial wealth was invested in AI-related companies, directly or through superannuation, up from 2 per cent in late 2022.[2] Hunter said returns had so far flowed to a small number of businesses and their equity holders.[2] Hunter also warned disappointing investment returns could expose some asset prices to sharp repricing and possibly disorderly adjustments.[1]