Australia's Treasury has just given Treasurer Jim Chalmers something more useful than another AI hype speech: a sober, hedged look at what the technology will actually do to growth and jobs. The report is part of the groundwork for September's Intergenerational Report, where Chalmers has said AI will be front and center of the government's economic story for the next 40 years.
The headline number is a long-term productivity growth assumption of 1.2 percent a year. Treasury is clear that AI will help get there, but cannot do it alone. Australia's productivity actually fell last year and has been mostly flat for a decade, so this is a real gap to close, not a small tweak.
Treasury lays out a range instead of a single confident number. If AI adoption goes well globally and spreads to Australia, growth could reach 1.5 to 2 percent. If it goes badly, growth stays stuck around 0.5 to 0.8 percent, close to where it sits today.
That is a wide gap, and it shows how much of this is still a guess dressed up as economics. Other Australian forecasters land in very different places, which is worth knowing before anyone builds a business plan around a single figure. The Productivity Commission has floated an AI boost worth more than 116 billion Australian dollars over a decade.
Commonwealth Bank economists, working from similar data, put the likely annual gain at just 0.4 percentage points, well below what Treasury and the Productivity Commission are hoping for. When experts disagree by this much, treat any specific AI growth forecast as a rough guess, not a plan.
The jobs story matters more for most readers than the growth math. Past waves of automation mostly hit routine physical work. This one is aimed at higher skilled, non routine office jobs, the kind done by accountants, analysts, and administrators.
Treasury expects some of these jobs to be boosted by AI and others to be replaced, and it admits the effect will land unevenly by industry and location. Finance, insurance, telecommunications, and professional services are already adopting AI fastest, which lines up with where white collar automation risk is highest.
Two out of three Australian businesses report some AI use, but fewer than one in ten call it significant, so most companies are still at the experimenting stage rather than the rebuilding stage. That gap between talking about AI and actually running on it is the real story right now, more than any productivity forecast.
There is also a less obvious thread about interest rates. Big AI companies worldwide have been borrowing at record levels, more than 121 billion US dollars in bonds during 2025 alone, to pay for data centres. Treasury flags that this kind of borrowing could push up the neutral interest rate worldwide, the rate at which central bank policy stops helping or hurting growth.
That would make borrowing more expensive for everyone, not just tech firms. Locally, Australia already has 162 data centres running and close to 130 more proposed, and the government has now agreed on national rules forcing new centres to bring their own renewable power. Queensland and the Northern Territory got some flexibility to use surplus fossil fuel power instead.
The honest takeaway is that Australia's government is betting on AI to fix a productivity problem that predates AI by a decade, while admitting it might not work, and while most businesses have barely started using the technology seriously.