One line in this week’s Treasury analysis of the dizzying economic implications of AI said it all: “In Australia, most productivity growth at the firm level related to AI will come from adopting and adapting technology developed elsewhere.”
It is one thing for a US-leaning think tank to suggest Australia pursue “sovereignty through managed dependence”, quite another for the government’s lead economic adviser.
But according to Treasury, whether Australia can pull off a second coming for productivity growth is in direct proportion to its ability to keep up with the “pace of innovation at the global frontier”.
Far from making our own luck, Australia’s hopes appear to rest on how hard and fast businesses can adopt — or “diffuse” — AI models developed, predominantly, in the United States.
This may simply reflect experience over the last 30 years, but as assistant technology minister Andrew Charlton has been at pains to point out over recent months, AI is an altogether different beast.
“AI raises the stakes. Because this time, what we risk importing is not simply a manufactured product. We risk becoming dependent on imported intelligence itself,” he told the ANU Crawford School of Public Policy last month.
Dr Charlton, in the same speech, set out a two-pronged path: adopting AI to boost productivity, and building around it to “ensure Australia captures a meaningful share of the value AI creates”, using data centres as the starting point.
“We should not choose between productivity and value capture. We need both. Because GDP tells us how much Australia produces. But our prosperity depends on how much of the resulting income Australians receive,” he said pointedly.
Treasury must have missed the memo, because nowhere in the brief does it talk up the opportunity for Australia to build further up the AI stack, where the real economic value sits, and capture the productivity gains on offer.
Nor does it acknowledge that by chasing productivity gains through overseas AI products, Australia risks cannibalising its own 1,500-and-growing AI companies. It’s almost as if productivity itself demands dependency.
Instead, its narrow focus leads it to data centres, which, although far from certain, could well go on to kickstart the local ecosystem with the right guarantees.
That’s not without cause; Australia does stands to benefit from an expected $150 billion in data centre investment — equal to around 5 per cent of GDP.
But as Treasury also points out, data centre investment “will not contribute equally to domestic economic activity”, with the construction phase and fit-out of imported ICT hardware soaking up most of the capital.
“The import intensity of data centre investment is estimated at around two-thirds,” it said. In short, much of what goes in, goes out, be that to upstream AI economies like the US, Taiwan, South Korea, or those downstream.
It is perhaps why Treasury’s has played it safe, forecasting long-term productivity growth at 1.2 per cent — halfway between the upside and downside scenarios — and warning of uncertainty about the “magnitude and timing” of the impending change.
All of this, of course, is deeply hypothetical. If this year has taught the world anything, we simply don’t — can’t — know what will play out as AI models become smarter, cheaper and harder to predict.
AI adoption across industries where Australia already holds a competitive advantage is, on its own, the most obvious way to turn around the country’s flailing productivity story.
But by treating data centres as the extent of value creation, Australia risks repeating its economic story of extraction, which is inextricably tied to long-standing productivity woes.
Do you know more? Contact James Riley via Email.