The Aussie AI Data Centre debate: profit, power & public trust

While technology issues do occasionally make it into political agendas, rarely does one transcend so many different political realms. The debates around AI cover everything from the environment and climate change to energy policy, infrastructure, town planning and personal ethics. With AI sitting at the nexus of multiple policy areas, it's not surprising that the debate can veer into the apparently chaotic.

Data centre policy

The digital sector is now Australia's second-largest economic contributor behind mining, with the Productivity Commission estimating that AI will add around $116 billion to the economy over the next decade.

Federal Member for Casey Aaron Violi said "For decades, Australia's economic story has been defined by what we could dig out of the ground or grow from our soil. But today, technology presents to us an opportunity to grow a future not defined by our geography, but by our ability to adopt technology and scale it through the digital economy."

Reaping that harvest requires a regulatory and social environment that encourages investment. The Federal Government is planning to introduce legislation compelling data centre operators to completely offset all power consumption with renewables.

While National Cabinet agreed to work towards nationally consistent mandatory standards for large data centres, Queensland and the Northern Territory argued that states need flexibility to set energy rules tailored to their own communities and energy mixes.

While physical infrastructure is important, it's not the only consideration. Australia currently has some 285 data centres in operation with at least another 225 in planning and development. That will, inevitably, lead to increased load on the electricity grid. Coupled with the increased electrification of homes and transport, the grid must add generation and transmission capacity.

In its annually updated ten-year forecast, the Australian Energy Market Operator (AEMO) said 14% of all power consumption will be from data centres in the coming years - up from around 3% today. While it's possible for that amount of generation and supporting infrastructure to be added, there is a challenge.

It takes less time to build a data centre and have it ready to operate than it does to build the energy generation it needs. While the federal government's intentions might be reasonable, it remains to be seen whether they're possible.

Do we need that many data centres?

The underpinning assumption of the data centre build is that large, centralised models will be the norm. But the ability to run smaller, bespoke models using on-prem hardware is rapidly increasing.

Kiraa AI co-founder Errol Brandt said there are about 1.5 billion AI capable devices made by Apple in the wild. And when the number devices made by other manufacturers is added, it brings into question whether the massive investment in large data centres is completely justified. Or at least, whether the apparent 'full speed ahead' mania we're seeing is completely justified.

Professor Toby Walsh from the University of New South Wales likens the speed at which data centres are being built with the massive investments made in fibre optics at the turn of the century.

"If you look at the $2T that were invested in fibre optics at the great fibre rollouts in the late 90s and early 2000s, that proved to be a bubble. The value of the telecom companies in the United States fell in value by about $2B, so it was mostly completely wiped out. And most of that fibre was dark for the first ten years because there was a huge over capacity. Today, that fibre's all lit and it's all being used to screen Netflix movies and the other stuff that we do, so it proved to be a reasonable long-term investment. But most of the money that was put in at the start was transferred from telecoms and people investing in telecoms to the companies now that are benefiting from it."

It's possible that we're seeing an overinvestment today and the companies and investors providing the funding today may not be the ones that receive the benefits.

Walsh posits that many of today's data centre investors are in the real estate business, and not the technology industry. As property speculators, they  procure land and build, only to walk away when it's time for what he sees as inevitable need to inject more capital when technology advances in the future.

This may be behind Microsoft's pivot away from building its own AI data centres. Dr Joseph Sweeney, Advisor on the Future of Work at Australian analyst form Intelligent Business Research Services (IBRS), said the lack of profitability with the major AI players was a factor.

"What we're seeing already is AI infrastructure spending is now having to be offset. Microsoft, about two years ago, saw this coming and pivoted. They didn't stop building their own data centres, but they certainly postponed about $500B of them. And they started doing $5B investments with neoclouds for five-year contracts. Why would you do that? You do that because you realise that your overinvestment isn't going to pay off or maybe you're not sure if it's going to so you're hedging your bets. You're pushing up that financial risk."

What if the AI bubble bursts?

There has been massive investment by the Magnificent Seven which comprises Microsoft, Nvidia, Alphabet, Apple, Meta, Tesla and Amazon. But they are burning through cash faster than they can make it.

The sector is running at a loss. In June 2026, around $2.3 trillion has been wiped off the value of their collective enterprises.

And the market is maturing with a realisation that much of the value proposition offered by large AI providers is easily replicated.

"The problem is AI, or generative AI, or large language models or more specifically vector databases are an algorithm. You can't copyright them. You can't intellectually protect them in any way whatsoever," said Sweeney.

If what we are seeing is indeed a bubble, much like the dot com bubble, then the trillions of dollars invested in the Magnificent Seven and others may evaporate. That's not to say AI will disappear. But we'll see a groundswell of new players and possible some market consolidation.

We're already seeing some volatility in the market capitalisation of large AI companies with values rising and falling on the back of daily announcements.

"I suspect the same will be true of the investments into AI data centres that we will see a vast transfer of money from the people making those investments into the AI companies who eventually end up in the longer term using it," said Walsh.

Public sentiment

Perhaps the biggest unknown facing the AI industry is how public sentiment impacts its uptake and the building of data centres. This is one of the areas the federal government intends to address through its new AI standards.

"The purpose of this standard is to make sure that we attract an outside share of global compute but do it in a way that maintains social license, avoids the backlash that we see in other in other countries and delivers real tangible benefits to Australians," said Andrew Charlton, the Assistant Minister for Science, Technology and the Digital Economy.

Much of the AI debate is focussed on the construction of data centres and whether AI will result in significant unemployment.

When it comes to the impact on employment, one thing almost every expert agrees on is that, at the very least, there will be employment displacement. That is, while some roles will be automated out of existence, new jobs will be created.

Some of this debate echoes the use of industrial robots in automotive manufacturing. However, over time, the car industry employed more people. Routine and repetitive tasks were automated while higher value jobs evolved. It's possible we'll see the same happen with AI.

Data centre construction, however, has a very direct impact today on communities. While they do result in construction jobs – it can be difficult to hire an electrician for domestic work in some areas as many are engaged in data centre projects – once the construction is complete, these buildings require very few people for their operation.

Some of the public anger stems from the nature of the building process, as Walsh explained.

"Most of these data centres are sufficiently large planning applications that they don't require local consultation, and they go up to the state level. So, the public hasn't literally represented it. They're also right to be upset that they're being built in their backyard."

In Walsh's view there's no reason to build a data centre in residential areas. Given the speed data moves across fibre optic, you could build a data centre in a less populated area, such as where a former coal mine operated.

"As an example, to build a data centre in Lithgow, there's an older coal fired power station that they can close down. There are electric cables that go to Lithgow. I did the maths. It would take less than a millisecond for the electron to get from Lithgow to Sydney, which is a latency that most businesses would be happy with. I suspect the community would love the investment, would love to see the jobs, and would love to see something replacing that coal-fired power station that's going to be closed."

These construction decisions are building public distrust in AI as it coalesces with negative sentiment over data centre builds, energy use and the push towards renewables he added.

The debate over AI and its infrastructure highlights a core tension for Australia. There is the drive to harness a technology that could contribute up to $116 billion to the economy while simultaneously navigating the practical realities of energy supply, regulatory flexibility, and community impact.

The federal push for renewable‑offset data centres, paired with the National Cabinet's aim for mandatory standards, demonstrates an intent to balance growth with sustainability. Yet, the speed at which data centres can be erected contrasts sharply with the slower pace of adding grid capacity, raising legitimate concerns about whether the current investment trajectory is proportionate to the actual demand for large, centralised models.

Public confidence will ultimately shape the sector's trajectory. If data centre development is perceived as a top‑down process that bypasses local consultation and places environmental and economic burdens on small communities, the social licence to operate will erode.

A more measured approach that leverages smaller, bespoke models, situating facilities in low‑population or repurposed industrial sites, and ensuring clear benefits to the communities that host them could mitigate backlash while still delivering the economic upside.

The future of AI infrastructure in Australia hinges on aligning regulatory ambition, market realities, and public acceptance in a way that recognises both the opportunities and the responsibilities the technology entails.