Australia Pushes Back on Datacentre Fears as Officials Call Backlash a US Import
Australian politicians and datacentre executives say the country's growing backlash against AI datacentres is an inauthentic import from the United States, with one minister declaring on 26 September: "We are not the United States."

The argument, reported by The Guardian on 26 September, rests on a scale comparison. Australia has 252 datacentres with 1.4 gigawatts of capacity, according to Belinda Dennett, head of the peak lobby group Data Centres Australia. The US has 5,400 datacentres and 53.7 gigawatts. The Guardian notes measurements of US datacentre counts vary.
What Dennett omitted, the Guardian points out, is that Australia has just under one datacentre per 100,000 people against 1.5 per 100,000 in the US. Australia's population is 28 million; the US figure is 342.7 million.
Regulation and gas
Canberra is moving to set national rules. The Australian government plans to override states and ban gas being used to power datacentres. New rules would require operators to bring as much new renewable energy into the system as the centre would otherwise take. State governments, keen for economic growth, have broadly welcomed the investment, in some cases overriding local council approval processes.
Victoria's then-minister for economic growth, Steve Dimopoulos, said in July that his state was second only to New South Wales on datacentres and was outstripping Australia's most populous state in growth for new centres. He blamed the resentment on US social media. "We're seeing some US states move to ban new large datacentres. And those campaigns and mega-developments in the US shape public opinion in Australia," he said. "Because when it's on your phone, you think it's right here, or you think it might just come here."
Dennett, speaking in March, was blunter: "Australia's datacentre debate is being fought on America's facts." She said the industry is depicted as "energy vampires, water hogs and job killers," and that most such headlines originate in the US, "where the industry is operating at a completely different scale, under completely different market rules."
Where the two sides agree
Not everyone buys the import framing. Rob Nicholls, professor and senior research associate in digital policy at the University of Sydney, told the Guardian the US social media input is "very vocal and rationally based." His summary of the argument: "It's going to increase the cost of electricity and it's going to increase the cost of water and I don't get a benefit from it." He called that a rational position.
The construction pipeline is the pressure point. Several large datacentres of 1 gigawatt or more are in planning. TikTok and Instagram videos have highlighted a 2.4 gigawatt centre planned for outer Melbourne and a 2 gigawatt gas-powered centre in the Northern Territory. Nicholls said the NT project is equivalent to 2m households' annual electricity consumption. The NT's population is just over 265,000.
The Australian debate has produced calls for a national moratorium, community protests, three government inquiries and state frameworks, according to the Guardian. That is a lot of political machinery for a market whose total installed capacity is roughly 2.6% of the US figure Dennett cited.
The chip supply behind the buildout
The accelerators that fill these halls are moving through supply chains that Washington is still trying to control. On 23 September, Tom's Hardware reported that UK neocloud Nscale built its business on a deal with Spring, a Singaporean ByteDance subsidiary. Spring accounted for $24 million of Nscale's $33 million revenue in 2025.
That contract gave Spring access to 2,304 Nvidia B200 chips at Nscale's Glomfjord, Norway, datacentre. Tom's Hardware, citing the Financial Times, said the arrangement was legal and exploited loopholes in US export controls. The detail surfaced only because Nscale filed S-1 documents for a US IPO. ByteDance and Spring were not named in the S-1 itself but in a supporting document, a $105 million loan agreement with Macquarie. Macquarie required Nscale to monitor Spring's chip use and report "compute anomalies or suspicious configurations" that might breach export controls.
Nvidia's exposure grew after that. Tom's Hardware reports Jensen Huang committed more than $660 million to Nscale. Nvidia's commitments later rose above $2 billion, plus an $860 million lease guarantee for a Texas facility. Nscale has since signed a $44 billion deal with Microsoft and a $45 billion contract with Anthropic, and says Spring will fall below 20% of revenue.
Design tools shift too
The silicon side is changing as well. Semiconductor Engineering's week in review, published 25 September, reports Cadence released an agent for spec-to-RTL generation and early PPA optimization inside its ChipStack AI Super Agent. It claims about 24% lower area and 18% lower power versus pure foundation model code generation. The same roundup notes Alibaba's Zhenwu V900 AI accelerator, with 216GB of memory and 1,200GB/s inter-chip bandwidth, claimed at 3x the performance of its M890 predecessor, with mass production planned for Q1 2027.
Also in that roundup: CXMT said its fifth-gen G5 DRAM entered mass production with a new LPDDR5X line, claiming an 11.95nm active-area half-pitch and at least 50% more dies per wafer than the previous generation, without disclosing actual yields. Samsung reportedly plans to at least double HBM4-family output in 2027, according to Seoul Economic Daily. And imec researchers modeled a 3D DRAM-on-GPU architecture that cut peak temperature from 121.7C to 103.4C versus an HBM-on-GPU baseline.
On the same day, Semiconductor Engineering flagged a paper from TU Munich, the University of Modena and Reggio Emilia and Applied Materials titled "System-Technology Co-Evaluation of A7 CFET and A10 NSFET Technologies from Cell Parasitics to Chip Reliability." The authors describe a thermal- and aging-aware flow linking calibrated device models, standard-cell generation, GDS-to-TCAD conversion, full RTL-to-GDS implementation for an AI accelerator, multiphysics thermal analysis and bias temperature instability aging evaluation.
Cadence's Matt Graham told Semiconductor Engineering on 24 September that super agents represent "a new category of tools," with complexity and value potentially as big as EDA tools themselves. Synopsys' Prith Banerjee compared multi-agent workflows to human teams, where a power expert does clock gating, area grows, and an area expert steps in. Movellus CEO Mo Faisal put the boundary bluntly: "Stay in your lane."
None of that settles the Australian argument. The country's buildout is smaller in absolute terms, and its regulators are writing rules before the gigawatt projects land. The residents near the proposed sites are arguing about local power and water, whatever the origin of the talking points.
Sources
5- 01Australian backlash to datacentres is faux import from US, officials say: 'We are not the United States'EN
- 02China's ByteDance gained access to over 2k Nvidia B200 chips through NorwayEN
- 03Chip Industry Week In ReviewEN
- 04Comparing A7 CFET and A10 Nanosheet FETs From Parasitics to Chip Reliability (TUM, UNIMORE, Applied Materials)EN
- 05When AI Agents Cross Chip Design SilosEN
All figures and quotations in this text come from the sources listed below.
Content prepared by the editorial team with AI assistance.
Comments
0- No comments yet — be the first.