On 28 September, Goodman Group withdrew its development application for Project Mars, a 90 megawatt data centre in Sydney's Lane Cove. The project, reported at around $798 million, sat about 20 metres from the nearest house and 160 metres from Lane Cove West Public School. In its letter to the NSW planning authority, Goodman said the policy and regulatory environment for large-scale data centres had "evolved significantly" since it began the approval process, and cited community feedback.
It's one of the clearest examples yet of local opposition stopping an Australian data centre. It won't be the last. The same week, the Greens called for a moratorium on AI data centres, and experts in Tasmania urged rules to hold data centre operators to account.
Why This Is an AI Strategy Story, Not a Property Story
Most organisations never think about where their AI actually runs. It's a line in a cloud contract. But much of the push toward onshore and sovereign AI in Australia, from regulated sectors keeping data in the country to government preferences for local hosting, assumes that onshore capacity will be there when it's needed.
Data centres need land, power, water and community acceptance. Lane Cove shows that the last of those can't be taken for granted, especially near residential areas. If large projects face more scrutiny, onshore capacity may grow more slowly or cost more than many AI roadmaps assume.
Social Licence Is Now Part of the Equation
The phrase that matters here is social licence. Communities are starting to ask what AI infrastructure gives back in return for the power, water and land it uses. Those questions will increasingly reach the organisations that use the capacity, not just the ones that build it. Boards and customers asking about the environmental footprint of an organisation's AI use is no longer far-fetched.
This doesn't mean onshore AI is a bad bet. It means sovereignty plans should be built with capacity, cost and community expectations in mind, rather than assuming unlimited local supply at today's prices.
What to Build Into Your AI Plans
A few practical steps follow. Work out which AI workloads genuinely need onshore hosting for regulatory or contractual reasons, and which only prefer it, because the answer shapes how exposed you are to capacity constraints. Ask your cloud and AI providers about their Australian capacity plans and region commitments, rather than assuming them. Keep architectures flexible enough that workloads can move between regions or providers where the rules allow. And track the energy and resource footprint of significant AI workloads, because someone will eventually ask.
What This Means for Your Organisation
What we see across implementation engagements is that data residency gets decided once, early, and then forgotten until something forces a rethink. Lane Cove is a reminder that the infrastructure behind sovereign AI is physical, local and contested. Organisations that plan for that reality, by separating genuine residency requirements from preferences and keeping their options open, will be better placed than those that assume onshore capacity will simply keep growing to meet demand.
Key Takeaways
- Goodman Group withdrew its 90 megawatt Project Mars data centre in Lane Cove on 28 September, citing an evolving regulatory environment and community feedback.
- The withdrawal is one of the clearest cases of local opposition stopping an Australian data centre, alongside calls for moratoriums and new rules on data centre operators.
- Sovereign and onshore AI strategies often assume local capacity will be available, and growing community scrutiny may slow that capacity or raise its cost.
- Organisations should separate genuine onshore requirements from preferences, confirm provider capacity plans, keep architectures portable and track the resource footprint of major AI workloads.
How Trusenta Can Help
AI Strategy Enterprise builds AI roadmaps that account for onshore capacity, cost and data residency requirements, not just model and use case choices.
Enterprise Architecture maps which AI workloads carry genuine residency requirements and designs them to move between regions or providers where the rules allow.
AI Integration Services builds AI integrations that avoid hard dependencies on a single hosting location or provider.
Conclusion
Lane Cove was a local planning decision, but its effects reach well beyond one Sydney suburb. The capacity behind Australia's sovereign AI ambitions has to be built somewhere, and communities now have a stronger say in where. Organisations that factor that into their AI strategy now will face fewer surprises when capacity, cost or community expectations shift.
