The 6.4GW Power Bank: Why Land and Electricity Are the Real AI Infrastructure Moats

The Constraint That Defines the Opportunity

Australia has approximately 6 gigawatts of potential data centre capacity in its development pipeline — roughly four times the operational capacity recorded at the end of 2025. But pipeline announcements are not the same as operational facilities. The critical differentiator between projects that proceed and those that stall is access to electricity.

PwC’s Global Data Centre Outlook identifies power availability as the leading factor shaping where AI infrastructure investment flows, followed by connectivity, security, policy certainty, and community consent. In practical terms, this means the value of a data centre site is determined less by its location and more by its proximity to high-capacity grid connections and reliable, affordable electricity.

Goodman Group’s 6.4-Gigawatt Power Bank

Goodman Group (ASX: GMG) has quietly assembled one of the most valuable power portfolios in Australian real estate. The company has secured a power bank of 6.4 gigawatts — equivalent to more than the entire operational capacity of Australia’s existing data centre fleet. Data centres now account for 73 per cent of Goodman’s development pipeline, which was on track to reach $18 billion by June 2026.

This power bank is not a paper asset. It represents secured grid connections, substation capacity, and in some cases, on-site generation arrangements that allow Goodman to offer tenants faster time-to-power than competitors. In a market where grid connection queues can stretch for years, this speed advantage is a genuine competitive moat.

The Community Consent Challenge

Power is necessary but not sufficient. The withdrawal of Goodman’s $1.2 billion Project Mars data centre on Sydney’s north shore in September 2026 illustrates the growing importance of community consent. The company cited evolving policy and regulation around large AI data centres, noting that the planning environment had changed significantly since the project was first proposed.

This is not an isolated incident. As ABC News reported, the Project Mars withdrawal came amid “growing opposition to AI infrastructure development across the country”. Community concerns typically centre on three issues: water consumption for cooling, electricity price impacts on local grids, and noise and visual amenity of industrial-scale facilities.

For investors, this introduces a new dimension of risk assessment. A company’s social licence to operate is becoming as important as its balance sheet. Projects with strong community engagement strategies, transparent environmental disclosures, and local benefit-sharing arrangements will have a structural advantage over those that do not.

The Regulatory Response

Australian governments at both federal and state levels are grappling with how to manage the data centre boom. The Prime Minister has stated that Australia has “both the desire and capacity to attract investment from major AI companies, including data centres, as long as they met certain requirements”.

These requirements are likely to include renewable energy sourcing, water efficiency standards, and community consultation frameworks. Companies that proactively meet or exceed these standards will be better positioned to navigate the planning process and secure the power and land they need.

Investing in the Power Layer

For investors, the implication is clear: the most defensible AI infrastructure investments may not be the data centre operators themselves, but the companies that control the scarce inputs — power, land, and grid connections. Goodman Group’s power bank is a case study in this dynamic. Its shares have faced scrutiny over the pace of its data centre pivot, but the underlying asset — secured power capacity in supply-constrained markets — remains a valuable strategic resource.

As Australia’s data centre construction cycle progresses from its early stages, the gap between projects with secured power and those without will become increasingly apparent. The winners will be those who understood that in AI infrastructure, electricity is not a utility — it is the product.