An 800 MW data centre proposal in Australia has exposed the collision course between hyperscale computing demand and the physical limits of grid decarbonisation, requiring four dedicated substations, two new transmission lines, and enough on-site diesel to run for four days without grid power – a single facility that would rank among the nation’s largest electricity users and test whether renewable-heavy grids can deliver the firm, continuous capacity that AI workloads demand.
The Infrastructure Reality Behind an 800 MW Load
The project, proposed by hyperscale operator AirTrunk for a site in Sydney’s western suburb of Horsley Park, would draw roughly the same continuous power as a mid-sized aluminium smelter or a combined-cycle gas turbine plant. Four zone substations – each typically serving thousands of residential customers – would be built solely to feed this one campus. Two new 132 kV transmission lines would connect it to the broader National Electricity Market (NEM). The scale is not incremental; it is city-scale infrastructure dedicated to a single private tenant.
What makes the proposal notable is not just the peak demand figure but the reliability architecture. The development application lodged with the NSW Department of Planning shows 48 diesel generators rated at 2.5 MW each – 120 MW of backup generation – backed by fuel storage sufficient for 96 hours of full-load operation. That is roughly 28.8 million litres of diesel on site, assuming typical generator consumption rates of 0.25 litres per kWh. Batteries are included in the plan, but their role is framed as bridging the seconds-to-minutes gap until generators start, not as a substitute for the diesel plant.
This configuration reflects a hard operational reality: hyperscale tenants – the cloud providers and AI model trainers who ultimately lease this capacity – typically contract for 99.99% or 99.999% uptime with financial penalties for any interruption. The NEM, even with its growing renewable share, cannot yet offer that level of firmness without transmission augmentation and long-duration storage that does not exist at this scale. The diesel plant is not a transitional choice; it is the only proven technology that can deliver four days of continuous, independent power at 120 MW within the project’s timeline.
Grid Integration and the Firm Capacity Gap
The Horsley Park proposal illustrates a structural mismatch that will repeat across every major grid hosting large-scale data centre growth. Australia’s NEM is undergoing a rapid transition: coal capacity is retiring – Liddell (2 GW) closed in 2023, Eraring (2.9 GW) is scheduled for 2025 – while wind and solar now regularly supply 40-50% of daytime energy. But the grid’s firm capacity – the power available during wind droughts, evening peaks, or transmission outages – has not grown at the same pace. The Australian Energy Market Operator’s 2024 Integrated System Plan projects a need for roughly 45 GW of new dispatchable capacity by 2040, much of it long-duration storage or gas-fired peaking plant that has yet to be financed.
An 800 MW baseload addition changes the calculus for network planners. Transgrid, the NSW transmission network service provider, must now justify the cost of two new 132 kV lines and associated switchgear – likely several hundred million dollars – against a single customer’s load. Under the current regulatory framework, the connection charges are negotiated between the customer and the network, with the Australian Energy Regulator providing oversight. But the precedent matters: if this connection proceeds on terms that socialise any portion of the augmentation cost, it sets a benchmark for the next ten hyperscale campuses currently in various stages of planning across Sydney, Melbourne, and Brisbane.
That points to a broader trend: data centres are becoming the primary driver of transmission investment in the NEM, displacing renewable energy zones as the anchor loads that justify new poles and wires. In the 2023-24 financial year, data centre connection enquiries to Transgrid and Powerlink (Queensland) exceeded 10 GW in aggregate – more than the total coal capacity retired in the past five years. Not all will proceed, but the pipeline is real, and each project of this scale forces a re-evaluation of whether the grid’s decarbonisation pathway can accommodate large, inflexible, 24/7 loads without new firm generation.
Implications for Renewable Integration and Storage Economics
The inclusion of batteries in the AirTrunk plan – described in the development application as “grid-scale” but without a disclosed capacity figure – signals a secondary value stream: frequency control and wholesale price arbitrage. In the NEM, batteries earn the bulk of their revenue from Frequency Control Ancillary Services (FCAS) markets, not energy arbitrage. A data centre with on-site storage can bid its batteries into FCAS when not needed for backup, creating a revenue offset. But the economics only work if the battery is large enough to be a meaningful market participant – typically 50 MW / 100 MWh or more – and if the site’s load profile leaves sufficient headroom.
If this trend holds, we will see more data centres deploying batteries not for resilience but as merchant assets, effectively turning their backup infrastructure into virtual power plants. That creates a tension: the same battery cannot simultaneously be fully charged for a four-day diesel-outage scenario and actively cycling for FCAS revenue. Operators will need to define reserve margins explicitly, and regulators may need to clarify whether on-site storage at large loads counts toward system reliability standards or remains a private asset.
By comparison, the largest grid-scale battery currently operating in Australia is the 300 MW / 450 MWh Victorian Big Battery. The AirTrunk battery, even at a speculative 100-200 MWh, would be a significant distributed asset. Aggregated across ten similar campuses, on-site storage could reach 1-2 GWh – a non-trivial share of the NEM’s total installed battery capacity, which stood at roughly 2.5 GWh at the end of 2024.
Who This Affects
- Transmission network planners: Must model 800 MW+ block loads as firm, non-shiftable demand in reliability assessments, not as flexible or interruptible load – this raises the firm capacity requirement for the NEM by roughly 1% per facility.
- Long-duration storage developers: The four-day diesel benchmark defines the performance target: any storage technology claiming to replace diesel must deliver 120 MW for 96 hours (11.5 GWh) at competitive cost – far beyond current lithium-ion economics and into the realm of flow batteries, compressed air, or thermal storage.
- Gas turbine and peaking plant operators: Each hyperscale campus that proceeds with diesel backup represents a lost opportunity for gas-fired firming capacity; if 10 GW of data centre load materialises with on-site diesel, that is 10 GW of potential gas plant revenue displaced.
- Policy analysts and regulators: The NSW planning approval process will test whether state governments can impose renewable procurement or emissions intensity conditions on private infrastructure of this scale – the outcome will shape the template for every subsequent project.
What to Watch Next
- Transgrid’s connection offer and negotiated terms: The scope of shared vs. dedicated assets, and whether any augmentation cost is recovered from other network users, will signal the regulatory precedent.
- AirTrunk’s renewable PPA strategy: Whether the operator contracts 800 MW of firmed renewable supply (wind/solar + storage) or relies on grid average emissions intensity will determine the facility’s actual carbon footprint.
- NSW Department of Planning conditions of consent: Any requirements for staged diesel reduction, hydrogen-ready generators, or minimum on-site renewable generation will set the policy bar.
- Competing hyperscale proposals in Western Sydney and Melbourne: At least three other 500 MW+ campuses are in early planning; their grid connection applications will reveal whether the AirTrunk template becomes the industry standard.
Bottom line: The Horsley Park project is not an outlier – it is the leading edge of a structural shift where data centres become the primary anchor loads shaping transmission investment, firm capacity requirements, and the practical limits of grid decarbonisation in Australia. The four-day diesel tank is the honest measure of how far the system still has to go.
Read the full report at RenewEconomy
Original source: RenewEconomy (Australian clean energy news)
Note: facts and figures attributed above to RenewEconomy (Australian clean energy news) reflect that outlet's original reporting. Broader context, cross-sector connections, and forward-looking scenarios reflect independent analysis by our editorial team.
About this article: Drafted by Energy Ai with AI-assisted research and writing based on public reporting, then reviewed under our editorial process before publication.
Leave a Reply