Artificial intelligence is emerging as a two-sided challenge for Australia’s electricity system, simultaneously giving cyber attackers more powerful tools and driving a rapid expansion of energy-hungry data centres.
The Australian Energy Market Operator (AEMO) has warned that advances in generative AI are lowering the technical barriers to sophisticated cyberattacks against power stations, electricity networks and other operational infrastructure.
AI tools can help attackers automate reconnaissance, identify vulnerabilities and coordinate multi-stage intrusions that previously required highly specialised technical skills.
That raises the prospect of a growing number of actors gaining the ability to target critical infrastructure, including energy, communications, financial services, healthcare and government systems.
Australia’s energy network is particularly exposed because it is becoming more decentralised and digitally connected as renewable energy, batteries, smart meters and internet-enabled equipment are added to the system.
AEMO describes energy infrastructure as a high-value target for cybercriminals, state-sponsored groups and hacktivists, with a successful attack potentially affecting electricity supply and broader economic activity.
Data centre demand adds pressure
At the same time, the AI systems creating new security risks are placing unprecedented pressure on the physical grid.
The expansion of cloud computing and generative AI is driving investment in large-scale data centres, which require substantial amounts of continuous electricity to operate servers and cooling systems.
Australian data centre electricity consumption has been forecast to increase from about 4 terawatt-hours in 2025 to more than 21 terawatt-hours by 2035.
Unlike some industrial facilities, hyperscale data centres can require power around the clock and may be concentrated in locations where transmission capacity is already constrained.
AEMO has also identified the potential for large data centres to create grid stability risks. Sudden voltage disturbances could cause several facilities to disconnect simultaneously, removing a large block of demand and creating a shock across the electricity network.
The issue is not simply whether Australia can produce enough electricity. New transmission lines, substations, storage and firming capacity will also be needed to deliver reliable power to rapidly growing digital precincts.
Tougher rules on the way
The Federal Government is moving toward rules requiring major data centre developers to fund their grid connections and secure adequate electricity supplies rather than passing infrastructure costs on to households.
Operators may also be required to reduce electricity use during periods of grid stress and improve their management of water, land and community impacts.
These measures reflect an emerging principle: companies profiting from the AI infrastructure boom will be expected to help build and stabilise the energy system supporting it.
There may also be a technological solution. Research suggests AI computing workloads can be shifted between locations or delayed during peak demand, allowing data centres to act as flexible grid participants rather than fixed, always-on consumers.
Tests involving GPU clusters have demonstrated rapid load reductions, sustained power curtailment and the ability to move less urgent computing tasks away from periods of grid stress.
The AI boom therefore presents Australia with a difficult balancing act. Data centres could attract billions of dollars in investment and support a growing digital economy, but without stronger cybersecurity, flexible power use and accelerated grid investment, the same boom could expose weaknesses in the infrastructure on which it depends.