Data centres are one of the biggest drivers of new energy demand.
AI will drive new demand for electricity, and we have to be clear-eyed about how this demand could overwhelm the grid, and push up household power bills.
Those risks are real. We have already seen overseas what can happen when data centre demand grows faster than electricity supply.
The government’s triple lock on data centres – bring your own supply, be demand flexible, and cover your grid connection costs – highlights potential benefits we should not overlook.
Planned properly, data centres could strengthen our electricity system — making them an asset to our energy system, not a strain.
The first is new generation.
Data centres are huge, long-term electricity customers. That means they can sign the contracts that help get new renewable energy projects financed.
We are already seeing this in Australia. Amazon has contracted almost one gigawatt of renewable capacity here. In April it announced nine new agreements adding 430 megawatts of wind, solar and batteries in NSW and Victoria.
Instead of competing with households for the electricity we already have, data centres can help finance more of it.
The second benefit is flexibility.
Electricity becomes particularly expensive when everybody wants it at once. We have to build parts of our electricity system to cope with a relatively small number of extreme peaks – like a scorching summer afternoon when millions of air conditioners are running.
But some computing can move.
An AI training job may make little difference to anyone if it finishes at 8pm rather than 6pm. Shifting that demand out of peak periods can reduce pressure on the grid and reduce the amount of expensive generation and network capacity needed just for those few critical hours.
In South Australia, Firmus and NVIDIA have agreed to reduce energy demand for up to 220 hours a year when wholesale electricity prices exceed agreed thresholds.
But the third opportunity – which is significant, and possibly the least understood – is spreading the fixed costs of the energy network.
When you pay your electricity bill, you are not just paying for the electricity you use. You are also paying for the poles, wires, substations and transformers that deliver it.
Network charges are the biggest cost line on an electricity bill – accounting for around 40 per cent of a typical electricity bill.
Data centres will need some new infrastructure, but they will also share costs of a lot of infrastructure we need anyway.
The poles are already standing. The wires are already strung. The substations are already built. Consumers have to pay for them whether they are carrying a little electricity or a lot.
Think of Uber pool. If a trip costs $60 and you are the only passenger, you bear the whole cost. Add another passenger travelling in the same direction and you can split it. Add another and each person’s share falls further.
Electricity networks can in some situations work in much the same way. If a network has spare capacity and a large data centre connects, it starts contributing towards infrastructure households are already paying for. More electricity flows through the same poles, wires and substations.
The fixed cost is spread — or fractionalised — across a much larger amount of electricity.
And we’re also making sure that if data centres create or accelerate the need for more network builds, they have to fully cover the cost.
That can put downward pressure on the amount each existing consumer needs to pay.
There is evidence this can happen. A 2026 US study by Asa Watten, John Bistline, and Geoffrey Blanford found that data centres modestly reduced average retail electricity prices between 2015 and 2024. The researchers found economies of scale across generation, transmission and distribution: as durable electricity demand grew, fixed costs were spread more widely. But America also provides the warning.
In the PJM electricity market, projected data centre demand surged while electricity supply was constrained. PJM’s independent market monitor estimated data centres accounted for 63 per cent of the increase in its 2025-26 capacity auction prices – around $US9.3bn ultimately recovered from customers.
That contrast is the point.
Data centres will not automatically increase electricity prices. Nor do they automatically reduce them.
Australia needs rules that are deliberately designed to capture the upside and prevent the downside. Rules that we are committed to establishing now – not once it’s too late.
The Prime Minister’s National AI Standards will set a framework to harness these benefits, without sacrificing the needs and values of Australians across the country.
Data centres should pay for new infrastructure their demand causes, underwrite new generation and reduce demand when the system is under stress. They should seek to locate where the grid has spare capacity.
The goal should not simply be to protect the electricity system from data centres.
It should be to use this enormous new source of demand to build more supply, use our existing grid more efficiently, and spread its fixed costs across more customers.
Get that right, and data centres could play a more positive role in the energy system than many people might expect.