Data centres consumed around 4 TWh of electricity across the National Electricity Market in 2024-25, close to 2% of grid-supplied demand. AEMO expects that to nearly triple to about 12 TWh by 2030, roughly 6% of the NEM. For a commercial energy buyer, the headline number matters less than where it lands. A large new block of demand is arriving in the two states where most commercial portfolios sit, on a timeline that overlaps the contracts being signed right now.
This guide covers what AEMO actually forecasts, why the connection queue overstates what will be built, what the price modelling does and does not say, and the two decisions a procurement team genuinely controls: when you contract, and for how long.
How Much Electricity Do Data Centres Use in Australia?
In 2024-25, data centres connected to the NEM used approximately 4 TWh, a little over 2% of grid-supplied electricity. AEMO forecasts that figure reaching nearly 12 TWh by 2030, around 6% of grid-supplied electricity in the NEM, and about 34 TWh by 2049-50, roughly 12%. That is a growth rate near 25% a year for the rest of this decade.
Two things are worth noting about those numbers. The first is that this is the first time the market operator has separated data centres from other industrial loads in its official planning inputs, which tells you the category has become material enough to model on its own. The second is that 6% of a market is a meaningful shift in the supply and demand balance, but it is not the kind of shock that rewrites a bill overnight. The effect shows up gradually, in the forward curve, over exactly the period a three or five year contract covers.
Why Does the Connection Queue Overstate What Is Coming?
At the end of the March 2026 quarter, 11 large-scale data centre projects above 5 MW, representing 5.4 GW of maximum demand, were progressing through the transmission connection process. Around 60% of that capacity sits in New South Wales and 40% in Victoria, and most projects are still at an early stage. AEMO's experience is that a large data centre connection takes roughly two years from application to energisation.
The wider pipeline is far larger and far less reliable. Network service providers reported around 44 GW of data centre connection requests during AEMO's 2025 inputs and scenarios process. Analysis of that pipeline estimates that roughly six in every seven megawatts is phantom demand: the same project lodging enquiries at multiple sites and networks, counted several times over, with only one of them ever getting built.
Treat the queue as a measure of developer ambition rather than a forecast of load. Anyone quoting the raw connection pipeline at you as evidence that prices must rise is quoting a number that AEMO itself discounts heavily. AEMO has also updated its forecasting methodology specifically to handle rapidly growing large loads, which is a fair signal that the market operator regards the uncertainty as unresolved.
What Does This Do to the Price You Are Contracting Against?
Start with where the market actually is. NEM-average wholesale spot prices averaged $74/MWh in the June 2026 quarter, down 47% year on year and the lowest second-quarter average since 2020, with renewables supplying a record 42.1% of generation. Softer wholesale conditions are the backdrop against which retailers are quoting you today. Large-load growth does not change that spot outcome. It changes the balance across the later years of a contract you sign now.
On the price effect itself, the most cited Australian figure comes from the Climate Council's Clouded future report, published 3 June 2026. It models one specific scenario: if new data centre demand is met mainly with gas generation rather than additional renewables and storage, wholesale prices could be 26% higher in New South Wales and 23% higher in Victoria by 2035, with grid emissions around 14% higher than a renewables-led pathway.
Read that carefully before you build a business case on it. Those are scenario projections published by an advocacy organisation, not an AEMO forecast, and the two should not be blended into a single implied consensus. The conditional matters as much as the percentage. What the market operator and the modellers do agree on is the mechanism: the price outcome depends on whether new generation and storage arrive alongside the new load, or behind it.
Which Rules Are Changing Around Large Loads?
Connection standards. The AEMC published a draft determination on 12 March 2026 proposing new access standards for large inverter-based loads, with a 30 MW threshold written into the National Electricity Rules. The concern is technical rather than commercial: if many large facilities disconnect simultaneously during a voltage disturbance, the risk of cascading instability rises. A final determination was scheduled for 29 October 2026.
Who pays for new generation. At their July 2026 meeting, energy ministers other than those from Queensland and the Northern Territory agreed to pursue national rules requiring data centres to offset their electricity demand by investing in additional renewable generation, sited near the facility unless local authorities opt out. Further consideration was set down for September 2026.
For a buyer, the second one is the consequential change. A regime where large new loads bring their own generation produces a very different price path from one where they simply join the existing pool. It is also the reason the 26% figure above is a scenario rather than a prediction: policy is actively moving against the assumption that scenario rests on.
What Can a Procurement Team Actually Do About It?
You cannot influence how much load connects in Sydney or Melbourne. You control two things, and both are decisions about your own contract book rather than the market.
Know your exposure window. Build a single register of every contract end date across the portfolio, with the consumption behind each one. Most buyers discover they have a cliff: a large share of load re-contracting in one quarter, which hands the timing decision to the calendar instead of to them. Staggering renewals is the least glamorous risk management available and one of the most effective.
Decide tenor deliberately. A short contract keeps you exposed to re-contracting into whatever the market looks like in two years. A long one locks a comparatively soft curve and gives up the benefit if prices keep falling. Neither is right in the abstract. What makes it a decision rather than a guess is knowing your own load shape, your renewal profile, and how much of your bill the wholesale component actually is.
Keep the wholesale component in proportion. Network charges make up roughly 40 to 50% of a typical commercial electricity bill, and those are regulated rather than tendered. A 26% movement in the wholesale component is not a 26% movement in your bill. Buyers who model the whole bill rather than the energy rate make better calls on both timing and tenor.
Have tender-grade data ready before you need it. Validated interval consumption history across every NMI is what turns a market view into a defensible position. It sets your load shape for pricing, supports a shorter and sharper tender, and lets you test scenarios against your own numbers rather than a published average.
How Utilified Turns Market Movement Into Decisions
Most procurement teams cannot answer the timing question quickly, because the inputs live in different places. Contract end dates are in a spreadsheet, consumption history is in a retailer portal per site, and the market view arrives as a broker email. Utilified brings every connection, contract, and invoice into one unified system, with network tariffs, loss factors, and AEMO data built in rather than bolted on.
That gives you a live contract register with renewal dates, validated interval data behind every site, and a market intelligence view covering NEM spot prices and futures curves in the same place as your own portfolio. Ask Joule, our AI utility assistant, which sites re-contract in the next twelve months and what they consumed, and the answer comes back from your own data. When you go to tender, the evidence base is already assembled.
See every contract end date and the load behind it before your next renewal. Book a demo of the Utilified platform →
Frequently Asked Questions
How much electricity do data centres use in Australia?
Data centres connected to the National Electricity Market used around 4 TWh in 2024-25, a little over 2% of grid-supplied electricity. AEMO forecasts close to 12 TWh by 2030, about 6% of the NEM, and roughly 34 TWh by 2049-50, around 12%.
Will data centre demand increase my electricity bill?
It depends on whether new generation and storage arrive alongside the new load. Climate Council modelling published in June 2026 found wholesale prices could be 26% higher in New South Wales and 23% higher in Victoria by 2035 if the demand is met mainly with gas. That is a scenario projection rather than a market operator forecast, and wholesale energy is only part of your bill: network charges alone are typically 40 to 50%.
What is phantom demand in the connection queue?
Phantom demand is connection capacity that is counted more than once because a single project lodges enquiries at several sites or networks before choosing one. Around 44 GW of data centre connection requests were reported to AEMO in its 2025 scenarios process, and analysis suggests roughly six in every seven megawatts will never be built. The firmer figure is the 5.4 GW across 11 projects actually progressing through the transmission connection process as at the March 2026 quarter.
Should I lock in a longer energy contract because of data centre growth?
Not on that basis alone. Wholesale prices fell 47% year on year in the June 2026 quarter, so a longer tenor locks in a comparatively soft curve but forgoes further falls. The better sequence is to map your contract end dates and load shape first, stagger renewals so no single quarter carries most of the portfolio, then decide tenor against your own numbers.