Australia's data centre pipeline is quoted in GW capacity, and many different figures have hit the headlines. The most extreme points to 44 GW of connection capacity given to or enquired about by data centres. However, these GW figures do not tell the full story. For the system, what really matters is the power demand that the data centre fleet actually draws, and how quickly.

In turn, this relies on three key factors: what projects will actually get built, what share of connection capacity they draw at maturity (utilisation), and how long they take to get to this point (ramp).

This article steps through Modo Energy's forecast of data centre demand in the NEM. This sees demand from data centres increasing rapidly in the next ten years, nearly quadrupling by 2030 and increasing more than six-fold by 2035. This will turn the sector from a relatively minor energy consumer to representing over 12% of total system demand in just ten years.

The key driver of today's rapid deployment and the shape of future demand growth is the bit-watt spread. That's the gap between the economic value of a unit of compute (a "bit") and the marginal cost of the power to produce it (a "watt"). Today bits are scarce relative to watts which means the spread is wide and investors are racing to close it.

But as more compute lands and efficiency lifts, the spread will compress. The marginal motivation to build new data centres will fall relative to running existing infrastructure more efficiently. Long-term power demand stops being investment-driven and starts depending more on end-use consumption.

In brief

• ​The NEM’s operational data centre fleet drew 3.9 TWh in FY25. Demand is forecast to increase to 15 TWh by FY30 and 39 TWh by FY50. High scenario demand reaches 57 TWh, driven by Australia becoming a hub for APAC regional inference and training load.
• 90% of FY30 demand is already operational or under construction. Today's regulatory debate shapes the post-2030 trajectory rather than near-term growth.
• Project phasing and utilisation ramps push maximum demand out 10-15 years post-commissioning for large projects. This maximum sits between 30-60% of connection capacity for most projects in Australia.
• Demand flexibility is being pressed for but is not expected to be a major function of the fleet. Batteries could instead play a role in mitigating data centre demand at peak times.

Data centre demand growth set to accelerate

The growth of the NEM’s data centre demand fits into two distinct time horizon buckets. Pre-2030 the primary driver is execution: existing and projects in construction ramping utilisation. Post-2035 sees projects reaching full construction and mature utilisation.

2030-2035 sits in the middle, with a combination of existing projects reaching maturity and new projects starting to ramp. This period sees the steepest acceleration in NEM data centre demand, especially in the High scenario. Uncertainty in post-2030 pipeline conversion drives increasing divergence between forecast scenario outcomes.