AI Is Hungry for Power: What Australia's Data

Every time you ask ChatGPT a question, it consumes roughly ten times the electricity of a standard Google search. Scale that across hundreds of millions of users, and you start to understand why AI is being described as one of the biggest new electricity loads the world has ever seen.

In Australia, that energy demand is reshaping where data centres are built, how the electricity grid is managed, and ultimately what businesses pay for cloud services. Here’s what’s happening and why it matters.

The Scale of the Problem

The International Energy Agency (IEA) estimates that global data centre electricity consumption will roughly double by 2026: largely driven by AI workloads. In Australia, the numbers are moving in the same direction:

AWS, Microsoft Azure, and Google have each committed to major new Australian data centre expansions: billions of dollars of investment in Sydney and Melbourne facilities announced between 2023 and 2025.
The hyperscale data centres supporting these expansions consume hundreds of megawatts each: comparable to a mid-sized regional city.
AEMO (the Australian Energy Market Operator) has specifically flagged large data centre loads as a material factor in its demand forecasting for the NEM (National Electricity Market).

Where the Power Is Coming From

Australia’s electricity grid is in the middle of a significant transition: from coal-heavy to renewables-heavy. Data centres are accelerating this transition in two ways:

Driving demand for renewable PPAs: Microsoft, Google, and AWS have all committed to 100% renewable energy matching for their Australian operations: driving investment in new solar and wind projects that are contracted specifically to supply data centre loads.
Creating grid management challenges: Large, relatively inflexible data centre loads complicate grid balancing: particularly as more variable renewable generation comes online. AEMO is actively working on mechanisms to make data centre loads more responsive to grid conditions.

The Snowy 2.0 Connection

Snowy 2.0: the massive pumped hydro project in the Snowy Mountains: is often discussed in the context of renewable energy storage. But AI data centres are a significant part of the demand story that makes large-scale storage like Snowy 2.0 economically necessary.

Pumped hydro stores excess renewable energy (generated when supply exceeds demand) and releases it during peak demand periods. As AI data centres run 24/7: creating a consistent baseline load that doesn’t follow the sun or the wind: the interaction between AI energy demand and storage capacity becomes more complex.

Snowy 2.0 has faced significant cost overruns and delays, but its long-term purpose: providing dispatchable storage to back up a renewables-heavy grid: becomes more important, not less, as data centre loads grow.

What It Means for Australian Business Energy Costs

The energy economics of AI create competing pressures on Australian business electricity prices:

Upward pressure: Large data centre loads increase overall electricity demand, which (all else equal) puts upward pressure on wholesale prices. In periods of high demand and constrained supply, data centres compete with industrial and commercial users for available generation.
Data centre operators’ investment in renewable PPAs is driving new renewable generation that increases supply and reduces reliance on expensive gas peaking generation. More supply, over time, puts downward pressure on wholesale prices.
Location effects: Data centre investment is concentrating in areas with good grid connectivity and renewable resource: primarily NSW and Victoria. This may affect the relative electricity pricing in different states over time.

Cloud Costs and AI

For small businesses, the most direct impact of AI energy costs is in cloud service pricing. Running AI models is expensive: the compute costs for inference (running AI responses) are significantly higher than for standard web services. AWS, Azure, and Google pass these costs through in their AI service pricing.

This is why AI API costs (what it costs to call the ChatGPT API, for example) have been declining as providers scale and optimise: but remain significantly higher per query than traditional database or web service calls. For small businesses building AI-powered features into their products or workflows, this cost structure matters for what’s economically viable.

🦅 The practical read for small business: AI energy costs are real but mostly invisible to small businesses using AI through subscription tools (ChatGPT Plus, Claude Pro). You’re paying a flat monthly fee that absorbs the compute cost. Where energy costs become visible is if you’re using AI APIs directly: in which case, model choice and prompt efficiency genuinely affect your monthly bill. For most small businesses, this is not yet a material cost driver: but it’s worth understanding as AI use scales.

Energy market data sourced from AEMO and publicly available industry reports. This article is general information.

Sources

📊 Compare AI tools side by side | 💼 Free resources & AI prompt packs

📖 More deep dives: News Deep Dives

Frequently Asked Questions

What is the scale of the problem?

Where the Power Is Coming From?

What is the snowy 2.0 connection?

What It Means for Australian Business Energy Costs?