Power first approach can help solve the data centre dilemma

Data centres are on the rise, driven by the surge in data usage, the rapid expansion of digital technologies and the growth in Generative AI and machine learning.

Australia is uniquely positioned to maximise capacity – we are ranked in the top 5 data centre players in terms of capacity globally. According to research from Morgan Stanley Australia’s data centre capability currently stands at around 1,000 MW and is expected to reach 2,500MW by 2030. To put that in perspective a more established market like the US, has a data centre capability of around 9,000MW today.  

While the opportunity for growth is exciting for both developers, investors and the community, there remains a key challenge – power availability. Powering data centres has become a critical issue for the industry globally. The vast amount of energy needed to power data centres especially hyperscale projects is challenging grids already under pressure from decarbonisation, intermittent nature of renewables, ageing infrastructure and changing energy requirements.¹  

According to Morgan Stanley, by 2030, data centres will potentially be making up 8% of the grid's demand in Australia, an increase from the current estimate of 5%. Beyond that point, power availability could become a limiting factor. In Ireland the grid directs 20% of electricity to data centres and in the Dublin metro area it sits at around 35%. Power is already constrained in some areas like Ireland and Singapore where we have seen Governments intervene to protect the network. 

So, if data centres are only going to get bigger and need more energy, how can we solve this dilemma? 

The three biggest challenges for data centres connecting to the grid 

Our geographical location makes Australia a key strategic hub for data centres, serving both the Asia-Pacific region and local markets. Which is why addressing the existing challenges will be essential to realising our potential.  

Location, location, location 
Historically the prime real estate for data centres has been in urban areas, close to people to enable low latency. Low latency refers to the minimal delay or lag in processing and transferring data, ensuring faster response times in networks and applications.  However, with the development of AI and computer learning it has made way for non-latency dependent locations meaning data centres can be developed outside of the metro/urban areas. This is starting to play out primarily in the US and EMEA.  

While this helps with the existing land constraint, regional areas in Australia may not have the telecommunications, water or road infrastructure to cope with the energy demands of a hyperscale data centre campus.   

High energy demands and network limitations 
With the growth of artificial intelligence related IT services and a manifold increase in the demand for cloud computing, we anticipate exponential growth in new data centres across Sydney and its surrounds, necessitating substantial growth in load. Recent estimates from Distributed Network Service Providers (DNSPs) suggest around 850 MW of potential new data centre load is expected to connect to the distributor networks in the next 10 years.² 

The distribution network is not designed to accommodate such large-scale connections en masse, making it challenging to identify suitable locations within the network that can support multiple high-capacity data centres. Especially as data centres require complete network reliability and power outages continue to be cited as the single biggest cause of concern for developers – primarily in the US.³  

Transgrid maintains a 99.99999% reliability rate for its transmission network, delivering stable electricity across 13,160 kilometres of transmission lines in New South Wales⁴.  

Historical development patterns  
Historically, data centres have identified and secured land first, then solved power availability issues by bringing power to the land.  

This approach worked because zoned land appropriate for development was typically located in metropolitan areas, where the sub-transmission network existed. This allowed data centres to connect easily to the distribution network and given the smaller size of the data centres this was not a challenge for the network.  

However, this dynamic is now changing.  

The distribution network can only supply limited capacity, which gets quickly consumed and it must be retained for residential customers to ensure they are not impacted through increasing load demands. This leaves limited capacity for future growth and expansion particularly for data centres. This is evident from current growth challenges for data centres in the Macquarie Park Sydney region.  

Why it’s time for a power first approach  

With the expansion of AI and machine learning we need to find a different approach in setting up next generation data centres. They are going to be bigger, stronger and have higher energy demands. Which is why we are suggesting a shift in the supply chain, where its power first, then land. By adopting a power first strategy we can consider power planning before land acquisition to ensure investment certainty for customers, flexibility of growth in the future, provide network stability where it is needed and avoid power shortages.