The rising energy demand from data centres: A global perspective

In recent years, the world has experienced a significant surge in the number of data centres, resulting in a rise in energy demand, posing both challenges and opportunities for the global energy sector. This raises several questions: What has caused the exponential growth in data centres? What types of energy sources will be required to support this growth? What implications are there for existing grid networks? How might the rise of new and emerging Artificial Intelligence (AI) software influence the ongoing expansion of data centres?

Just how quickly has the data centre industry been growing?

With 250 data centres spread across the country and a further 175 predicted to be needed by 2030, Australia is one of the world’s top five data centre hubs. The data centre market in Australia is set to increase exponentially over the next few years, which will greatly impact many aspects of Australian business, especially when it comes to the demand for power. Currently, data centres in Australia consume about 5 per cent of the national electricity generation, which is expected to grow to 8 per cent by 2030. Some estimates even suggest that data centres could require up to 15 per cent of total grid power by then.

What has caused the exponential growth in data centres?

Several factors are contributing to the rapid increase in the number of data centres worldwide:

  • Cloud computing: The shift from on-site IT infrastructure to cloud-based services has expanded centres to host and manage vast amounts of data. Businesses of all sizes are migrating to the cloud, driving the need for more data centres.
  • AI and machine learning: Advanced AI algorithms require substantial computational power, which demands high-performance data centres. The recent explosion in generative AI applications has further accelerated this trend and will likely be the key driver behind future demand growth.
  • Digital transformation: The advent of 5G networks and the increasing adoption of digital solutions across various industries has significantly increased the volume of everyday data generated. This data needs to be stored and processed, driving a need for data centres.

The growth in demand arising from each of these factors is only going to increase, and as AI becomes all-pervasive, the demand for an ever-increasing number of larger data centres will increase.

Why do data centres need so much energy?

One of the key challenges for Australia will be how best to provide the energy for the growth of data centres and, even more so, how to supply this using clean energy. Data centres consume vast amounts of energy due to several key factors.

Firstly, the sheer volume of data being processed, stored, and transmitted requires significant computational power, and servers, storage devices, and networking equipment all need electricity to operate efficiently.

Secondly, these components generate substantial heat, necessitating robust cooling systems to maintain optimal operating temperatures and prevent overheating. Cooling alone can account for nearly 40 per cent of each data centre’s total energy consumption.

Of course, data centres generally run non-stop to ensure continuous availability and reliability, leading to constant energy usage. Hyperscale data centres, operated by tech giants like Google, Amazon, and Microsoft, are particularly energy-intensive due to their massive scale and high-performance requirements.

What energy sources are needed to meet this demand? 

Balancing the required energy with Australia’s carbon emission targets will mean that Australia can’t rely solely on fossil fuels to meet this demand. To support the burgeoning number of data centres, a diverse mix of energy sources are required:

  • Renewable energy: Data centres are increasingly using wind and solar energy due to their clean power capabilities. However, their weather-dependent nature makes them less reliable as standalone providers. Data centres need constant power, meaning that to be viable, renewable energy sources need to be coupled with cost-effective and reliable energy generation and storage solutions.
  • Gas-powered generation: To ensure a stable and continuous power supply, gas-powered generation can be used as a supplementary energy source. It provides a reliable backup when renewable sources are insufficient and can help bridge the gap during peak demand periods.
  • Nuclear energy: Small modular reactors (SMRs) could offer a reliable and sustainable energy source that seamlessly integrates with intermittent renewables like solar and wind. SMRs provide steady power and can assist with the cooling needs of AI data centres. However, public perception around nuclear energy together with significant development and deployment timeframes remain limiting factors. Nuclear energy should be considered as a long-term option rather than short-term solution.

What will be the impact on existing grid networks?

The rapid growth of data centres has already put a strain on existing grid networks, leading to several connectivity issues, and this is only going to get worse due to factors including:

  • Aging infrastructure: Many regions have outdated transmission lines and substations that cannot handle the rising demands of data centres, leading to congestion and instability in the local grid.
  • Grid connection queues: In some areas, the wait times for grid connections can be several years, delaying projects and impacting their viability. This is particularly problematic in regions with limited transmission capacity.
  • Power supply issues: Power shortages will force data centres to not only bid against other users for access to electricity, but also to install additional on-site power-generating equipment to guarantee 24/7 operations. Additional considerations include:
    • Future power and utility plans are already a central factor in deciding where to build data centres
    • Data centre operators will install significant on-site power generation and storage to enable continuous operation when available utility power is reduced or fails. This is more than traditional backup power—on-site generation must be able to provide extended or continuous runtime when needed
    • Major data centre operators are looking at emerging clean technologies such as nuclear, hydrogen, geothermal and even fusion for longer-term power generation.

To mitigate some of these concerns, advanced conductors, including technologies like high-temperature superconductors and innovative alloys, are being explored to improve electricity transmission efficiency and reduce grid strain.

Is this increase in demand from AI inevitable?

With new AI technologies becoming more and more prevalent, new dynamics in the energy demand equation for data centres are emerging:

  • Reduced energy consumption: Some AI technologies are becoming more efficient in training high-performance models with less electricity, potentially reducing the overall energy consumption of data centres.
  • Increased AI usage: Conversely, the reduced cost of using AI due to these efficiencies could lead to increased usage of AI applications, thereby driving up energy demand.
  • Uncertain projections: The emergence of these efficient AI technologies has added uncertainty to projections of future energy demand for data centres. While they may reduce the need for massive hardware outlays, they could also lead to higher overall usage of AI, complicating energy demand forecasts. There is likely to be a wider divergence of projections for AI use, reflecting the uncertainty of what the next innovation might be and whether it will increase or decrease energy demand.
  • Impact on energy investments: The efficiency gains from AI technologies may influence investment decisions in the energy sector, particularly in nuclear and natural gas capacity, as the perceived need for new power plants may be reassessed.

In a significant move to address the rising energy demands of its data centres, Google has announced a partnership with Kairos Power to build seven small nuclear reactors in the U.S. The first reactor is expected to come online by 2030, with the remaining reactors deployed by 2035. This initiative aims to provide 500 megawatts of power to Google's AI technologies, highlighting the tech giant's commitment to clean, round-the-clock power sources.

Conclusion

The huge increase in data centres is a double-edged sword, driving technological advancements and causing significant challenges for the energy sector. A diverse mix of energy sources, coupled with innovations in grid connectivity and AI efficiency, will be crucial in sustainably meeting growing energy demands. As new technologies evolve, their impact on energy usage will need to be closely monitored to ensure a balanced and sustainable approach to powering our digital future.

How BDO can help

BDO’s natural resources & energy experts combine deep industry knowledge with a global perspective to help clients effectively navigate the complexities of the natural resources sector. For more information on how we can support your needs in the natural resources sector, contact us.