Power hungry: What role can renewables play in the AI revolution?

The rapid advancement of artificial intelligence is already bringing about significant changes, revolutionising industries from healthcare to finance, the public sector to, well, management consultancy. But this revolution comes at a cost: a substantial increase in energy consumption.

While we at Nous are firm believers in the potential of AI, we are equally committed to the world’s net zero goals. It would be remiss of us not to acknowledge the issue and to think about ways that industries might address it.

Recent data highlights that AI energy requirements are exceedingly high. For example, a single ChatGPT query consumes roughly 10 times the energy of a Google search. Gartner projects that AI data centres will consume 500 TWh in 2027, 2.6 times more than in 2023. The International Energy Agency forecasts that data centres around the world could consume as much power as the entire nation of Japan by 2026. 

AI players are scrambling to secure the required power using a range of energy technologies. Microsoft has inked a deal to reopen the infamous Three Mile Island nuclear power plant, primarily to meet the energy needs of its AI operations, and has committed $1.5 billion toward 10.5 GW of new renewables capacity. Meanwhile, Elon Musk’s xAI, uses 15 gas turbines – which reportedly emit hazardous air pollutants, including formaldehyde, at dangerous levels – to power its Colossus supercomputer. These escalating energy needs are leaving a heavy emissions footprint. For example, Google’s CO₂ emissions have risen by 50 per cent since 2019, largely due to data centre demand, which is increasingly being driven by AI and its requirements.

Energy use is becoming both a rate limiter on AI progress and a major cost driver. While strides are being made to improve efficiency, the gains are complex. China’s DeepSeek model marked a breakthrough in efficient model training, but the energy it saves upfront could be largely offset by its heavier demands during inference (the generation of outputs in real time), a characteristic of reasoning models. Meanwhile, a shift toward on-device AI, designed to reduce reliance on the cloud and prioritise energy efficiency over raw computing power, is gaining traction.

But greater efficiency doesn’t mean lower overall energy consumption. On the contrary, as AI becomes cheaper to deploy, companies are increasing investment in training and inference, pushing total energy use even higher.

All this underscores the magnitude of the technology’s energy requirements and environmental impacts, which are layered atop the already existing demands of the tech industry, including cloud storage and cryptocurrency mining, which are themselves significant drains on energy. The challenge is to balance this rapidly increasing demand with the imperative to reduce carbon emissions and achieve our net zero goals.

One promising solution – especially for Australia, with our excellent renewable resources – is the rise of green data centres. As countries search for sustainable ways to meet their growing energy demands, Australia has the potential to lead in this domain. Do we have what it takes to position ourselves as a global exporter of digital energy solutions?

AI adoption is no longer a matter of choice, but a necessity. At the same time, Australia and the world face non-negotiable net zero ambitions and obligations. There is a genuine risk that the allure of competitive advantage in the digital sphere could prompt some to prioritise technological development over environmental sustainability.

However, there is an alternative course that may be taken. In this scenario, the push towards sustainability and renewable energy doesn’t halt in the face of technological demands. Instead, precisely because the energy demand is so significant, and the risk of meeting it with non-renewable sources so potentially damaging to the planet, development and adoption of renewables accelerates instead. 

Several considerations come into play here. Load shifting around the availability of solar and wind power, optimising infrastructure use, and increasing energy efficiency will be obviously crucial. Additionally, the pace of infrastructure development needs to keep up – no mean feat – with the increasing demand of the tech giants. Expanding domestic data processing capacity is also necessary, not only as a result of AI, but also due to existing capacity constraints. This, too, would seem to point at the importance of ramping up renewables now.

At the same time, we should recognise the potential for a symbiotic relationship between AI and renewable energy, leveraging this to move closer towards our sustainability goals. Rather than seeing AI and renewable energy as competing priorities, it's crucial to understand how they can complement each other. AI can be used to optimise energy usage, support grid stability, and improve renewable energy efficiency. While large data centres place heavy demands on the grid, they also have the potential to offer benefits to it through load shifting, demand response, and even ancillary services. It is here that Australia might carve out a role for itself, not only as a renewable energy leader, but also a leading player in the AI and data centre revolutions.

Australia stands at a unique intersection of opportunity and obligation. With abundant solar and wind resources, the country is well-positioned to lead in the development of green data centres. Such infrastructure could significantly bolster Australia's energy market and contribute to global sustainability efforts. GreenSquareDC’s Perth data centre, WAi1, could potentially be the first of many: a purpose-built facility for resource-intensive AI processing powered by clean energy.

By positioning itself as a regional hub for green data centres, Australia can not only meet its own data capacity needs, but also export digital energy solutions, including AI model training. The country's expertise in renewable energy can also attract investments and foster partnerships that further enhance its standing in the global energy landscape.

However, several challenges must be addressed to realise this potential. The scalability and efficiency of green data centres are essential concerns. Ensuring that these centres can operate at scale while maintaining efficiency will require substantial technological and infrastructural investments. Location, too, will be key: the idea of positioning green data centres in our most sun-kissed regions sounds easy enough, but overlooks both latency issues – data centres need to be near users to keep latency low – as well as the heat generated by AI servers. (WAi1 uses liquid cooling systems and groundwater cooling.)  

Additionally, Green data centres face many of the same hurdles as other infrastructure and renewable energy projects, from lengthy approval processes to supply chain bottlenecks, workforce constraints, and social license challenges. Addressing these pressures will require coordinated action from policymakers and industry to build a supportive environment for sustainable data centres and to accelerate investment in renewable energy.

Looking ahead, the integration of AI and renewable energy offers a significant opportunity. The following steps can help in achieving this vision:

• Strengthening policy support. Policymakers must create robust frameworks that incentivise the development of green data centres, renewables, and storage mechanisms.
• Investing in research and development. Continuous investment in R&D is crucial to advance technologies that improve the efficiency and scalability of green data centres. Universities and private companies should collaborate on research initiatives that push the boundaries of what is possible in this space.
• Public-private partnerships. Collaboration between the public and private sectors can accelerate the deployment of renewable energy solutions. Successful models from other industries should be studied and adapted to the energy sector to foster a spirit of cooperation and shared goals.
• Education and training. Building a workforce capable of supporting the green revolution in data management is essential. Developing educational programs and training initiatives that focus on sustainable practices and renewable energy can ensure that Australia has the necessary human capital to lead in this domain.
• International collaboration. Australia should seek to establish strong international partnerships to share knowledge, technology, and best practices in renewable energy and sustainable data management. These collaborations can help position Australia as a global leader in green technology. In areas where Australia does not yet have the technological expertise, importing that expertise may be necessary, with the stipulation that experts being brought in should be required to help train Australians and build local knowledge capacity.

The convergence of AI and renewable energy presents both a challenge and an opportunity. Australia's abundant renewable resources offer the country a unique advantage that can support the burgeoning energy needs of AI technologies. 

Through policy support, investment in innovation, and a canny use of public-private partnerships, Australia may just be able to navigate the seemingly but not necessarily incompatible demands of technological advancement and environmental responsibility. In doing so, the country can set a benchmark for sustainable development in the age of AI, showcasing how competing revolutions might in fact turn out to be one and the same.

Get in touch to discuss the intersection of AI and renewable energy. 

Connect with David Diviny, Tom Strawhorn, and Mohamed Daif on LinkedIn.