02 Jun The environmental challenges of AI-driven data centres
Artificial intelligence (AI) has become a game changer for the public sector, automating tasks and improving decision-making processes. However, as AI adoption accelerates, concerns surrounding the environmental impact of AI-driven data centres have quite rightfully emerged. With significant power usage, carbon emissions, water consumption and embodied carbon in construction, it is crucial for governments and public sector organisations to address the sustainability challenges posed by AI at the core data centre level.
The overlooked environmental impact of AI
While AI offers efficiency and accuracy across various applications, its environmental costs are often underestimated. Generative AI, which requires approximately five times the power of traditional data centre storage and processing, significantly exacerbates existing environmental issues within the data centre industry. Legacy data centres designed for general-purpose computing simply cannot efficiently accommodate new AI-based GPU servers, their densities and storage capacities.
For instance, traditional data centres have a standard power density of around 8 kW per rack. In contrast, GPU-based servers for AI model training demand 45 to 55 kW per rack or more. As a result, conventional air-cooling systems are incapable of dissipating the heat generated by these high-powered processors.
Sustainable solutions for AI-driven data centres
To reduce AI’s environmental impact, public sector organisations must implement sustainable practices and solutions in their data centres. Key measures include:
- Green power: Ensuring data centres operate on 24/7 renewable energy sources, such as solar, wind and hydroelectric power, can significantly reduce carbon emissions associated with AI computations.
- Efficient cooling: Liquid cooling systems are more effective for AI-driven data centres, reducing power usage by up to 40% while managing heat dissipation.
- Waterless cooling solutions: Data centres consume significant amounts of water, but waterless cooling technologies like immersion cooling or groundwater cooling can minimise water consumption while maintaining optimal temperatures for AI servers.
- Recycling and e-waste management: Responsible recycling and e-waste management policies can reduce embodied carbon in construction and mitigate the environmental impact of decommissioned equipment.
- Collaboration and research: Public sector organisations should collaborate with private enterprises, academia and international organisations to promote research and development of innovative and sustainable data centre technologies.
Incentivising sustainable data centres
Governments can encourage the adoption of sustainable practices in data centres by offering financial incentives, such as tax breaks or grants, to organisations that invest in green technologies or achieve specific sustainability milestones. These incentives can accelerate the transition to environmentally friendly data centres and help reduce the overall environmental impact of AI.
Data centre certification programs
Establishing certification programs for data centres can set clear standards for energy efficiency and environmental performance. These programs can help organisations identify areas for improvement and provide benchmarks for comparing data centre sustainability. For example, the Leadership in Energy and Environmental Design (LEED) certification, widely recognised in the construction industry, can be adapted to create a similar certification program for data centres.
Raising awareness and education
Increasing awareness about the environmental impact of AI-driven data centres is essential to drive change in the industry. Governments and public sector organisations can partner with educational institutions to develop training programs and workshops that focus on sustainable data centre design and management. By educating professionals in the field, we can ensure that sustainability becomes an integral part of data centre operations.
Green procurement policies
Adopting green procurement policies can help public sector organisations ensure that their data centre partners prioritise environmental sustainability. By requiring data centre providers to meet specific energy efficiency, waste reduction and renewable energy targets, public sector organisations can use their purchasing power to drive positive change in the industry.
As well as changing the design of data centres, the drive for sustainable AI will also inform where they are built. Access to low-cost, zero-emissions power is critical and was part of the reason why GreenSquareDC chose to build our 96 MW AI-enabled data centre, WA1, in Perth, Western Australia. The state already enjoys some of the lowest electricity costs in the OECD, and with access to more renewable energy projects than virtually anywhere else on earth. This makes WA an ideal location for building AI-ready data centres, which benefit from access to abundant, cheap renewable energy.
AI has the power to revolutionise the public sector, but its environmental impact cannot be ignored. By implementing sustainable practices and investing in environmentally friendly data centre technologies, governments and public sector organisations can mitigate the negative consequences of AI adoption while continuing to leverage its potential for positive change.
By taking a proactive approach to addressing AI’s impact on the environment, the public sector can set an example for other industries to follow, ensuring a sustainable future for all. The key to striking a balance between AI-driven innovation and environmental sustainability lies in a holistic strategy that encompasses efficient resource management, infrastructure upgrades, innovative cooling solutions and responsible e-waste disposal.
Image credit: iStock.com/imaginima
By Walt Coulston, CEO and founder, GreenSquareDC