Need for Greener Practices in Data Centres

A recent report released by the National Engineering Policy Centre (NEPC) underscores the critical necessity for data centres to implement environmentally friendly practices, especially given the momentum behind the government’s AI Opportunities Action Plan.

The report, titled Engineering Responsible AI: Foundations for Environmentally Sustainable AI, was produced in partnership with esteemed organisations including the Royal Academy of Engineering, the Institution of Engineering and Technology, and BCS, the Chartered Institute of IT.

It emphasises that while data centres facilitating AI technologies can indeed be designed to be more efficient in their utilisation of resources such as energy and water, achieving these efficiencies requires a well-aligned regulatory framework and infrastructure.

Harnessing AI’s Potential While Reducing Environmental Footprints

AI is increasingly viewed as a powerful tool for fostering economic development, generating new job opportunities, and enhancing quality of life. Central to the UK’s technological strategy, the AI Opportunities Action Plan aims to catalyse economic growth, secure future employment, and enhance everyday life.

Communities are already reaping the benefits of AI in various sectors, such as accelerating pharmaceutical research, predicting weather patterns, optimising energy consumption, and supporting climate studies and sustainability initiatives. However, the rising dependency on AI brings notable environmental challenges stemming from the infrastructure necessary to support it.

Data centres, which are integral to AI operations, are significant consumers of energy and water. This escalating demand raises concerns over competition for scarce resources like sustainable energy and potable water. Notably, tech giants like Google and Microsoft have documented an increase in water usage within their data centres annually since 2020, much of which is drawn from drinking sources, further intensifying fears regarding resource depletion.

With plans to reform the UK’s planning system to facilitate data centre construction already underway, the report advocates for immediate policy measures to manage the environmental repercussions of these facilities. Currently, there’s a lack of transparent, reliable data on resource consumption, making it difficult for policymakers to gauge the real extent of environmental impacts and act decisively.

Key Steps Toward Sustainable AI Development

The NEPC has implored the government to take the lead in establishing sustainable practices in AI development, outlining five actionable recommendations for policymakers to position the UK as a frontrunner in resource-efficient AI:

1. Enhance environmental reporting requirements.
2. Improve communication regarding the environmental impacts of the sector.
3. Establish sustainability standards for data centres.
4. Reassess data collection, storage, and management strategies.
5. Pursue government investments as a model for sustainable practices.

Central to these recommendations is the push for mandatory environmental reporting, which would necessitate data centres to track their energy sources, water usage, carbon emissions, and electronic waste recycling efforts—thus providing the necessary data to inform policy decisions.

Increased public awareness is also paramount. By effectively communicating the environmental costs linked to AI, it can incentivise developers to streamline AI applications, minimise data usage, and employ more energy-efficient methodologies. The report further advocates for integrating environmental design and sustainability themes into computer science curricula at both educational levels.

Innovations for Eco-Friendly Data Centres

A major focus of the report is the urgent need to reengineer data centres to lower their ecological impact. Recommendations include innovations such as waste heat recovery systems, eliminating the use of potable water for cooling, and committing to 100% carbon-free energy.

One inspiring initiative is being implemented at Queen Mary University of London, where excess heat from a data centre is redirected to supply heating and hot water across the campus, showcasing the potential for a more sustainable technological infrastructure.

Additionally, the report suggests a revision of policies concerning data retention mandates to alleviate the excessive environmental burden posed by the prolonged storage of large datasets. Proposals for creating a National Data Library could promote best practices by centralising and optimising data management.

Professor Tom Rodden from the University of Nottingham and head of the working group behind the report, emphasised the urgency of addressing these challenges: “Recent years have seen the evolution of AI systems directed by an insatiable quest for scale, necessitating greater computational power. Consequently, AI services are expanding at an unprecedented pace compared to other high-energy systems, often with little consideration for resource efficiency.”

He added, “This trend poses grave risks, threatening irreversible damage to our environment. Acquiring reliable data on these impacts is essential; by accurately monitoring the environmental costs, we can start to prioritise efficiency in our development and use of AI, paving the way for a sustainable AI landscape in the UK.”

Dame Dawn Childs, CEO of Pure Data Centres Group, stressed the vital role engineering plays in boosting efficiency. “While advancements in AI models and hardware can drive down energy consumption, we must ensure the data centres housing these technologies operate sustainably, prioritising renewable energy sources, minimising water usage, and reducing emissions from the outset,” she suggested.

The Call for Transparency on AI’s Environmental Costs

Despite its increasing importance, public awareness regarding the environmental impact of AI remains alarmingly low. Research conducted by the Institution of Engineering and Technology (IET) revealed that less than one in six individuals in the UK are cognizant of the significant environmental consequences associated with AI technologies.

“AI providers must be forthright about these implications,” remarked Professor Sarvapali Ramchurn, CEO of Responsible AI UK and a Fellow of the IET. “If we cannot accurately measure these impacts, we cannot effectively manage them or ensure equitable benefits for all. The suggestions put forth in this report will facilitate national dialogues on the sustainability of AI systems and the compromises entailed.”

As the UK takes bold strides towards establishing itself as a leader in AI development, maintaining a focus on environmental sustainability should be paramount. By implementing policies and practices suggested in the NEPC report, the government can nurture growth within the AI sector while safeguarding limited resources for future generations.