E X P E R T O P I N I O N f 2023 was the year

AI took the world by storm, 2024 is the year that regulators took notice. Around the globe, rules are being drafted, passed and implemented to ensure that the development of AI doesn ' t infringe upon human rights or universal values and priorities, such as environmental stewardship.

Nowhere is this more evident than in Europe, where regulators have been busy approving new rules over the use of AI, as well as its underlying infrastructure, including data centres. At the same time, regulators claim they want to foster an environment that encourages AI innovation. According to the European Union:‘ The European AI Strategy aims at making the EU a worldclass hub for AI and ensuring that AI is human-centric and trustworthy.’

Implementing regulations without stifling innovation is always a challenge. Enacting sustainability rules will be especially challenging now, as AI dramatically increases the digital economy ' s energy consumption. In Europe, sustainability is at the forefront, but this new dynamic accounting for 2.7 % of the region ' s electricity demand. By 2030, it could reach at least 3.2 %.

With AI on the rise, as well as other energy-intensive technologies like virtual reality and cloud gaming, Ireland ' s data centres could double their electricity use by 2030, according to IEA estimates, while Denmark ' s data centres could see a six-fold increase in the electricity use.

has left the path forward somewhat unclear. It will be up to regulators and industry stakeholders alike to live up to sustainability expectations while supporting Europe ' s fledgling AI ecosystem.

The energy demands of AI

Artificial Intelligence became a mainstream tool in late 2022, with the launch of OpenAI ' s virtual assistant, ChatGPT, an advanced Large Language Model( LLM). Not long after, the energy implications of AI became clear. The International Energy Agency( IEA) noted that ChatGPT uses on average 2.9 watthours( Wh) of electricity every time it answers a user request. By comparison, a typical Google search takes just 0.3 Wh.

In 2022 – with little to no LLM deployments – data centres consumed 460 terawatt-hours( TWh) of electricity globally, accounting for 2 % of the world ' s energy usage. However, with the emergence of purpose-bult AI data centres for LLM workloads, that is changing quickly. Earlier this year, the IEA forecast that by 2026, the electricity consumption of data centres could more than double, surpassing 1,000 TWh.

In the EU, data centres in 2018 consumed 76.8 TWh of energy,

Europe ' s new data centre sustainability rules

Amid this environment of new, energyhungry AI applications, the EU is about to implement new sustainability rules. Starting in September, data centre operators in EU nations will have to follow the updated Energy Efficiency Directive. This will require them to regularly disclose their energy and water consumption, demonstrate their use of renewable energy sources and measure the effectiveness of cooling systems, among other things. The goal is to reduce EU energy consumption by 11.7 % by 2030, relative to the forecast energy consumption for 2030 made in 2020.

Data centre operators will also have to follow new sustainability requirements included in Europe ' s pioneering AI rules. Earlier this year, the EU passed the AI Act, the world ' s first major piece of legislation to impose rules governing the use and development of AI.

The law creates a framework of principles for regulating AI. For instance, given the risks associated with biometric identification systems, the use of remote biometric identification in public spaces for law enforcement purposes is, in principle, prohibited.

www. intelligentdatacentres. com 43