How Data Centres Can Make the Global AI Race Sustainable
The continued proliferation of artificial intelligence (AI) is driving a power surge, with data centres bearing the brunt of its energy demands. Yet, as these needs keep swelling, so are concerns over carbon emissions.
The International Energy Agency (IEA) has already indicated that data centres already accounted for 1-1.5% of global electricity consumption in 2022, ahead of peak AI interest. Likewise, the organisation found that global electricity demand grew 2.2% in 2023, warning this figure could double by 2026.
This, in turn, could double carbon dioxide (CO2) emissions from data centres (2022-2030), meaning that data centres must harness AI sustainably so as to not threaten net zero targets, in addition to global energy security.
The fight to keep data centre emissions down
Data centres continue to dominate global conversation with regards to innovation and sustainability. In September 2024, the UK government designated data centres as Critical National Infrastructure (CNI), which demonstrates the growing importance that these facilities hold in today’s digital economy.
This was hailed as good news for the UK economy, alongside AWS’ intentions to invest £8bn (US$10.5bn) into UK data centres moving forward. However, such rapid expansion efforts could threaten the overall sustainability of the industry.
In fact, this week (September 2024) it was reported that emissions from in-house data centres of tech giants Google, Microsoft, Meta and Apple may be 7.62 times (662%) higher than official figures.
In recent months, these organisations have struggled to keep overall emissions down, with Google in particular reporting a 13% increase earlier in 2024 on account of its AI data centres.
“The race for AI dominance is heating up, but at what cost? The increase in emissions is largely down to legacy data centres unable to cope with the power demands of AI,” comments Elio van Puyvelde, Chief Information Officer at Nscale. “And while big tech invests heavily in renewables, the sheer scale of the AI boom threatens to overwhelm those efforts.”
Morgan Stanley suggests that data centre carbon emissions are set to triple by 2030 on account of AI, emitting 2.5 billion tonnes of CO2 in the process. As the industry confronts AI innovation, the challenge to keep energy consumption low becomes both a technological and ethical challenge.
To mitigate these issues from escalating long-term, Elio suggests that data centres can and should be built with sustainability in mind.
“Ensuring data centres are built where there is a stable supply of renewable power is one thing, but the industry must focus on maximising efficiency too,” he says. “The reality is AI workloads are so often poorly optimised, wasting energy use and delivering poor returns. Accelerating AI hardware and software optimisation is one of the best routes to managing AI energy consumption levels.
“Companies must optimise and fine-tune AI models as well as invest in energy efficient AI accelerator hardware.”
Making sure innovation is sustainable
As data centre investment continues on its upward trajectory, businesses involved are having to consider how compute infrastructure needs to adapt to suit the demands of AI.
“Organisations across the globe are rushing to be the biggest and best AI innovator, a goal driven by the transformative power of this technology,” comments Dave King, Senior Principal Product Engineer at Cadence.
“The extreme power densities and power demand have caused a cataclysmic shift in how data centres must power and cool the technology behind the AI revolution. At a time when many organisations are embarking on sustainability projects to meet incoming legislation, balancing environmental responsibilities with being a progressive AI champion is proving to be a difficult challenge.”
“As data centre industry leaders face the daunting task of weighing AI’s capabilities against its environmental toll, the challenge becomes not just technological but ethical.”
There are, however, some ways of reducing the carbon footprint of data centres that embrace the power of technology. Used responsibly, these could enable companies working within the sector to address both current and future capacity.
Dave suggests digital twin technology as an example, a virtual representation of an object or system designed to accurately reflect a physical object.
“These virtual replicas of the physical data centre empower facility managers to reduce the carbon footprint of AI by enabling them to address not only current but future stranded capacity,” he says. “It also allows data centres to improve power management and assess the effectiveness of heat rejection. The benefits continue as they can even be used to identify the least amount of energy needed to cool a facility without impacting its operational effectiveness.
“By prioritising data centre efficiency through digital twins, AI's growth doesn't need to burden the environment. Rather, this shift fosters responsible innovation and will reduce the need for endless data centre construction.”
******
Make sure you check out the latest edition of Data Centre Magazine and also sign up to our global conference series - Tech & AI LIVE 2024
******
Data Centre Magazine is a BizClik brand
