Simon Yeoman, CEO at Fasthosts, discusses the integral role sustainable practices play in creating AI-era ready data centres.

Underneath the bonnet of our digital lives hums an unseen engine: data centres. They collect and process the lifeblood of our online world, from the social connections we foster and financial transactions that power our economies to the music and films we stream.

However, this persistently humming mechanism carries a significant price tag.

As we venture deeper into the AI age, the role of data centres becomes even more vital. The rapid growth of this technology demands ever-increasing storage capabilities and computational power with generative AI potentially requiring up to 33 times more energy than conventional software.

The UK AI market, currently valued at over £16.8 billion, is outlined to surge to £801.6 billion by 2035, stressing the serious need for sustainable solutions. While data centres have historically focused on tackling energy efficiency with advancements in server technology, renewable energy and cooling, the increasing demand for water to cool these facilities poses a new threat. 

The double burden

Data centres are known for their substantial energy use, and although efficiency gains have been made, the relentless growth of AI is pushing consumption to unforeseen heights. Beyond energy, data centres are now also grappling with excessive water usage. 

The escalating demands of data processing and AI require robust cooling systems, which in turn require higher water usage. This dual challenge of energy and water consumption presents a substantial sustainability barrier for the industry to navigate.

Leaning into change  

One innovative solution is the adoption of liquid cooling technologies. These systems use closed-loop mechanisms that drastically reduce water usage and allow heat reuse. By circulating a coolant liquid through server components, these systems efficiently absorb and dissipate heat, conserving water and improving cooling effectiveness.

In regions with suitable climates, air-cooled systems offer a more sustainable alternative. These systems use ambient air to cool data centres, reducing the reliance on water-based cooling. Combining air cooling with advanced air filtration techniques can further enhance sustainability by cutting down water use and keeping temperatures within optimal ranges.

Alternatively, existing water-based cooling systems can be improved by water recycling and reducing the amount of water that evaporates. This allows data centres in drought-prone areas to reuse water within their cooling cycle and minimise water loss. This approach conserves water and helps the data centre remain efficient and operational even during water shortages.

As we move forward, strategically locating new data centres in regions with abundant water resources or cooler climates, and migrating existing ones to such regions, should be a top priority. 

Redefining sustainability in the AI-era

Undeniably, innovative solutions are crucial, but the inevitable surge in AI adoption necessitates a paradigm shift. That’s why the focus can no longer be on simply lowering the number of kWh or litres of water used. Instead, these resources must be used as efficiently and effectively as possible, ensuring every bit contributes to maximising performance and sustainability.

To address this challenge, managing modern data centres also requires sophisticated tools for real-time monitoring and optimisation. Data enmeshing, which involves integrating data from various sources, such as power consumption, cooling efficiency and environmental conditions, provides a comprehensive view of operations.

This approach helps spot inefficiencies, such as overworked cooling systems, and can make targeted improvements. For example, detailed power usage effectiveness (PUE) reports can identify outdated servers that need replacing with more energy-efficient models.

Improving environmental efficiencies  

Proper capacity planning is another critical aspect of efficient data centre management. It involves predicting future data centre needs and making sure that the infrastructure can handle increased loads without over-provisioning resources. This helps prevent wasted energy and resources on underutilised servers and systems, while also enabling data centres to scale to meet growing demands.