As data centre energy demand surges, can the industry still deliver sustainable growth at scale? James Pennington, Global Sustainability Services Director at Lenovo, argues that the answer depends on taking a wider, system-level view of digital infrastructure.
Concern over data centre energy demand is rising fast, yet much of the debate remains narrowly focused on individual sites rather than the digital infrastructure and energy systems that support them.
Today, global data centre energy consumption already stands at 415 TWh, according to the International Energy Agency (IEA) – that’s 1.5% of global energy demand. This is projected to more than double to around 945 TWh by 2030, more than the total energy consumption of Japan.
With data centre energy use growing roughly four times faster than overall demand, the question is no longer whether efficiency matters, but how digital infrastructure is designed to integrate with wider energy and resource systems.
This challenge cannot be addressed by optimising individual data centres in isolation. Data centres interact directly with power grids, water systems, and global supply chains, and their impact needs to be assessed accordingly. Understanding and managing these dependencies is essential if sustainability efforts are to move beyond incremental gains. Meaningful progress depends on a system-level approach that considers energy use, water demand, and how equipment is designed, reused, and recycled over its full lifecycle.
Driving physical efficiency
The AI boom is powered by data centres filled with energy-hungry GPUs, and improving the efficiency of these environments is an important part of the sustainability challenge. Data centre efficiency is commonly measured using power usage effectiveness (PUE): the ratio of total facility power to the power consumed by IT equipment. Lower PUE scores indicate less energy lost to overheads, making accurate, end-to-end measurement critical.
This is where liquid cooling has an advantage over traditional air cooling systems. Water is more efficient at removing heat than air and can improve thermal efficiency. Water cooling can reduce power consumption in the data centre by up to 40%. Some liquid-optimised facilities are now reporting PUE figures as low as 1.1, and in some cases closer to 1.04. At a PUE of 1.1, data centres use significantly less overhead energy, leaving relatively little lost to cooling.
A circular approach
Applying circular-economy principles to data centre hardware and cooling infrastructure can support sustainability objectives. Today, only a small proportion of data centre infrastructure, such as servers, is reused or recycled. Whether through services that extend hardware lifespans or asset recovery programmes that ensure assets are reused and recycled effectively at end of life, data centre operators can make measurable progress on emissions, with circular-economy approaches offering a practical starting point for many organisations.
The benefits of circularity range from reclaiming critical minerals to improving asset reuse and supporting a more sustainable electronics industry. To realise these benefits, business leaders need to take a clear view of the full lifecycle of their data centre infrastructure and servers. Everything from how components are manufactured and designed to shipping and how they are disposed of at end of life will have an impact. Shifting towards as-a-service approaches can also help avoid overprovisioning and reduce carbon emissions.
Circularity also extends to cooling systems in data centres. With today’s warm water cooling systems, hot water can be reused for heating to further reduce the environmental impact of the data centre. Closed-loop cooling systems can also play a role: modern water cooling systems are moving beyond evaporative cooling, where hot water is sprayed onto pads to be cooled, towards closed-loop systems using liquid-to-air heat exchangers, significantly reducing the water demands of data centres.
Improving workload and utilisation efficiency
A key factor often overlooked in conversations around sustainability is workload efficiency. Put simply, the best way to save energy is to avoid unnecessary consumption in the first place. Data centre operators need to ensure every watt spent is translated into meaningful computing output. At the infrastructure layer, virtualisation helps achieve this by consolidating workloads onto fewer servers, reducing idle capacity and increasing overall utilisation. By tuning workloads to make full use of the hardware they run on, operators can link performance gains more directly to lower energy consumption.
Perhaps counter-intuitively, updating older systems can also deliver results, with newer architecture providing more performance at a lower energy cost. This strengthens the case for switching to an as-a-service approach to infrastructure, reducing the up-front capital expenditure associated with technology upgrades. Warm-water cooling systems also enable components to run at higher performance levels without overheating. Starting with one workload at a time, data centre operators can incrementally transform their environments and build momentum towards wider systemic change.
A smarter approach
Improving efficiency in the data centre is only going to become more important as AI workloads drive rising energy demand in the years ahead. Operators will need to take a multi-faceted approach, covering everything from physical efficiency, including energy sourcing and cooling systems, to workload efficiency, such as software design and model choices.
Circular approaches are also likely to become more significant, as operators look more closely at the sustainability case for recycling and reusing equipment, alongside the efficiency benefits associated with as-a-service models.
Treating data centres as part of a wider digital and energy system is now essential if AI-driven growth is to remain economically and environmentally viable.
