
Why Data Centre Sustainability Needs a System-Level Approach


Why Data Centre Sustainability Needs a System-Level Approach

Global data centre energy consumption is at 415 TWh and could more than double to 945 TWh by 2030, surpassing Japan's total energy use, according to the International Energy Agency.
As AI workloads accelerate this, the industry faces a choice between optimising individual components in isolation or adopting a system-level approach that addresses efficiency, resource use and circularity together.
James Pennington, Global Sustainability Services Director at Lenovo, says the shift requires operators to look beyond immediate performance metrics.
Lenovo has set a target of a 50% improvement in energy efficiency of servers in the data centre by 2029/2030, using innovations in cooling technology, workload optimisation and circular economy practices to meet growing compute demands without proportional increases in environmental impact.
Warnings against optimising data centre components in isolation
The scale of projected energy growth demands structural changes now rather than incremental adjustments later. James says operators must factor in everything from growing energy demands and water use to how equipment is reused and recycled.
"The main change is to take a system-level view of the impact of data centres, factoring in everything from growing energy demands and water use to how equipment is reused and recycled," James says. "That means making design, cooling and workload decisions that remain efficient under sustained demand, rather than just current loads."
He argues that organisations should also take a wider view of how digital infrastructure is integrated into society and systems such as energy and water supply.
"Without this shift, operators risk optimising individual components in isolation while locking in inefficiencies that become increasingly difficult and expensive to reverse," James explains.
Treating facilities as ecosystems requires coordination across power and cooling
The ecosystem James describes approach centres on three areas: physical efficiency of data centres, workload efficiency and circularity in equipment reuse and recycling. But James says data centres interact with power and water supplies in ways that demand careful planning.
"By thinking carefully about how these interactions take place, and focusing on the fundamentals, data centre operators can make explicit sustainability trade-offs,” James says. These include factors such as “balancing higher compute density against the energy and cooling overheads it creates, or prioritising lower emissions and resource efficiency over short-term performance gains, rather than addressing each in isolation,” he adds.
James notes this requires closer coordination between facilities teams, IT operations and energy providers, so that decisions about capacity, cooling and workload placement are made with a shared understanding of their long-term sustainability impact.
Lenovo Neptune warm water cooling delivers energy reductions
Physical efficiency interventions vary by facility, but cooling represents a significant opportunity. In many data centres, cooling uses a significant amount of energy and as much as one third can be used to drive fans in the servers alone. As rack power density grows, liquid cooling is becoming standard practice.
James says Lenovo's Neptune technology uses warm water cooling to drive measurable reductions in energy use.
"In suitable environments, Neptune can reduce energy by up to 40% and removes the need to chill water," James says.
Neptune-enabled supercomputers have been ranked highly in the Green500 and Top500 lists, delivering high-density computing while reducing facility power use. James notes that warm water cooling has become necessary as traditional air cooling can no longer match the power demands of dense AI and HPC systems. With no need for chillers and easy integration, such systems support energy efficiency targets and long-term emissions reduction.
- 94,000 metric tonnes – Lenovo customer products the company has reused or recycled since 2020
- 50% – Lenovo’s goal for energy efficiency improvements in its data centre servers by 2029/30
- 40% – Potential energy reductions enabled by Lenovo Neptune technology
The technology takes a step further than standard liquid cooling because fluids are more efficient than air at removing heat. James says Lenovo consultants can add value to customer decisions because there is no one-size-fits-all design, with many factors to consider, including the load being generated by the IT equipment.
Workload efficiency represents untapped lever for AI environments
While physical infrastructure improvements attract attention, workload efficiency remains underutilised as a sustainability strategy. James argues it can be a hugely important component for data centre operators.
"Workload efficiency is all too often overlooked as a sustainability strategy, but it can be a hugely important component, even in AI-heavy environments," James says. "Data centre operators should adopt a 'one workload at a time' approach to drive efficiency, building momentum in this way towards wider systemic change."
He says that as far as possible, every watt consumed should be tied to productive computing output.
Virtualisation has a role to play here, helping to minimise idle capacity and maximise usage by allowing multiple applications to run on the same server. Updating older systems can also help, with newer architectures delivering more performance at a lower energy cost.
James notes that 'as a service' approaches can help data centre operators upgrade to newer hardware with lower upfront costs and increase efficiency by only installing what is required, reducing redundant capacity.
Asset recovery and Lenovo Certified Refurbished extend hardware lifecycles
Circularity is growing in importance in the data centre ecosystem, with companies and countries setting targets on waste generation. Asset recovery services help ensure assets are reused at end of life, alongside refurbishment services which extend hardware lifecycles.
Lenovo works across 43 markets globally to ensure devices are recycled and repaired responsibly, with 94,000 metric tonnes of products from customers recycled or reused since 2020. James says programmes such as Lenovo Certified Refurbished can help data centre operators extend hardware lifecycles without compromise.
"Devices are inspected, restored and tested through a certified refurbishment process, with renewed warranty and security assurances," James says.
"When done properly, refurbishment and reuse can also reduce exposure to supply-chain disruption and embodied carbon, while giving operators greater flexibility in how and when they refresh their infrastructure."
Certified Lenovo standards ensure devices are reliable, secure and ready to work, addressing concerns about performance or resilience when extending equipment life.
Future designs integrate facilities into communities with agentic AI optimisation
The future of data centre design could see facilities integrated more closely into communities, including everything from spas to living spaces, helping to reuse the heat generated by the equipment.
Lenovo partnered with engineering firm AKT II and architects Mamou-Mani to explore how data centres might be designed differently as operators contend with tighter power constraints, sustainability targets and higher-density AI workloads.
The designs showcased ideas ranging from data centres buried deep in bunkers to minimise land usage, to riverside data centres woven into urban communities.
Referencing Lenovo research The Data Center of the Future, James notes that 45% of IT leaders said that their current infrastructure does not support carbon reduction or energy reduction goals – highlighting the need for more sustainable and power optimising technology.
Future data centre designs should also prioritise the use of technologies that can measure energy efficiency in real time, such as advanced modelling and energy analytics. Agentic AI is expected to double workforce productivity worldwide by 2027 and can make an impact on the data centre too.
"If given the required intelligence, it can be deployed both measure efficiency and optimise workflows in real time," James says. "If AI growth continues on its current trajectory, sustainability will only be achievable if it is built into how data centres are designed, operated and upgraded, rather than treated as a parallel initiative."

