Experts Share: The Environmental Cost Of Data Centres Part 2

When we ran the first part of this discussion, we had a flood of industry experts writing in with such sharp, considered takes that a single article simply couldn’t hold them all. If anything, that tells you just how much this subject is weighing on the people who work closest to it.

And that’s rather the point really. The environmental cost of data centres has gone from a niche concern to one of the defining infrastructure debates of our time. With AI pushing demand to dizzying new heights, the stakes continue to climb.

The big question is still hanging over it all: what can the industry realistically do to clean up its act? As before, we’ll let the experts take it from here.

 

Our Experts

 

• Aoife Foley: IEEE Senior Member & Chair in Net Zero Infrastructure, The University of Manchester
• Ashlesa Mohapatra: Senior Quality Engineer at ServerLIFT
• Alex MacColl: PM EMEA at DataMove
• James Owen: Director at Click Intelligence
• Matt Thompson: Managing Director at Airsys (UK) Limited
• Keith Hills: Founding Director at Eco Green Group
• Luthando Bonani: Founder & Aspiring Tech Entrepreneur
• Elena Popkova: Professor at Tashkent State University of Economics
• Helen Munro: Head of Environment & Sustainability at Pulsant
• Polina Kuperman: Community & Social Impact Lead at Krystal
• Matt Evans: CEO of Apx Data Centre Solutions
• Scott Geisinger: Co-Founder & COO of Nexus Quantum
• Gary Watson: Managing Director of Stellanor
• Joe Higgins: Group Chief Executive Officer of Envirogen
• Alan Brejnholt: Affiliate Professor of Corporate Social Responsibility and International Management at ESCP Business School
• Greg Roberts: Data Centre Sector Lead at Ramboll

 

Aoife Foley, IEEE Senior Member & Chair in Net Zero Infrastructure, The University of Manchester

 

 

“The International Energy Agency expects data centre electricity demand to more than double by 2030, reaching around 945 terawatt-hours, slightly above Japan’s annual use. This not only indicates greater energy demand but also underscores the importance of balancing digital growth with sustainable energy practices.

Although it is impossible to calculate precisely, the entire ICT sector is estimated to account for about 1.4 per cent of CO2 emissions globally. Infrastructure and operations leaders have a responsibility here and need to consider the unnecessary waste associated with data storage and commit to generating power from more renewable sources. By eliminating unstructured data and improving operational efficiency, organisations can reduce maintenance costs and improve regulatory compliance.

AI workloads consume significantly more energy than traditional cloud computing tasks, and although hyperscale operators are investing in renewable energy to soften the impact, this alone is not enough.

Cooling innovations such as liquid immersion and direct-to-chip systems add further efficiencies, yet they still address symptoms rather than the deeper inefficiencies in model design and compute intensity – and growing dependence on water-based cooling places additional pressure on local resources. These impacts can be reduced through smarter model optimisation and a closer alignment between data centre strategy and regional renewable generation.”

 

Ashlesa Mohapatra, Senior Quality Engineer at ServerLIFT

 

 

“As data centres have grown, one of the less visible but important environmental concerns has been how much water is used for cooling. A lot of traditional cooling systems rely on evaporative cooling or water-based chillers, which can end up using large volumes of freshwater, especially in hotter regions where systems are working harder for longer. This becomes a real concern in water-stressed areas, where the same water is also needed for communities and agriculture.

Because of this, the industry has been actively shifting how cooling is designed and managed.

One of the biggest changes has been moving toward more closed-loop and air-based cooling systems, where water use is reduced or recycled instead of being continuously consumed. In many cases, companies are also using free-air cooling, where outside air is used to cool equipment when weather conditions allow, which reduces the need for water altogether for part of the year.

There’s also been a lot of development in liquid cooling at the server level, like direct-to-chip or immersion cooling. These systems are more efficient at removing heat, which means the overall cooling demand goes down, and with it, both energy and water usage.

On top of the technical changes, companies are also thinking more carefully about where data centres are actually built. Many are avoiding high water-stress regions or choosing locations where they can use recycled or non-drinking water sources instead of freshwater. Some are even working with local utilities to reuse treated wastewater for cooling.

Overall, what’s changing is not just the technology, but the mindset. There’s a clear move toward designing data centres that are less dependent on freshwater and more aligned with local environmental limits.”

 

Alex MacColl, PM EMEA at DataMove

 

 

“The conversation surrounding data centres and the environment is usually targeted purely at power and cooling. A huge part of the footprint is missed – the hardware itself – and the vast resources locked up in their manufacture.

The numbers are eye watering. To manufacture an average 1U server involves a huge amount of energy – emissions estimated at 1.5-1.8 tonnes of CO2 before it’s even had a power cable plugged in. The integrated circuit boards require an average of 8,000 litres of water in the manufacturing process and a chip fabrication plant can consume more than 20 million litres of water a day – often in areas that face supply issues.

This is a major point the industry and its wider commentators overlook. The environmental cost is partly spent the moment the hardware exists. Hardware is often prematurely replaced when relocations / consolidations take place as most decisions are based on 3, 5 or 7 year budget plans. Writing off all this embodied carbon and material usage is a major problem – regardless of how green the new facility or energy source might be.

In our work relocating infrastructure and supporting data centre facilities, we routinely see a lack of hardware sustainability and actively promote relocation and redeployment, or erasure and second life use with our customers.”

 

James Owen, Director at Click Intelligence

 

 

“When people think about how they use the internet – it could be sending an email, using AI, using streaming platforms like Netflix, or shopping online – they rarely think about what is physically making it all possible. The reality is that data centres sit behind almost every digital interaction we have. As we increase our reliance on technology, so does the demand we place on these facilities.

Data centres consume significant amounts of energy; there’s no denying it, and the industry has had to confront that head-on. However, I think it’s crucial that the conversation doesn’t just stop at the problem. Across the sector, we’re seeing substantial investment in renewable energy, smarter cooling systems, and more efficient hardware that can deliver greater computing power with lower energy consumption.

As someone working in digital marketing, I’ve also noticed that businesses are becoming much more conscious of the footprint they are leaving. Important questions around sustainability are no longer reserved for manufacturers or logistics companies. Organisations are increasingly asking how their websites, cloud services, and digital operations can be run more responsibly.

The challenge moving forward is balancing innovation with sustainability. Technologies like AI are creating exciting opportunities, but they also increase demand for computing resources. The good news is that environmental responsibility is now a major consideration for data centre operators, technology providers, and the businesses that rely on them. Progress is being made, and collaboration across the industry will be key to ensuring that our digital future is also a more sustainable one.”

 

Matt Thompson, Managing Director at Airsys (UK) Limited

 

 

“Our appetite for data shows no signs of slowing. From streaming and social media to AI-driven services, the demand on data centre infrastructure grows daily and so does the associated carbon emissions — and with it, the environmental conversation intensifies.

It’s worth noting that data centres, despite their rapid growth, still represent a relatively small share of global power consumption and associated carbon emissions, compared to heavy manufacturing and industrial processing. But that’s no licence for complacency. Operators have a genuine responsibility to act, whether through delivering new, energy-efficient builds or modernising existing facilities with free cooling solutions and improved airflow management in the white space to meaningfully reduce carbon emissions.

Water consumption deserves equal attention. Water-based technologies, such as evaporative cooling, has rightly drawn scrutiny over the years, trading electricity for water, placing enormous stress on local resources. Although its adoption has slowed as the industry seeks alternative solutions.

A more pressing focus now is unlocking power within existing sites and applying that unlocked power to additional compute requirements. Improving Power Compute Effectiveness (PCE) through highly efficient, non-compressor based cooling technologies is key to achieving this.

Ultimately, industry effort alone isn’t enough. Government regulation must play its part — and encouragingly, Europe is leading the way, driving change through data centre efficiency directives and the transition to low GWP refrigerants. That combination of operator accountability and regulatory ambition is what meaningful progress looks like.”

 

Keith Hills, Founding Director at Eco Green Group

 

 

“Artificial intelligence is transforming how we live and work, but the physical infrastructure supporting this cannot come at the expense of the natural environment. The challenge for our industry is ensuring that data centres become assets for biodiversity and climate resilience rather than isolated industrial developments.

As demand for AI and computing accelerates, we have an opportunity to redefine what critical infrastructure looks like. Green infrastructure is becoming a fundamental component of responsible development.The most successful data centres of the future will combine technological performance with environmental stewardship.

At Eco Green Group, we’re championing solutions including bio-solar roofs, blue roofs and living walls that help improve water management and create more resilient developments.
Green roofs allow developers to maximise the ecological value of a site while supporting planning objectives and long-term sustainability commitments.

Our goal should be to ensure that the growth of digital infrastructure leaves a positive environmental legacy for future generations. The sector has the chance to become a global leader in sustainable infrastructure by integrating renewable energy generation, water management and biodiversity-enhancing design from day one. We’re working to ensure that the buildings powering our digital economy contribute positively to the environment and communities around them.”

 

Luthando Bonani, Founder & Aspiring Tech Entrepreneur

 

 

 

 

“Data centre deployment rate has raised concerns about high natural resource consumption and creates socio-economic pressure for local communities. Renewable energy sector is simply not growing fast enough to meet the needs of AI (2025).

Among environmental impacts, infrastructural strain is an emerging area of concern (2025). Data-centres cause power disruptions, and therefore existing energy infrastructure should be updated – costs which are subsidised by residents through rising utility bills and regular blackouts. In response, extreme proposals have been floated, such as the small-scale modular nuclear reactors. But even when data centres produce their own energy, local grid still provides them with significant backup infrastructure. New units will inevitably be powered by fossil fuel sources, causing further environmental distress.

Six major datacenter hubs – Singapore, Netherlands (Amsterdam), Ireland, Germany (Frankfurt), USA (Virginia) and UK (London) – changed their strategies in response and follow the regulatory patterns (2024):
1. stricter energy accuracy standards may facilitate planning sustainable data centre growth;
2. policy responses may yield better results when paired with multi-regional infrastructural strategies matching IT infrastructure needs with available energy capacity.

Policymaking approaches need to consider different constraints, as digitalisation interferes with other areas of the agenda – such as the environmental one, which should pressure a reevaluation of the national digitalisation strategies.”

 

Helen Munro, Head of Environment & Sustainability at Pulsant

 

 

 

 

“The data centre industry is moving in the right direction, but it’s moving with the business case in mind rather than the climate.

The headlines around AI’s energy appetite speak to this; the IEA predicts that consumption will double to 945 TWh by 2030. But what often gets lost in the message is that this has been a problem for years, rather than a recent occurrence, and some of the solutions are genuinely looking promising. The tech sector accounted for around 40% of all corporate power purchase agreements for renewables signed in 2025. And the pipeline of conditional agreements between data centre operators and small modular reactor projects grew from 25 to 45 gigawatts in less than a year.

The problem is accountability. Too many of these improvements are self-reported and voluntarily provided, and not always verifiable independently. A company is still able to claim green credentials without any meaningful checks being done, and that’s where things need to change.

Regulation is starting to help here. Germany is leading the way, with the country now mandating 100% renewable energy for data centres from 2027, but the rest of the world has catching up to do. Progress without transparency won’t do here, because while the technology exists, the will to be held accountable is still lagging behind.”

Matt Evans, CEO of Apx Data Centre Solutions

 

 

“There are widespread misconceptions about the environmental cost of modern data centres, and the problem largely stems from a lack of awareness. In fact, 9 in 10 (89%) UK adults say they know little or nothing about them.

That lack of understanding breeds fear and mistrust, and the industry hasn’t always helped its cause. We are, in effect, a windowless warehouse, and the deals struck inside them are often kept under wraps. Yet the reality is more mundane than people think. Most modern data centres use less water than a typical golf course.

Behind that facade, people miss the fact that technology companies around the world are investing millions in precision engineering for their data centres. The question they are asking is a simple one: how do we deliver the same energy to meet demand, but do it more efficiently? That shift in mindset has been a catalyst for cooling specialists like us, designing tailored, hyper-efficient systems built around our customers’ exact needs and priorities.

We are part of a conscious movement working to change the wider environmental story, and our approach to cooling is already shifting. As the active load moves towards a higher liquid portion, we are designing for direct-to-chip liquid cooling to meet what today’s customers and end users demand. Crucially though, if we want to break the unhealthy cycle of replacing short-lifecycle equipment, we have to build infrastructure that lasts.”

 

Scott Geisinger, Co-Founder & COO of Nexus Quantum

 

 

“Data centres, as we know, now exact a steep environmental cost. A single hyperscale AI campus can cover 200 to 2,250 acres and several million square feet, and draw hundreds of megawatts, for instance, Meta’s Hyperion campus is scaling toward 5 GW.

Collectively, data centres consumed roughly 415 TWh in 2024 and are on track for around 945 TWh by 2030, close to 3% of global supply, alongside billions of gallons of cooling water and a rising carbon burden. The industry is responding with liquid cooling, efficiency gains and dedicated nuclear and renewable power, but these only slow a curve they cannot bend, because the constraint is fundamentally one of compute.

The attention mechanism that lets large language models turn accessible data into context scales quadratically, doubling the context roughly quadruples the computation, and so supporting it with classical silicon demands ever-increasing land and power.

Nexus tackles the root cause. Our ISR runtime keeps quantum circuits alive long enough to run these workloads. We are already delivering the first quantum attention computation completed on production hardware, a 16.7× fidelity gain on real GPT-2 embeddings. Because quantum hardware draws kilowatts, not megawatts, one ISR-enabled quantum computer could ultimately do the AI work of hundreds of data centres.”

 

Gary Watson, Managing Director of Stellanor

 

 

“Much of the issues surrounding sustainability and the cost to the environment stem from how data centres are built and their power capacity. A major consideration for the industry is how we prioritise extending the life of existing assets through refurbishment rather than defaulting to demolition and rebuild. This is a great way in which we can avoid significant embodied carbon with every site repurposed.

It also matters how we power them, prioritising green electrons – renewable power matched to where and when we draw it, keeps the operational footprint down. Longer term, gains come from investing in high efficiency, future-ready infrastructure such as advanced cooling technologies that improve water use effectiveness, denser racks, and optimisation software that manages power and workloads in real time to deliver greater computing power per watt as AI requirements intensify.

The conversation of sustainability must also extend beyond immediate facilities. Adopting a community centric growth model, where local areas benefit from grid investments, job creation, and heat reuse, is essential for the sector to maintain its social license to operate.

As AI drives demand to levels we’ve never seen, the question for us isn’t whether we build, but how, from the inside out.”

 

Joe Higgins, Group Chief Executive Officer of Envirogen

 

 

“Data centres can have a significant environmental footprint, but advances in water management are helping operators reduce resource consumption and improve long-term resilience. As AI workloads grow and liquid cooling becomes more common, water is emerging as a critical infrastructure consideration alongside power, connectivity and land availability.

Increasingly, we are seeing data centre operators focus less on efficiency metrics alone and more on building long-term water resilience. This includes improving Water Usage Effectiveness (WUE) and adopting circular water strategies that enable the treatment and reuse of non-potable sources such as rainwater, condensate, and treated wastewater, where site conditions allow.

We’re also seeing advanced technologies, including high-recovery reverse osmosis (RO) and direct nanofiltration (dNF), being evaluated as part of wider water reuse and recovery strategies. Some operators are exploring Zero Liquid Discharge (ZLD) approaches to maximise water recovery and minimise environmental discharge, particularly in water-stressed or discharge-limited regions.

Ultimately, we expect water availability and long-term resilience to play an increasing role in where facilities are built and how they scale. Specialist water treatment providers have an important role to play in helping developers and operators integrate water strategy alongside power and cooling decisions from the earliest stages of project design.”

 

Alan Brejnholt, Affiliate Professor of Corporate Social Responsibility and International Management at ESCP Business School

 

 

“Data centres operated by AI companies have become a major sustainability concern because the rapid growth of artificial intelligence requires vast computing power. These facilities consume large amounts of electricity, water, land, and critical minerals, creating environmental externalities including resource depletion, pollution, land degradation, and biodiversity loss.

According to the International Energy Agency (IEA), global data-centre electricity demand is expected to more than double by 2030 due to the expansion of AI. In 2024, data centres consumed approximately 415 TWh of electricity, around 1.5% of global demand, with consumption growing by roughly 12% annually. The IEA also estimates that natural gas and coal could provide over 40% of the additional electricity required by data centres through 2030.

The industry is responding through a combination of decarbonisation and efficiency improvements. In the UK, 500+ AI data centres exist (3rd largest, globally). Major technology firms are purchasing renewable electricity through Power Purchase Agreements (PPAs), investing in more energy-efficient chips and advanced cooling technologies, and exploring low-carbon energy sources such as small modular nuclear reactors.

Companies are also extending server lifetimes through refurbishment and reuse while improving recovery of critical minerals from retired equipment. However, significant challenges remain. AI infrastructure depends on minerals such as lithium, cobalt, and rare earth elements, whose extraction can damage ecosystems. Furthermore, efficiency gains may create rebound effects, where cheaper AI services increase overall demand for computing and electricity, potentially offsetting environmental improvements.”

 

Greg Roberts, Data Centre Sector Lead at Ramboll

 

 

“The environmental challenge posed by data centres is undeniable, with demand for power in particular placing pressure on both net zero ambitions and grid infrastructure.

However, industry leaders are taking steps to address this. For instance, the need to optimise data centre energy efficiency is encouraging new investment in carbon capture, clean power technologies, and Power Purchase Agreements (PPAs), boosting growth in these sectors.

Early-stage ecological surveys and biodiversity metric tools can assist developers to embed biodiversity protections throughout a data centre’s lifecycle, and many have already committed to voluntary biodiversity net gain ahead of this becoming mandatory in November. Water stress can similarly be prevented through water reduction and reuse strategies, as well as avoiding evaporative-based cooling wherever possible.

Prioritising using local and low carbon materials also offers an opportunity to reduce the cost of embodied carbon – Microsoft’s timber data centre and investment in low carbon steel is an example of this approach.

Data centres are critical infrastructure and mitigating their environmental impacts will prove crucial for building a resilient future for the UK. If energy efficiency, circularity, and biodiversity are prioritised alongside growth, data centres could become the heart of sustainable urban communities.”