In this article, George Bryden, Process Consultant at Bryden Wood, explores why data centre sustainability reporting needs to go beyond PUE – and how clearer disclosure on water, noise and carbon could help operators earn the trust of the communities around them.
Do data centres have a perception problem?
From the outside, a data centre gives little away. It is a black box in more ways than one: a sealed building whose demands on power, water and the surrounding soundscape are largely invisible to the people who live next to it. That mattered less when data centres were background infrastructure. It matters a great deal now.
Goldman Sachs expects US data centre power demand to reach around 66 GW by 2027, more than double the 31 GW of 2025.
As these facilities become the physical face of AI, concern about their environmental and community impact is growing just as quickly – and, left unaddressed, it could make the next project harder to deliver. It need not.
Greater transparency on water use, and clearer communication about noise and carbon, would help operators build trust with the communities they serve.
Two factors are driving this need for better communication.
First, the interconnectedness of AI and data centres ultimately makes data centres the physical embodiment of AI and its associated concerns about technological change. AI’s growing integration into our way of life has turned data centres into lightning rods for public outcry and rejection. This can be seen in the results of a Gallup poll from May this year, which found 7 in 10 Americans were against the construction of AI data centres near their homes.
Second is the growing awareness of how resource-heavy operating data centres can be. Among poll respondents against the development of data centres near to their homes, about a fifth mentioned specific fears over water and energy usage, and a sixth raised concerns over noise, air, and water pollution. Right now, there is a need to de-escalate the rhetoric around data centres and fix the disconnect between the sustainability information operators currently report and the information that regulators and local communities increasingly expect to see.
It is no longer enough to just measure a facility on only a Power Usage Effectiveness (PUE) metric. Operators should be focusing on a more holistic story, incorporating data for water, noise, and carbon output. Building credibility should be incumbent on each individual facility, disclosing broader types and more depth of operating data so that ‘closed loop’ reporting can be diminished, and we can manage these facilities more responsibly. The scale of data centre production and use is reaching levels where this cannot remain a discussion and must transition towards setting targets for measurably better community cohesion.
Water
WUE (Water Usage Effectiveness) is sufficient for benchmarking but ultimately lacks in terms of revealing how much of the water used is potable vs non-potable. Potability is a standard of how clean and safe the water used is. Uptime Institute’s 2024 survey found only about 40% of operators collect water-use data for both potable and non-potable sources, so disclosing the percentages of each should become normalised in the reporting of operators.
Alongside water quality, measuring water stress and seasonal variation in water throughput are also becoming more criticalmetrics to evaluate. Data centre water usage has become a significant point of debate and as we move towards hotter and more drought-laden summers, will become a much more sensitive topic. Absolute consumption figures, such as a data centre using a certain number of litres of water per day, remain useful for illustrating scale and understanding potential pressure on local resources. However, these figures alone provide an incomplete picture. More constructive reporting frameworks should place equal emphasis on factors such as water source, potable versus non-potable use, local water availability, seasonal demand fluctuations, and wastewater management. A facility consuming large volumes of recycled water in a water-abundant region presents a very different sustainability profile from one drawing potable water in a water-stressed catchment.
Whilst the US has the most data centres in the world, the UK has the third most, ahead of China in fourth place, and the UK is continuing to expand capacity. Water UK, the trade association for the UK's water industry, claimed this year that data centre water demand is projected to increase from 6.6 million litres per day to 19.8 million litres per day by 2030. These figures are valuable because they demonstrate the potential scale of future demand, equivalent to around 60,000 households, with summer consumption potentially increasing eightfold to 77 million litres per day. However, the sustainability implications of these volumes cannot be understood without additional context on where the water comes from, whether it is potable or recycled, and the degree of local water stress.
It doesn’t take a crystal ball to see that operators need to set expectations ahead of public unease during more water strained periods and the best course of action is building credibility now through clearer and more consistent reportingin line with environmentally sensitive design. Setting targets could alleviate public worry and could include committing to use more non-potable water where possible and adopting greater sustainability policies like iterating on Singapore’s greywater recycling policy where businesses that use more than 60 megalitres of water a year are required to recycle 50% of the water they use.
Noise Pollution
Data centre noise pollution is a characteristic of data centres that seems to be rarely talked about. That is, until one is built too near your community.
In designs and planning where noise is inadequately addressed, a persistent tonal whine can become part of everyday life for nearby residents, raising issues about both their living environment and the value of their homes. Such concerns deserve a substantive response and not simply reassurance. According to the Environmental and Energy Study Institute, these sites can reach 96 decibels and can be heard up to 800m away. This is a factor we must discuss more openly, consider more carefully in our designs, and, as responsible designers, mitigate.
Compliant design can still generate complaints if sound modelling doesn’t properly account for nighttime background levels, tonality, topography, and wind direction. We also must not treat all data centre noise as equal and instead should distinguish between the routine plant-cooling hum and the generator load testing. Alongside this, operators should publish the predicted and measured sound levels at the nearest homes and where people actually experience it, and not just at the boundary of the site.
Credibility for noise pollution is imperative, because losing community trust could lead to stricter policies governing permits, causing friction in future site selection and negotiations with local stakeholders. If policies on tolerated noise pollution tighten after public pushback, additional costs could be required to comply for land buyouts or retrofittingaround the facilities to comply with stricter rules. And the impact of all this on operators? Cost. We know that balancing being sustainable now and cost efficient is the challenge, but this is a fast-developing market and as demand grows, so too will be the discourse surrounding their operation.
Carbon
From a wider perspective, industry average PUE has plateaued in contrast to AI causing significant hikes to rack densities, and this means that the challenge to cooling is the number one priority for operators dealing with greaterthroughput. Balancing energy and water usage is a difficult equilibrium to manage and often leads to cooling choices trading lower water usage for higher energy use and costs. Embodied carbon results are largely defined by early structural and operating decisions and therefore, value is derived in designing facilities to be flexible to changes in process, allowing for the extension of a building’s life beyond its current IT strategy. This should not be treated as aprocurement afterthought. Proven measures include extending server lifespan through improving durability and operating conditions, alongside design-led reductions in embodied carbon associated with cooling systems, cabling, and pipework through material efficiency and system rationalisation. Emerging approaches such as server circularity, refurbishment, and reuse offer significant potential to reduce lifecycle embodied carbon but need alignment across design, procurement, and operational models. As a result, improved reporting on embodied carbon should focus on metrics that not only discuss output but also discuss a site’s situation, for example facility restraints, give context on decisions made in the design phase, and the measures taken to diminish the effects of those choices on the environment and local community.
Designed to belong
Designing any facility requires both optimising the building and its processes and integrating them with environmental objectives to minimise impact. Given the additionalperception challenges facing data centres, making the effort to maintain community trust is a worthwhile investment. As these facilities become more visible and embedded in their surroundings, their long-term viability will increasingly depend on social acceptance and environmental integration. They are not isolated assets; they are infrastructure. Better reporting does not mean producing a constant waterfall of data. It means prioritising relevant information and explaining the context in which each facility operates. When design, operations, and environmental context are aligned, sustainability becomes an outcome of the design process rather than a competing objective. A more structured framework for measuring and reporting performance would make discussions about resource use and emissions more constructive and demonstrate that operators intend to participate in the communities around their facilities, not merely sit alongside them.
