The Cloud Looming Over Your Neighbourhood

Don’t you hate it when you’re trying to work but all you hear is your neighbour’s leaf-blower? Can you imagine if you could hear it all day and night, every day and night? Well, that is the case for some residents in communities close to data centre developments. Taking the AI glamour out of data centres, they are industrial facilities: large, energy-intensive, water-dependent and increasingly clustered into hyperscale hubs. While the benefits of data centres are global – powering AI models and cloud services – the consequences of their facilities are intensely local.

The true energy footprint
A conventional data centre draws around 10 to 25 megawatts (MW) of power, roughly equal to the energy use of 10,000 to 25,000 households. Hyperscale AI‑oriented facilities can reach capacities of 100MW, matching the annual electricity consumption of around 100,000 homes. As most new developments are AI‑focused, they place unprecedented strain on existing power grids. This rising demand forces utilities to upgrade infrastructure for both generating power and transporting it: expensive work that ultimately flows back to residents through higher rates and taxes. To manage grid stress, many data centres deploy fleets of gas or diesel back-up generators – meant to be used sparingly. However, they often run more frequently than permitted and contribute to local air pollution. One high‑profile example is xAI’s Colossus 1 facility, which at one point operated 35 gas turbine generators, some without air quality permits. xAI battled legislators for eighteen months over these turbines, claiming that they were exempt from requirements for air quality permits, with the US EPA finally ruling that they were not exempt in January 2026. While this case resulted in a regulatory win for the community, many similar incidents do not.

Water: the hidden cost
Beyond electricity, data centres are also enormous consumers of water. Cooling systems alone can use vast volumes, all of which evaporates and must be replenished. A relatively small 1MW data centre can require 26 million litres of water annually, just for cooling. Even more significant is the water indirectly consumed through electricity generation, particularly in regions reliant on thermoelectric power plants, where water is heated to create steam and drive turbines. This water often comes from shared municipal or aquifer sources. The Colossus facility, for example, uses 3 to 18 million litres of water per day from the Memphis Sand Aquifer, which also provides water to residents. While this may be manageable in water‑abundant regions, high‑volume withdrawals can still stress groundwater systems. In drought‑prone areas – especially the western US where many data centres are located – the impact is far more acute, and can exacerbate competition for increasingly scarce water resources.

Silent but harmful: noise pollution
The physical footprint of data centres also extends into sound. Large fans, cooling towers, compressors and on‑site generators operate nearly 24/7 and create persistent noise, even when it isn’t always consciously heard. The constant audible hum that can be heard from some data centres is a nuisance of its own, turning once-quiet neighbourhoods into buzzing beehives (that don’t even provide honey). But more concerning is the non-audible low-frequency noise (LFN), sometimes with infrasound components: a number of studies suggest an association between LFN and adverse health effects, from annoyance and concentration problems to lower sleep quality and mood effects.

The path forward
Some data centres are being built near residential communities and even schools. As such, there’s a massive need for technology firms, data operators and governments to address these hyper-local impacts through technological investments and legislation. Software solutions, such as energy management and unified data centre management solutions, can help provide continuous monitoring and optimization for energy and water consumption to reduce strain on local infrastructure. Operators also need to find ways to better support their power capacity requirements beyond investing in higher-efficiency strategies and on-site infrastructure. Research from Rewiring America suggests that hyperscalers could meet their power capacity needs by paying for household electric upgrades, such as heat pumps, battery storage and rooftop solar panels.

Technology firms need to consider the impact on local resources and infrastructure when selecting new development sites and increase reporting transparency, such as publishing water and power usage effectiveness; Microsoft and Facebook already publish aggregated water data. To take it a step further, these organizations should also commit to noise monitoring (including LFN) and build community benefit agreements that address local costs – not just global sustainability slogans. Governments have a central role to play in this and in some places, such as Ireland, are stepping in where impact of data centre growth has been high.

For more insights into how data centres can account for their environmental impact, check out Market Insight: Accounting For AI’s Rising Emissions, and for further reading on technology offerings for data centres, see Smart Innovators: Unified Data Centre Management Platforms (2026).