Around the world, major hyperscalers and colocation providers are pouring capital into data centres in the race to power the next generation of AI and cloud infrastructure. Amazon Web Services (AWS), Microsoft Azure, Google Cloud Platform (GCP), Equinix, and Digital Realty lead the pack, followed closely by Nvidia, Blackstone, and NTT. But the physical infrastructure behind the digital economy comes with a darker side we don’t talk about a lot. Data centres are voracious consumers of water and electricity. They also emit pollutants and impose hidden costs on communities.
In parts of Africa where water scarcity is already acute, the adverse effects of data centres are particularly stark. Many regions lack reliable access to clean water. Yet, to compete globally in digital infrastructure, the continent must attract investment in data centres. Meanwhile, pollution and climate impacts from energy consumption and backup generators exacerbate public health burdens. In Texas, for instance, AI data centres have been cited in reports by the Environmental Integrity Project (EIP) as contributors to air pollution. Under the surface of this expansion lies a truth that cannot be ignored: the very backbone of tomorrow’s intelligence may be eroding today’s environmental, health, and social foundations.
The Hidden Costs of the Digital Infrastructure Boom
Data centres are often framed as clean, modern, “information age” infrastructure, especially when compared to smokestacks or mines. But that narrative obscures serious trade-offs. The following are some of the opportunity costs that Big Tech’s obsession with scaling data centres tends to externalise or minimise.
1. Public Health Impact from Air Pollution
The concentrated energy demand of data centres typically draws from regional power grids, many of which burn fossil fuels. The resulting emissions, such as particulate matter (PM₂.₅), nitrogen oxides, and sulfur oxides, harm local air quality. Backup diesel or gas generators further contribute to pollutant releases.
A study by UC Riverside and Caltech estimates that over a recent five-year span, air pollution from U.S. data centres has imposed more than $5.4 billion in public health costs, encompassing asthma, cardiovascular disease, cancer, and premature death. Other modelling suggests that by 2030, the annual health burden from AI infrastructure alone could reach $20 billion, rivalling classic polluting sectors.
Unfortunately, these burdens are not evenly distributed because economically disadvantaged and marginalised communities are often the ones that reside closest to industrial infrastructure and bear the worst health effects. In Memphis’s Boxtown neighbourhood, residents reported worsening asthma, sleepless nights, and respiratory distress after 35 unpermitted gas turbines were installed at a data centre site, sending nitrogen dioxide levels soaring. (You had already included this anecdote in your original framing.)
2. Stress on Finite Water Resources
Cooling is a fundamental necessity for modern data centres. Whether via evaporative towers, chilled water systems, or other techniques, many facilities rely on clean water. Moreover, a significant share of water demand is “indirect,” associated with electricity generation and supply chains. According to the IEA, about 60% of data centre water consumption is attributed to the energy (offsite) side.
A 2025 study by Carnegie Mellon’s team estimated water usage and efficiency metrics across 41 African countries and found that in some settings, running inference via GPT-4 for a 10-page report consumes up to 60 litres of water, while a lighter model like Llama-3-70B may use ~0.7 litres for the same task. Though those absolute quantities may appear modest, they must be understood in the context that many African regions already experience severe water stress, and water leakage in infrastructure exacerbates scarcity.
Moreover, in conventional data centre design, water use is sometimes quoted as “up to 26 million litres per megawatt per year” on average (for water‐cooled systems). Thus, when data centre operators situate new facilities in already water-stressed zones, they compete with agriculture, community use, sanitation, and ecosystem needs.
3. Carbon Emissions, Climate Risk, and Energy Demand
Data centres are electricity monsters. In 2022, the International Energy Agency estimated that data centres, cryptocurrencies, and AI jointly accounted for nearly 2% of global electricity use, and that share is projected to rise sharply. As more AI models are trained, and demand for large-scale inference grows, the power requirements amplify.
While many companies tout “100% renewable energy” goals, renewable supply cannot always match demand reliably. Some data centres still depend on grid power backed by fossil fuels, or maintain fossil-fuel peaker plants or gas turbines for reliability. FT has documented that some data centre operators in Ireland and elsewhere deploy gas turbines to maintain uptime when renewable supply falters.
Google, for instance, saw its greenhouse gas emissions rise 48% over five years, much of which is attributed to growth in its AI/data centre operations. The risk is that unbridled growth of high-consumption infrastructure may undermine climate mitigation goals and exert pressure on grids to expand fossil endowment.
4. Inequitable Distribution of Benefits and Burdens
That data centres generate enormous profit, value, and innovation is largely uncontested. But who captures the upside is a more contested question.
Often, the economic gains flow to shareholders, engineers, and urban growth zones, while the environmental and health costs are borne by local communities. Public subsidies, tax breaks, water rights, and access to land are frequently negotiated in favour of big tech professionals, sometimes at the expense of other sectors like agriculture, public services, or local industry.
Where regulatory oversight is weak or equity safeguards absent, decision making largely excludes affected communities, reducing local agency and increasing resentment.
5. Land Use and Opportunity Forgone
Massive data centres often require large tracts of land and associated infrastructure (power lines, cooling plants, roads). In fast-growing cities or peri-urban zones, land competes with agriculture, housing, habitat, or space for community amenities.
When priority is granted to digital infrastructure, farms, green belts, parks, social amenities become sidelined.
6. Technological Lock-in and Stranded Asset Risk
Once a hyperscale data centre is built, it is likely to operate for decades. But the architecture, cooling systems, and design choices may become obsolete as technologies evolve (e.g. more efficient chips, localised edge computing, alternative cooling methods). That risks creating stranded assets or requiring complete retrofits.
Furthermore, continued investment in the same patterns (high electricity, high water) can deflect capital away from more sustainable innovations—reduced compute models, chip redesign, or decentralised architectures.
7. Transparency, Accountability, and Democratic Oversight
Many data centre operators are opaque about resource use, energy consumption, emissions, and backup generator usage. Without standardised disclosure frameworks, stakeholders (communities, regulators, researchers) struggle to assess trade-offs.
Calls are growing for mandatory reporting of energy and water consumption tied to AI and data centre operations. For example, the National Engineering Policy Centre (UK) has recommended legal obligations for tech companies to report consumption and emissions as AI continues to scale.
Absent robust oversight, companies may externalise costs onto local populations, eroding public trust and the legitimacy of infrastructure projects.
8. Deferred Innovation and Opportunity Loss
When the majority of capital, engineering talent, and R&D is directed into brute-force scaling of data centre capacity, alternative paths can lose support. Innovations that might reduce resource intensity (e.g. liquid cooling, more efficient AI architectures, neuromorphic computing) may get underfunded.
In effect, we risk locking in a high-cost trajectory rather than exploring a leaner, more sustainable computing future.
Stay ahead in the world of AI, business, and technology by visiting Impact AI News for the latest news and insights that drive global change.
Got a story to share? Pitch it to us at info@impactnews-wire.com and reach the right audience worldwide!
Discover more from Impact AINews
Subscribe to get the latest posts sent to your email.
