Data centre build-out may face increasing push-back next year

As the global expansion of data centres continues to surge driven by AI and digital growth, advancements in efficiency and aggressive renewable energy procurement offer potential respite, but local infrastructure constraints and environmental justice concerns threaten to complicate the industry’s future trajectory.

The steady hum of the global internet increasingly masks a more uncomfortable reality: the infrastructure that keeps our digital lives running is becoming one of the fastest‑growing drivers of electricity and water demand. Data centres already consume an estimated one to two percent of global electricity, and the rise of artificial intelligence , with its appetite for constant, heavy computation , is amplifying that pressure, forcing industry, regulators and communities to confront trade‑offs between digital growth and resource limits. ^[1] (EarthTalk), ^[2] (Reuters)

Technical fixes are emerging that promise meaningful reductions in energy use. Academic researchers at the University of Waterloo, working with engineers from Fastly, have developed compact “code kernels” that speed internet traffic and can cut energy use for routing tasks by up to about 30 percent. According to the original report, the developers estimate that broad adoption could translate into substantial electricity savings across the internet ecosystem because most services rely on the same Linux‑based routing infrastructure. ^[1] (EarthTalk)

Hardware advances are another avenue for efficiency gains. Chipmakers such as Nvidia have released more powerful, more energy‑dense processors that can train AI models more quickly and therefore, per task, use less electricity. But those gains are double‑edged: the latest “superchips” pack hundreds of billions of transistors onto a single device and produce enormous heat, increasing cooling loads and, in many facilities, water consumption. Industry data shows major cloud operators relied on tens of billions of litres for cooling in recent years, prompting experiments with non‑water cooling approaches such as direct liquid (oil‑based) cooling to reduce both water and energy footprints. ^[1] (EarthTalk)

At the system level, power sourcing is shifting as firms seek stable, lower‑carbon supplies. Big cloud providers are expanding purchases of wind, solar and battery storage, and some are investing in nuclear energy to obtain firm, carbon‑free baseload power. According to the original report, companies including Google and Microsoft view nuclear as increasingly attractive despite high upfront costs and public concern because it can offer the steady output needed by large data centre campuses. Larger corporate procurements are evident in the marketplace: energy companies are closing multi‑gigawatt clean‑power deals with hyperscalers and pairing new campus builds with integrated generation and storage. ^[1] (EarthTalk), ^[3] (Reuters)

Those corporate and market moves are coinciding with broader, measurable increases in national electricity demand. The U.S. Energy Information Administration forecasts record U.S. power consumption for 2025 and 2026, driven in part by data centres supporting AI and other compute‑intensive activities. The EIA expects U.S. electricity use to rise above the 2024 record level and a modest shift in generation toward more renewables, even as shares for gas, coal and nuclear evolve. The EIA head has also signalled plans to begin formally tracking data‑centre energy use so policymakers and markets can better understand the sector’s footprint. ^[2] (Reuters), ^[7] (Axios)

The growth is straining local infrastructure and prompting contentious planning fights. In London, rapid expansion of data centres is consuming a growing share of city electricity and creating situations where housing developments have been delayed because grid capacity can’t be secured; similar dynamics are playing out in energy‑hungry regions worldwide. In the U.S. and elsewhere, utilities are proposing large capacity builds targeted at serving data centre demand , a move that has sparked debate about who bears the financial risk if expected commercial contracts do not materialise. Georgia Power’s multibillion‑dollar proposal to expand capacity primarily for data centres has become a flashpoint for regulators and ratepayer advocates; state staff have suggested tighter limits on approvals to avoid overbuilding. ^[4] (TechRadar), ^[6] (AP)

Environmental justice advocates warn the burden of rapid data‑centre build‑outs is falling disproportionately on marginalised communities. Reporting and campaign activity document cases where large AI facilities are sited in neighbourhoods already facing pollution, water stress and other climate impacts, prompting legal challenges and calls for stricter oversight. Civil rights groups are organising to highlight both the local health impacts and the uneven distribution of economic benefits from the industry’s growth. According to coverage of these developments, activists argue data centres often offer relatively few local jobs while consuming large volumes of water and power in vulnerable regions. ^[5] (Axios)

Policy responses and market innovation are converging on a few practical levers. Operators can cut the footprint of AI by training and deploying smaller, more efficient models where feasible; adopt software‑level efficiency measures such as the Waterloo code kernels; move to non‑water cooling technologies; and increase long‑term clean‑energy procurement coupled with grid‑modernisation investments. Regulators and utilities can reduce community risk by conditioning approvals on signed power contracts, expanding grid planning horizons, and requiring equitable siting and mitigation measures so local communities are not left to shoulder environmental costs. The EIA’s planned new data collection and revised long‑term outlooks are intended to inform those conversations. ^[1] (EarthTalk), ^[7] (Axios), ^[3] (Reuters)

Individual users also retain influence: reducing unnecessary data use, cutting needless AI queries, and choosing lower‑bandwidth behaviours can cumulatively temper demand growth. The original EarthTalk column argues that while efficiency gains are critical, without demand restraint efficiency often unlocks higher consumption , so personal choices matter alongside corporate and policy action. ^[1] (EarthTalk)

📌 Reference Map:

##Reference Map:

• ^[1] (EarthTalk / eMagazine) – Paragraph 1, Paragraph 2, Paragraph 3, Paragraph 4, Paragraph 8
• ^[2] (Reuters) – Paragraph 1, Paragraph 5
• ^[3] (Reuters) – Paragraph 4, Paragraph 7
• ^[4] (TechRadar) – Paragraph 6
• ^[5] (Axios) – Paragraph 6, Paragraph 7
• ^[6] (Associated Press) – Paragraph 6
• ^[7] (Axios) – Paragraph 5, Paragraph 7

Source: Fuse Wire Services