\n\n
Key developments: AI data centres are consuming massive amounts of electricity, with the United Nations University projecting a 945‑terawatt‑hour annual demand by 2030—roughly three times the combined power use of Bangladesh, Nigeria and Pakistan. Water consumption is also rising; each 20‑50 AI prompts can require about half a litre of water, and larger tasks such as image generation demand even more. New policy pushes, like proposals for net‑zero operation and limits on carbon‑offset reliance, aim to force firms to adopt renewable power, efficiency measures, and heat‑recovery systems. Meanwhile, high‑profile projects such as Meta’s $50 billion, 4,500‑megawatt data centre in Louisiana illustrate the scale of upcoming power needs.
Why it matters: The energy and water footprints of AI threaten to strain grids and water supplies, especially in regions already facing shortages. If unchecked, AI’s resource demand could outweigh its productivity gains and spark public backlash, as seen in debates over large‑scale data‑centre construction.
What to watch next: Expect tighter regulation on data‑centre emissions and water use, greater scrutiny of megaprojects like Meta’s Louisiana facility, and industry shifts toward ultra‑efficient chip designs and renewable‑energy‑focused data‑centre architectures. [[1, 2, 3, 4]]