United Nations warns AI could double global electricity use by 2030

A United Nations report released on June 3 projects that artificial‑intelligence (AI) systems will consume about 3 % of worldwide electricity by 2030, roughly doubling current demand. The study estimates AI data centres will require 9.3 trillion litres of water annually and could occupy land equivalent to ten times the size of Mexico City, while producing up to 2.5 million tonnes of e‑waste each year. It cites the “Jevons paradox,” where efficiency gains lead to higher overall consumption, and calls for responsible AI development to limit climate impacts.

A separate business‑focused analysis highlights that many AI providers, such as Anthropic, do not disclose detailed emissions data, complicating corporate Scope 3 reporting. Using spend‑based proxies, a £10,000 annual spend on Anthropic’s Claude tool is estimated to generate about 1.18 tonnes of CO₂e, compared with roughly 0.63 tonnes for Microsoft Copilot and 0.51 tonnes for Google Gemini. While per‑query emissions are modest – measured in grams of CO₂e – the cumulative effect of widespread AI adoption could become a notable source of corporate emissions, prompting firms to consider offset costs and reporting transparency.