News Articles

(SCMP, 22 June 2026) The environmental impact of artificial intelligence is widely known. A newly released United Nations report has now exposed its carbon, water and land footprints. Among 20 of the world’s largest data centre hubs, Hong Kong has the third-highest carbon-intensive grids due to its heavy reliance on fossil fuels. The carbon footprint of its grid was 43 per cent above the global average. 

More than 11,800 AI data centres are operating globally, according to an industry estimate in 2024. Hong Kong has 61 data centres in operation. It is expected to have 81 built by the end of 2030, including the largest, the Sandy Ridge data facility cluster.

We probably cannot avoid the accelerating growth of AI demand, but there are ways to minimise its threat to our warming planet. The Hong Kong government and AI-related businesses must reduce our carbon footprint by enhancing the energy and water efficiency of data centres.

First, we must mandate power usage effectiveness (PUE) standards for existing and coming data centres. The lower the PUE number, the higher the energy efficiency. Mandatory disclosure of PUE will drive improvements in performance.

Second, we need innovation. Data centres can be put in locations where they can make good use of natural cooling. In early June, for example, China launched the world’s first commercial-scale underwater data centre powered by an offshore wind farm off the Shanghai coast.

Third, there should be continuous public education on AI’s environmental impact. A short query posted to AI generally consumes 2.9 watt-hours of electricity, which is almost 10 times more than the energy used for the traditional internet search. Understanding the impact will help us make better decisions.

Edwin Lau Che-feng, founder, The Green Earth

South China Morning Post, Data centres must be made more energy efficient