A new United Nations study ranked data centers in Hong Kong among the world's most carbon-intensive infrastructures. According to the report, this is largely due to the city's electricity being heavily reliant on fossil fuels.
The study titled "The Environmental Cost of AI's Energy Use" examined the carbon, land, and water impacts of the infrastructure supporting artificial intelligence. The UN projects that by 2030, data centers could consume up to 945 terawatt-hours of energy.
Why Does Hong Kong Stand Out?
According to the UN press release, this amount is nearly three times the combined annual electricity consumption of Pakistan, Bangladesh, and Nigeria. The same assessment noted that Indonesia, India, and Hong Kong have power grids with carbon footprints well above the global average.
The report states that in Hong Kong, 67% of energy comes from fossil fuels, 32% from nuclear, and only 1% from renewable sources. This energy mix further increases the electricity-related emissions of data centers.
In comparison, countries with similar living standards such as the US, Germany, and Italy have carbon intensities below the global average. The US is 18%, Germany 24%, and Italy 32% lower, respectively.
Water and Land Impact Also Assessed
The study draws attention not only to carbon emissions but also to water and land use by data centers. According to the UN, AI-driven water consumption could equal the basic annual domestic needs of 1.3 billion people by the end of the decade.
The report highlights that the land footprint could exceed 14,500 square kilometers, roughly twice the size of the Jakarta metropolitan area. However, Hong Kong is among the lightest users in terms of land and water consumption.
Hong Kong’s Digital Infrastructure is Expanding
As a trade and logistics hub, Hong Kong hosts about 300 internet service providers. The city's telecommunications networks connect to 12 external submarine optical fiber cable systems, with new lines under construction.
The report also notes that the government is building a new 110,000-square-meter data facility at Sandy Ridge, with 90% dedicated to data centers. This investment indicates ongoing growth in the city's digital infrastructure.
The study states that daily AI operations account for 80 to 90% of total energy demand, while model training represents a smaller share. It mentions that ChatGPT processes about 2.5 billion requests daily, and image generation requires a thousand times more energy than a simple text query.
The report also shared that DeepSeek, launched in January 2025, reached over 20 million daily active users within three weeks and is expected to have around 125 million monthly active users by mid-2025.
Data center operators in Hong Kong say they are working to improve energy efficiency, reduce electricity consumption, and limit environmental impacts. Existing regulations in the city govern energy efficiency in buildings and fresh water use in cooling towers.
The Green Data Centers Implementation Guide, prepared by the Electrical and Mechanical Services Department, outlines steps for efficient design, procurement, operation, and disposal processes. The guide also encourages the use of assessment tools to measure environmental impact.