Hong Kong's data centres have come under scrutiny for their significant environmental impact, as highlighted by a recent UN study. The report, titled "Environmental Cost of AI's Energy Use," sheds light on the global challenges posed by the infrastructure powering artificial intelligence. One of the key findings is that Hong Kong's data centres are among the most carbon-intensive in the world, primarily due to the city's reliance on fossil fuels for energy generation.
What makes this particularly fascinating is the contrast it presents. While Hong Kong may be known for its technological advancements and thriving digital economy, the environmental cost of sustaining these industries is a growing concern. The report reveals that by 2030, data centres could consume a staggering 945 terawatt-hours of electricity, which is equivalent to the combined annual electricity use of several populous countries. This raises a deeper question about the sustainability of our digital future.
In my opinion, the study's focus on the carbon footprint of AI's energy use is a crucial step towards understanding and addressing the environmental challenges we face. It's not just about the energy consumption of data centres; it's about the broader implications for our planet. The report also highlights the water and land footprints associated with cooling these heat-intensive facilities, emphasizing the need for a holistic approach to sustainability.
One detail that I find especially interesting is Hong Kong's ranking as one of the lightest consumers of water and land. This may seem counterintuitive, given the city's high carbon intensity. However, the reason lies in its energy mix, which relies less on hydropower and renewable energy sources that require substantial water and land resources. This highlights the complex interplay between different environmental factors and the need for a balanced approach to sustainable development.
As a trade and logistics hub, Hong Kong's role in hosting data centres is significant. With over 300 internet service providers and extensive telecommunication networks, the city is an attractive location for data storage and processing. However, the government's plans to build a new data facility in Sandy Ridge, with a substantial portion dedicated to data centres, raise questions about the long-term environmental sustainability of this strategy.
The UN report also emphasizes the daily use of AI models as a major contributor to energy demand, rather than just model training. This insight challenges our understanding of AI's environmental impact and highlights the need for more efficient practices. The case of ChatGPT, processing billions of prompts daily, serves as a stark reminder of the energy-intensive nature of AI applications.
From my perspective, the study's findings should serve as a wake-up call for policymakers, businesses, and individuals alike. It's time to rethink our approach to AI and its infrastructure, exploring more sustainable alternatives. The environmental cost of AI's energy use is a global challenge that requires collaborative efforts and innovative solutions. By addressing these issues head-on, we can work towards a more sustainable and responsible digital future.
