The rapid expansion of the AI industry has raised concerns about the environmental impact of data centres, with a new forecast suggesting that the industry may not meet net zero targets by 2030. Fengqi You and his team at Cornell University in New York modeled the energy, water, and carbon usage of leading AI servers by 2030, considering different growth scenarios and potential data centre locations in the United States. While some tech giants like Google, Microsoft, and Meta have set net zero goals by 2030, not every AI company has followed suit.
The projections indicate that US AI server buildout could require between 731 million and 1.125 billion additional cubic meters of water by 2030, with carbon emissions ranging from 24 to 44 million tonnes per year. The forecast is dependent on the rate of AI demand growth, server production capacity, and data centre locations. To mitigate the impact, the researchers suggest focusing on location, decarbonizing energy supplies, and enhancing data centre efficiency. These strategies could potentially reduce emissions by 73% and water usage by 86%.
However, public opposition to data centre installations due to environmental concerns could hinder these projections. Residents in states like Virginia, which hosts a significant portion of global data centre capacity, have raised objections to further construction citing water resource depletion and environmental impact. Similar opposition has been seen in other states like Pennsylvania, Texas, and California. Despite successful campaigns against data centres, the actual energy and water usage of rejected projects remains unknown.
Experts in the field have cautiously welcomed the new findings, acknowledging the challenges of predicting AI’s environmental impact in a rapidly evolving industry. There is a call for more transparency and accountability in tracking and reporting energy and compute usage in AI development. The need to invest in renewable energy capacity and consider data centre locations is also emphasized.
In conclusion, the forecast highlights the importance of addressing the environmental impact of AI data centres and the necessity for sustainable practices in the industry. By implementing strategies to reduce emissions, water usage, and increasing transparency, the AI sector can work towards a more environmentally friendly future.
