Introduction
Artificial intelligence (AI) has revolutionized our lives, from virtual assistants to recommendation systems. However, behind this innovation lies a massive consumption of resources, especially water , that often goes unnoticed. This article explores how AI and data centers contribute to global water stress.
1. Water: the essential coolant of the digital age
Data centers, which house servers for processing and storing information, generate enormous amounts of heat. To prevent overheating, cooling systems are used that consume large volumes of water. For example, it is estimated that training AI models like GPT-3 may have used up to 700,000 liters of water, equivalent to the water footprint of manufacturing 320 Tesla electric vehicles.
2. It's not just AI: water consumption in traditional data centers
Water consumption isn't limited to AI. Enterprise data centers also require large quantities of water for their operations. In 2023, a 14% increase in their water consumption was reported, primarily used for cooling the data centers of the largest company in the sector. In regions like Virginia, USA, data center water consumption has raised concerns about water stress.
3. Myths and realities about digital water consumption
Contrary to popular belief, individual use of digital services has a minor impact on water consumption. The real challenge lies in the infrastructure that supports these services. For example, each ChatGPT query can use up to 25 ml of water, but training AI models and maintaining data centers account for much higher consumption. In some cases, this is potable water; in others, recycled or reclaimed water is used. However, its use still places significant pressure on local water resources, especially in regions experiencing drought or overexploitation.
4. Towards a more sustainable technology
Some companies are taking steps to reduce their water footprint, and some are already experimenting with waterless cooling systems, such as boiling liquid cooling, to meet their commitment to replenish more water than they consume. However, much remains to be done to achieve truly sustainable technology.
Conclusion
Technological innovation must go hand in hand with environmental responsibility. It is crucial that companies and users become aware of the water impact of technology and work together to develop more sustainable solutions.