Europe’s ambition to expand its AI capabilities is encountering a critical challenge due to increasing resistance against resource-intensive data centers, which require substantial amounts of water for cooling to prevent overheating, reports 24brussels.
This issue poses a significant hurdle, especially as the enhancement of AI technology has become a priority across Europe. The demand for these water resources conflicts with their necessity for agriculture and power generation, further exacerbating tensions amid rising instances of drought.
For instance, French President Emmanuel Macron, who organized an AI Action Summit in Paris in February, aims to position France as a premier destination for AI investment, necessitating the establishment of more data centers. Currently, France is home to 322 data centers and plans to streamline permitting processes to accommodate the more powerful infrastructures supporting the burgeoning field of generative AI.
However, this proposal has met with opposition from political opponents and NGOs concerned about the environmental implications, particularly regarding the significant energy and water consumption associated with these projects.
In May, the French NGO la Quadrature du Net called for a halt on new data center permits pending a public discourse on resource usage, a campaign that has yet to bear fruit. Movements opposing data centers are emerging throughout Europe.
A situation similar to France’s arose in Germany when Elon Musk announced plans for a Tesla electric vehicle plant in the environmentally sensitive lake-filled territories near Berlin. In Dublin, a hub for international tech companies, NGOs have been advocating for a moratorium on new data centers as their demands on the local energy grid escalate.
Spain’s drought-affected Aragón region has also garnered attention this year after local activist group, Tu Nube Seca Mi Rio (Your Cloud Dries Up My River), vigorously protested against the establishment of new data centers in an area where water resources are already critically low.
A trickle of data
Acquiring a comprehensive understanding of the water consumption of data centers is challenging, as companies often report varying levels of detail, and no single authority is responsible for tracking this data.
Google’s latest Environmental Report (for 2024) provides an estimate; the tech giant revealed that it consumed approximately 31 billion liters of water across its global operations (data centers and offices), which equates to irrigating 54 golf courses for a year. Notably, the figures for 2022 and 2023 showed a 17% rise in water consumption, indicating that despite a reduction in workforce, the operational and construction activities of its data centers have increased water usage.
Moreover, it is important to note that Google has not disclosed water usage data for shared third-party data centers that it does not manage directly, complicating accurate assessments of total water consumption.
In North Holland, Microsoft’s facility, which became operational in 2021, reported 75 million liters of water used for cooling, significantly exceeding its predicted 12-20 million liters during the permit application phase—a claim disputed by Microsoft, which indicated that the figure includes water utilized for construction.
Thirsty generative AIs
Dedicated facilities for processing information extensively require cooling mechanisms for the servers they house. In colder regions, these centers can utilize outside air to manage temperatures. However, when external temperatures surpass 25°C—common in Europe—this method alone is insufficient. AI processes, known for their heavy computational demands, further elevate the need for effective cooling systems.
Determining the precise water usage associated with AI applications presents additional challenges. A 2023 study estimated that every 10 to 50 queries directed at OpenAI’s GPT-3 chatbot necessitate the equivalent of a 500ml water bottle, not accounting for the extensive cooling requirements during the model’s initial development.
The EU’s water industry association predicts that the annual water consumption by data centers will surge by 52% by 2030, escalating from 62 million liters in 2024 to 94 million liters by 2030.
In regions grappling with water scarcity, leading cloud providers like AWS, Google, and Microsoft claim to not use water for cooling. This comes at a cost, necessitating increased electricity to operate air-based cooling systems. Simultaneously, these major corporations have promised to reduce their environmental impacts by committing to becoming “water-positive” by 2030, aiming to return more water to the environment than they consume, including initiatives to support wetland restoration. Nevertheless, it remains uncertain whether this additional water will effectively replenish supplies in the same regions from which it is extracted.
Amid these challenges, efficiency in data centers has been rapidly advancing, with innovative technologies enabling newer AI models to operate with significantly lower computational power requirements than their predecessors, consequently reducing waste heat.
Ever-growing AI demand
Despite these advancements, Europe’s AI market is projected to expand at an average annual rate of 26.3% between 2025 and 2031, meaning that while systems may grow more efficient, demand for AI technologies will continue to escalate. This situation has prompted climate advocates to assert that efficiency gains will be insufficient to offset the rising adoption of AI solutions.
“Companies now offer more integrated high-compute services into everything,” stated Matthias Spielkamp, founder of the NGO AlgorithmWatch, highlighting features like the “AI overviews” now prevalent in Google search results. “You can’t escape, you can’t turn them off,” he added, although Google maintains that its resource consumption has not changed since the introduction of these overviews.
The European Commission has ambitious plans to enhance AI development throughout the bloc. Its AI Continent action plan includes funding for five new AI “Gigafactories” aimed at conducting the most resource-intensive AI training tasks within Europe. The initial call for interest in these large-scale data processing facilities received responses from over 70 projects.
Additionally, the EU is evaluating methods to simplify permitting processes for new data centers through the upcoming Cloud and AI Development Act. This initiative is designed to facilitate access to local resources while encouraging innovations to enhance the resource efficiency of AI.
Furthermore, the Commission aims to bolster transparency concerning data centers’ water usage, aiding local authorities in forecasting shortages and refining resource management strategies. As of May 2024, larger data centers are mandated to report their water and energy use to the Commission under the EU’s Energy Efficiency Directive (EED), though no public data has yet been released. Through its Water Resilience Strategy, the Commission seeks to enhance the sustainability ratings of data centers by incorporating clearer criteria for water consumption.
Critics contend that the EED reports may not provide sufficient detail to effectively support local decision-making regarding this increasingly precious resource.
