20 Mar Sustainable data centres and AI water consumption
Artificial Intelligence is the most significant technological revolution of our era, optimising everything from medical diagnostics to global logistics. However, this surge in massive computing power has placed a vital resource in the spotlight: water. For AI to “think”, thousands of servers must operate at peak capacity, generating intense heat that must be mitigated to ensure system stability.
In this context, the transition towards sustainable data centres is no longer just an ethical goal; it is an operational necessity to ensure that digital growth remains compatible with environmental preservation.
The algorithm’s hydrological footprint: Why does AI consume water?
For many users, the link between a chatbot and water usage is not immediately obvious. So, why exactly does AI consume water? The answer lies in cooling. Data centres house thousands of chips that reach critical temperatures while processing data. To prevent system failure, cooling systems are employed, many of which rely on water evaporation to dissipate heat effectively.
Data that demands responsibility
When examining the scale of AI’s water consumption, recent academic studies offer revealing insights: it is estimated that for every 20 to 50 queries made to a large language model (such as ChatGPT), approximately one litre of water is consumed. When multiplied by millions of daily global interactions, the magnitude of the challenge becomes clear. Furthermore, manufacturing a single high-performance chip can require up to 8,000 litres of purified water.
Connectivity as a driver of sustainability
Given these circumstances, the telecommunications industry is not standing still. The key to achieving sustainable data centres lies in geographical and technical optimisation, where the optical fibre network plays a decisive role.
- Strategic relocation to cooler climates
A high-capacity, ultra-low-latency optical fibre network—such as lyntia’s—allows data processing to be moved to regions with cooler or more humid climates. In these locations, “free cooling” techniques (using outside air) can be used instead of constant water evaporation systems. Thanks to intelligent connectivity, data can travel thousands of kilometres in milliseconds to locations where cooling is most efficient and environmentally friendly.
- From copper to fibre: Less heat, less water
Replacing legacy copper cables with optical fibre is, in itself, an indirect water-saving measure. Fibre not only transmits more information but also generates significantly less heat within telecommunications exchanges. By reducing the thermal stress on the infrastructure, the need for heavy-duty cooling systems is drastically lowered, paving the way for “zero-waste” networks.
Spain: A responsible technological hub
Spain is establishing itself as a premier digital hub in Europe, with capacity projected to reach 4,000 MW by 2050. This growth is accompanied by a firm commitment to renewable energy and efficient management. Operators and data centres across the country are investing in closed-loop systems where water is continuously reused, minimising the extraction of new water resources.
The combination of a clean energy matrix and cutting-edge connectivity infrastructure allows Spain to lead the way toward an AI that is not only intelligent but also sustainable.
Conclusion: Moving toward positive-impact technology
The challenge is significant, but solutions are already being implemented. The future of sustainable data centres depends on the ability to integrate efficient hardware with an optical fibre network that optimises every process. At lyntia, we are committed to ensuring that the infrastructure supporting tomorrow’s AI is the same one that protects today’s natural resources.