United States

Cristian Fàbrega Gallego: "AI is not environmentally free, but it is not the monster that is presented to us either"

Professor of the Department of Electronic and Biomedical Engineering and coordinator of the Master's Degree in Renewable Energies and Energy Sustainability at UB

Cristian Fábrega
Miquel Rodrigo Ubach
26/07/2026 - 19:11 h.
4 min

BarcelonaCristian Fàbrega Gallego is a professor in the Department of Electronic and Biomedical Engineering and coordinates the master's degree in renewable energies and energy sustainability at the University of Barcelona. He speaks to ARA to explain the environmental impact and energy consumption of artificial intelligence data centers, which are multiplying around the world.

We always talk a lot about water, but what other resources do artificial intelligence data centers consume?

— The environmental impact of these data centers or any other system is diverse. It is true that in this case water is often mentioned, because at the time there were some somewhat viral news stories, but the main consumption is electric. Then there are the water resources it needs, especially for cooling systems. There are also other impacts that are quite more local, such as noise or heat emission, which can sometimes also cause temperature variations around the facility, as well as the need to connect to the electrical infrastructure and the impact on the areas where they are built.

He mentioned noise pollution. There are also people who complain about the light pollution from these centers. Can this have consequences for human health?

— If the data center is not well built and is close to populated areas, the sound aspect can have repercussions on health. The light aspect is not as evident. The centers are usually underground and can hardly have a very high impact on health. And all these effects can be mitigated, because there are technologies to do so. In other words, it is not a technology different from any other, with similar impacts.

Has the demand for energy increased with the arrival of generative AI models?

— The need for data centers that require a large amount of resources has skyrocketed. However, if they are well managed and well integrated into the areas where the centers are located, the impact should not be different from any other type of facility. The problem is that it is desired to do it so quickly and there is such a great need for data centers that not all the necessary criteria are applied to reduce their impact.

Is it known what part of the centers' consumption is dedicated to computing and what part to cooling?

— Refrigeration consumption is very high, because almost all the energy consumed by data centers is ultimately dissipated as heat. But when calculations are made, the water needed for the cooling systems that the power plants generating electricity for the data centers require is also accounted for.

Is air cooling really a viable alternative to water cooling?

— Yes, there are alternative technologies for cooling. In the case of air cooling, what is done is to move the problem elsewhere, because it consumes a lot of energy. The water resource will come from the extra energy that is needed for cooling. That is, the water impact depends a lot on the origin of the electricity consumed and the technology used for cooling. And this depends on the location and the electrical system it is connected to. There is no study that has given a universally applicable value; studies are done with respect to a very specific location or a very specific data center.

How many years of useful life can the centers have and how is the management of the waste they generate once obsolete planned?

— I couldn't say exactly what the lifespan is. It's true that these types of technologies are regularly updated because new, faster ones appear. I don't think this aspect is critical because we're talking about well-defined and controlled structures, and therefore management can be done easily. In fact, the problem with hardware is more serious with the computers we have at home. Here the problem is very decentralized and when our computer dies, we don't know what to do with it and a recovery and recycling circuit isn't established.

And if we talk about Catalonia, is it a region with the capacity to host centers of this kind? What regulatory measures should be applied to make its presence sustainable?

— Catalonia has a good electrical network. It is true that the integration of renewable energies is a bottleneck. For many years now we have had problems due to bureaucratic issues, many diverse implementation issues, and, above all, environmental ones. If regulations or rules are approved, the data centers will be perfectly viable, not only in Catalonia, but anywhere in Spain or Europe. The underlying problem is that these are barriers and, therefore, the managers of data centers will end up going to a country where there is no regulation.

Will the potential help of AI in developing greener and more sustainable systems ever offset its footprint?

— It is difficult to say. I think it will hardly compensate for the consumption it may have, but none of the technological evolutions we have had have compensated for the environmental part. Artificial intelligence is not environmentally free, but it is not the monster that is sometimes presented to us either. Its impact will depend on the design of the data centers, where the energy it consumes comes from, or whether we are able to make better use of the resources they use.

Consum d'aigua dels centres de dades d'IA
stats