Tools capable of generating images, text, videos, and analysing enormous amounts of information: this is artificial intelligence, a technology that has become one of the most revolutionary innovations of our time.
Applications such as Gemini, ChatGPT, and other tools based on generative artificial intelligence systems have changed the way people work, access information, and communicate.
Answers to users’ questions are generally generated within a matter of seconds, but behind every query submitted to AI lies a highly complex technological infrastructure. In fact, responses are not generated directly by the device from which the question is asked, but by data centres. Data centres are large facilities—true computing hubs—where thousands of computers operate continuously.
But as artificial intelligence continues to grow, what is the environmental cost of this technological revolution?
When people think about the digital world, they often imagine something non-physical. Indeed, online services are frequently perceived as intangible, existing without occupying physical space or relying on real infrastructures. The reality, however, is that every digital service depends on physical infrastructure capable of housing servers, electrical power systems, vast amounts of resources needed to keep computer networks operational, and cooling systems.
According to the International Energy Agency (IEA), data centres currently represent a significant and increasingly important share of global electricity consumption. The energy sector is therefore being required to respond to the growing demand for electricity, as the development of artificial intelligence requires computing power that is considerably greater than that of traditional systems.
Thus, while demand continues to increase, the expansion of artificial intelligence requires new infrastructures and higher energy consumption, creating an important challenge for environmental sustainability.
The question, therefore, is not whether AI is inherently positive or negative, but rather how it can be used and developed responsibly by seeking to create a balance between technological innovation and environmental protection.
So, where is a system of this scale actually located? As already mentioned, the physical place where artificial intelligence operates is known as a data centre. Consequently, the impression that AI resides entirely within our smartphones or personal computers is only partially correct; the most complex processing is carried out remotely, inside these enormous technological facilities.
Data centres can be imagined as vast digital factories. Rather than producing physical objects, they process data through thousands of servers specifically designed to handle enormous quantities of information while operating continuously.
Therefore, every time a user interacts with an artificial intelligence system by asking it a question, that request is sent to these computing centres, where the servers analyse it, process it, and, through complex mathematical models, generate a response. This process, however, requires a substantial amount of energy.
In its 2025 report Energy and AI, the International Energy Agency highlights how the growing demand for artificial intelligence is reshaping the role of data centres within the global energy system. According to the Agency’s estimates, the electricity consumption of data centres could increase significantly in the coming years. This growth would primarily be driven by the expansion and widespread adoption of generative artificial intelligence.
It should be clarified, however, that this does not mean that every use of AI is automatically significant or that it necessarily has a negative environmental impact. In fact, energy consumption depends on many factors, such as the type of model used, the energy source employed, the efficiency of computing systems, and the technologies used to cool the equipment.
Artificial intelligence models are trained through processes known as machine learning. During this phase, the system analyses enormous amounts of data and adjusts millions or even billions of internal parameters in order to recognise patterns and produce increasingly accurate results. These calculations are carried out using GPUs (Graphics Processing Units), which are specialised processing units. Originally, these components were used to enhance graphics in video games; today, they have become essential for artificial intelligence because of their ability to perform a vast number of calculations simultaneously.
This is why increasingly large and powerful facilities are required, inevitably leading to higher energy consumption.
So far, we have discussed environmental impact and electricity consumption, but another essential resource is water. Why water? As previously mentioned, a data centre consists of thousands of servers working together, and every operation generates heat.
Artificial intelligence, particularly during the training of advanced models, requires significantly greater computing power than traditional digital services. As a result, very high temperatures are reached, which, if left uncontrolled, could reduce performance or even damage the equipment.
To prevent these issues, data centres rely on sophisticated cooling systems. Some use large volumes of circulated air, others adopt hybrid air-and-water systems, while others rely primarily on water cooling. These mechanisms ensure that the equipment is cooled effectively.
According to Google, cooling represents one of the most important aspects of data centre design. In recent years, the company has invested in the development of more efficient technologies aimed at reducing both energy and water consumption by adapting cooling systems to the climatic characteristics of the different geographical areas in which its facilities are located.
In regions where water is more readily available, evaporative cooling systems are often used, whereas in other areas solutions that minimise water consumption are preferred. Water use has been at the centre of public debate for several years, as many have pointed out that the expansion of data centres could place additional pressure on local water resources. This does not mean that data centres are inherently harmful, but rather that appropriate strategies must be developed to ensure both efficiency and long-term sustainability.
This is why many companies are investing in new cooling technologies. Among the most promising solutions is liquid cooling, which uses refrigerant liquids in direct contact with electronic components, as well as systems capable of recovering the heat generated by servers to warm buildings or supply district heating networks. In this way, part of the energy that would otherwise be wasted can be reused.
The challenge, therefore, is not simply to build increasingly powerful devices, but to design infrastructures capable of reducing their environmental impact without slowing technological innovation.
So far, the focus has been on how the development of artificial intelligence requires increasingly large infrastructures. However, AI is not only a technology that consumes resources; it is also a tool that can contribute to managing them more efficiently. One of the areas in which artificial intelligence is demonstrating considerable potential is energy management. By analysing vast amounts of data, AI systems can forecast electricity demand and facilitate the integration of renewable energy sources, such as solar and wind power, whose production varies according to weather conditions.
Artificial intelligence is also finding applications in climate change research. Machine learning models can help scientists analyse enormous volumes of climate data, improving weather forecasting and enhancing the understanding of complex environmental phenomena. Rather than replacing traditional models, AI serves as a valuable supporting tool.
Naturally, these benefits do not eliminate the associated challenges. As highlighted by the IEA, the overall impact of artificial intelligence on the climate will depend on the decisions made in the coming years.
Artificial intelligence represents one of the most significant innovations of the twenty-first century. Its ability to process enormous quantities of information is transforming work, medicine, industry, scientific research, and many other sectors.
While the widespread adoption of artificial intelligence systems presents new challenges in terms of sustainability, the same technology also offers valuable tools for improving energy efficiency and supporting scientific research aimed at addressing increasingly complex environmental issues.
Therefore, the ultimate question is not whether artificial intelligence should continue to evolve, but rather how it should do so. Making technology compatible with sustainability goals is a major challenge, and only through a responsible and informed approach will it be possible to make a meaningful contribution to protecting our planet
Author: Ms Maria Lagani, psychology graduate- Master’s student. Research Team for JUMP staff (Italy)
References:
Google. Operating Sustainably. Google Data Centers.
International Energy Agency (IEA). (2025). Energy and AI. Paris: International Energy Agency.
Stanford Institute for Human-Centered Artificial Intelligence (HAI). (2025). AI Index Report 2025. Stanford University.
United Nations Environment Programme (UNEP). Artificial Intelligence and the Environment.
