The dark side of ChatGPT. Artificial Intelligence is devouring energy resources at an alarming rate

Although it has played a major role in reaching the threshold of zero carbon dioxide emissions, the tech industry also has a significant destructive side through its footprint on the environment. Large language models like ChatGPT are among the most energy-hungry technologies, writes Mariana Mazzucato, professor of economics at University College London (UCL) and director of the Institute for Innovation and Public Utility, in an opinion piece for The Guardian.


The tech industry is thinking about things that don’t physically exist, like apps or the internet browser on your phone. And yet, the infrastructure needed to store all this information – that is, data centers housed in office spaces and on the outskirts of cities – consumes huge amounts of energy. Despite its name, that infrastructure used for “cloud” accounts generates more greenhouse gas emissions globally than commercial flights.

For example, in 2018, the 5 billion YouTube views of the viral song Despacito required the same amount of energy that would be needed to heat 40,000 US homes annually.

This is the highly destructive side of the tech industry in terms of the environment. Although it has played a major role in reaching zero emissions by giving us smart meters and efficient solar power, it is still essential to turn our attention to its environmental footprint. In this regard, large language models such as ChatGPT are among the most energy-intensive technologies. Research suggests, for example, that an estimated 700,000 liters of water may have been used to cool the ChatGPT-3 training machines at Microsoft’s data centers.

Google’s global data center, as well as Meta’s ambitious plans for a new supercomputer called the Research SuperCluster (RSC), are a further illustration of the energy-intensive nature of the industry, raising concerns that these facilities could be responsible for a significant increase in energy consumption. Additionally, as these companies aim to reduce reliance on fossil fuels, they may decide to locate their data centers in regions where the cost of electricity is lower, such as in the southern US, which could exacerbate consumption issues of water in the drier areas of the world.

Additionally, while minerals such as lithium and cobalt are most commonly associated with batteries in the automotive sector, they are also essential for batteries used in data centers. The extraction process often involves significant water use and can lead to pollution, undermining water security.

In addition, when significant energy resources are allocated to technology-related activities, this can lead to energy shortages for essential needs, such as powering homes. Recent data from the UK shows that the country’s aging electricity grid is holding back affordable housing projects. In Bicester, for example, plans to build 7,000 new homes were put on hold because the electricity grid did not have enough capacity.

Making disclosure of practices and environmental impacts conditional on obtaining government support could ensure greater transparency and accountability. Similar measures could promote corporate responsibility in global mineral supply chains, enforcing greater respect for human rights.

Navigating the intersection of technological progress and sustainability, policymakers face the challenge of cultivating less extractive business models. It is not just about taking a step-by-step approach, but about taking a comprehensive systematic view that enables governments to build the necessary capacity for planning and implementation. Such an approach should eschew outdated top-down methods in favor of flexible strategies that integrate knowledge at all levels, from the local to the global. Only by adopting a holistic perspective can the technology industry’s significant impact on the environment be effectively mitigated.

Finally, despite the unprecedented wave of innovation in the 1990s, we have consistently neglected the repercussions of these advances on the climate crisis. As climatologists predict global warming will exceed the 1.5°C target, it is time to tackle today’s grand challenges systemically, so that solving one problem does not exacerbate another, concludes Mariana Mazzucato, professor of economics at University College London (UCL) and Director of the Institute for Innovation and Public Utility.