UN report warns about AI’s ecological footprint: Europe’s response

A new report from the United Nations University (UNU) paints an alarming picture of AI’s footprint. The hunger for computing power not only threatens our climate but also places enormous pressure on scarce natural resources such as water and land. Fortunately, European solutions—ranging from strict legislation to revolutionary computer chips and smart recycling—are already there.

The invisible ecological footprint of our data usage

The public debate on AI often focuses on the enormous amount of energy required to train large models. However, the UN University report shows that the real culprit lies elsewhere. As much as 80 to 90 percent of total energy consumption occurs during the daily use of AI services, known as the inference phase. The projections for the near future are downright alarming. By the year 2030, data centers worldwide could consume 945 terawatt-hours of electricity annually. This massive power consumption is accompanied by an equally massive thirst for water. By that time, cooling the servers will require an amount of water equivalent to the basic needs of 1.3 billion people. In addition, data centers are expected to consume 14,500 square kilometers of land and produce 2.5 million tons of electronic waste annually. The report also warns of the rebound effect. This means that more efficient systems lower the threshold for use, causing total demand and the resulting environmental damage to actually increase.

Europe’s active role

Europe is choosing an active role in addressing this ecological crisis. With the introduction of the AI Act, the European Union is setting a global standard for responsible technology. Article 40 of this legislation gives the European Commission the power to impose strict requirements on the energy and resource efficiency of AI models. This principle is known as “efficiency by design”.

Developers of both high-risk systems and general-purpose AI models are required to document the environmental footprint of their software. These rules are reinforced by other European directives, such as the Corporate Sustainability Reporting Directive (CSRD) and the Energy Efficiency Directive.

Companies are therefore forced to provide full transparency about their energy consumption. These laws aim to ensure that sustainability does not remain a marketing slogan but becomes a hard design requirement. Europe hopes to create a market in which green software becomes the standard. This not only protects the environment but also encourages the development of more efficient algorithms that are cheaper to use in everyday applications.

New ways of computing

To truly address the energy crisis of AI, a fundamental shift in computer hardware is required. Current computer systems are built on the traditional von Neumann architecture, where data constantly moves back and forth between the processor and memory. This data transfer consumes by far the most energy. Neuromorphic computing offers a radically different alternative. This technology mimics the human brain by physically integrating memory and processing. In theory, this approach could reduce the energy consumption of computer chips by a factor of 10,000.

For example, Neuromorphic Computing NL (NC-NL), the Netherlands, is taking an important step toward the next generation of intelligent and energy-efficient technology. Knowledge institutes, industry, and government are joining forces to ensure that the Netherlands remains an international leader in neuromorphic computing.

Recovering raw materials from Europe’s waste stream

Sustainable AI requires more than just reducing electricity use. The production of advanced microchips depends heavily on scarce minerals and metals.

To reduce this dependence, European research institutions are focusing on the so-called “urban mine”. TNO is exploring ways to recycle critical raw materials such as gallium and dysprosium from discarded electronics and solar panels on a large scale.

This is strategically important for European autonomy, as supply chains for these materials are currently largely controlled by countries outside Europe. At the same time, the Delft University of Technology is working on innovative methods to extract metals such as copper and platinum more efficiently from waste streams.

The need for a green digital transition

The coming years will be decisive for how we shape artificial intelligence. Warnings from the United Nations show that the current path is unsustainable.