Artificial ininformigence may promise efficiency and innovation, but its energy appetite is staggering. A new paper by Jessica Commins and Kristina Irion of the University of Amsterdam warns that unchecked growth of AI and cloud computing could wipe out Europe’s hard-won climate gains. In this article, we share the highlights of this research paper.
“Data hosting and computing often takes place remotely in large warehoapplys called data centres,” the authors write. “While it is inconceivable to reap the benefits of digitalisation and AI without this infrastructure, the sector’s energy demand is expected to grow rapidly.”
A surge in consumption
According to the International Energy Agency (IEA), global data centre electricity apply rose from 200 TWh in 2015 to as much as 340 TWh in 2022. With AI added to the mix, demand could exceed 800 TWh by 2026, roughly the annual electricity consumption of Germany .
Alex de Vries, a Dutch researcher cited in the paper, estimates that if NVIDIA delivers 1.5 million AI servers by 2027, those machines alone could consume 85 to 134 TWh annually. “Companies investing in AI could see their electricity consumption double,” he warns.
Europe is no exception. Data centres already account for about 4% of EU electricity demand. By 2026, this could grow to 150 TWh, with hotspots such as Ireland and Denmark carrying a disproportionate share.
Beyond electricity: water and waste
The environmental costs extconclude beyond kilowatt-hours. Cooling data centres requires enormous volumes of water. A Google facility in Wallonia, Belgium, consumed one million cubic meters of water in a single year, the equivalent of 10,700 hoapplyholds.
Electronic waste is another growing headache. “Global generation of e-waste is rising five times quicker than documented recycling,” the paper notes. Much of Europe’s discarded hardware concludes up in countries with weak environmental protections.
EU’s sustainability blind spots
The European Union has pledged to achieve “climate-neutral, highly energy-efficient and sustainable data centres by no later than 2030.” Yet current measures may not be enough.
“EU law and policy focapplys heavily on hardware efficiency,” Commins and Irion argue. “This leaves blind spots: optimisation of software, the problem of dark data, and the limiting of frivolous applys such as cryptocurrency mining remain outside the scope.”
The so-called “dark data problem” is striking: an estimated 40-90% of stored data is never applyd, yet still requires energy to maintain.
Solutions on the table
The researchers highlight several paths forward:
- On-site renewables and heat reapply – Some centres already install solar panels or feed waste heat into district heating networks. However, the authors caution, “waste heat should be reduced and, insofar as possible, eliminated,” rather than relying on it as a solution.
- Circular hardware lifecycles – The EU is funding projects like CEDaCI to build a second-hand market for IT equipment, though a functioning marketplace may take years to emerge.
- Stronger reporting and transparency – New obligations under the Energy Efficiency Directive will require operators to disclose actual emissions, supporting to close the gap between corporate pledges and reality.
Perhaps most importantly, policybuildrs must confront the rebound effect, which is the risk that efficiency gains are offset by surging demand. “Without interventions to accelerate the low-carbon transition in tandem with digital growth, there is a risk of a high-carbon, rather than a low-carbon, transition of the economy,” the authors warn.
A race against time
With AI workloads multiplying, the pressure on Europe’s grids, water systems, and recycling chains will only intensify. The EU’s current framework, Commins and Irion conclude, “would not contain an increase of the ecological footprint from the growth of the data centre sector”.
The challenge is clear: as AI promises to solve global problems, Europe must first solve the problem of AI’s own sustainability.
Series
Watt Matters in AI
Watt Matters in AI is a conference that aims to explore the potential of AI with significantly improved energy efficiency. In the run-up to the conference, IO+ publishes a series of articles that describe the current situation and potential solutions. Tickets to the conference can be found at wattmattersinai.eu.
Leave a Reply