- THE ECB BLOG
AI versus green: clash of the transitions?
25 March 2025
AI adoption requires enormous amounts of electricity. And so does greening the economy. Are the digital and green transitions clashing or can they be successfully achieved together? The ECB Blog takes a closer look.
This post is part of a miniseries related to the ECB conference “The Transformative Power of AI”, on 1-2 April 2025, bringing together researchers, practitioners, and policymakers. Learn more here.
Two of the biggest economic challenges Europe faces are the digital and the green transitions. The former seeks to harness the productive possibilities of new technologies, especially artificial intelligence (AI). The latter aims to transition away from carbon usage and green the economy. At a time when electricity generation needs to move away from fossil fuel, both transitions need vast amounts of electricity, to drive data centres, electric vehicles, and heat pumps. These competing demands appear to put the twin transitions onto a collision course. This post argues that the interaction between AI and the green transition goes much deeper than electricity demand. AI can in fact help develop technologies that substantially benefit the green transition. Addressing certain structural factors will benefit both societal goals.
An enormous appetite for energy
The data centres that underpin AI are voracious guzzlers of electricity. Research by Eurosystem staff demonstrates just by how much electricity demand could increase as AI adoption spreads. Take an everyday example: a standard Google search consumes around 0.3Wh of electricity. With that, one could power a standard 60W light bulb for 20 seconds. Once AI comes into play, that picture changes dramatically. A ChatGPT request needs around ten times the electricity to run. An AI-augmented Google search even requires 7-9Wh per request.[1] That is around 300 times the electricity demand, and the power needed to keep the same light bulb on for up to 9 minutes.
Electricity demand driven by the digital transition is already substantial in some Member States. For instance, data centres accounted for around 21% of the total electricity used in Ireland in 2023, more than all urban households combined.
And while the overall electricity demand is already high, data centres are expected to be a substantial driver of even higher EU electricity demand in the coming years (Chart 1). Yet, they are hardly the only. Their contribution is still less than that of the projected 9 million extra battery electric vehicles and 11 million new heat pumps over the same period. This growing demand for electricity makes it harder to decarbonise energy. And consequently, meeting the EU targets for greening electricity requires massive new renewables capacity. The greater the increase in demand, the greater still the increase in renewables needed.
Chart 1
Data centres and green transition to drive rebound in EU electricity demand
Estimated drivers of changes in EU electricity demand
TWh
Source: International Energy Agency.
Even without the increased demands of AI and green investments, meeting those renewable energy targets already looks tough. Expanding the supply of renewable energy is hampered by the sluggish pace of authorisations, which can take years, and long waits for connections to the energy grid. The International Energy Agency estimates that constructing and connecting all the wind and solar projects that have already been granted grid clearance would increase the capacity already in place in Italy at the end of 2022 by 45%, and even triple the equivalent capacity in Spain.
Given these competing demands, it would seem that the development of AI is at loggerheads with the goals of the green transition.
Technological transformation of the economy
It’s worth remembering, however, that at their core the widespread adoption of AI and the green transition involve the same thing: a technological transformation of the economy and society. Combating climate change requires the creation of new, carbon-free technologies and their diffusion throughout the economy. And this is a process that is already benefitting from the support of artificial intelligence. For the future, AI holds even greater potential to improve resource and energy efficiency.
One example is the AI-supported efficiency gains in the energy sector. AI has accelerated the modelling for new wind power sites by factor of up to 4,000. Rather than months, the time taken to calculate the optimal placement for new wind turbines can now be measured in minutes. Furthermore, AI can help with grid maintenance and predicting renewables supply and overall electricity demand, supporting the balancing of the grid. AI is also helping data centres themselves become more efficient. For example, Google used DeepMind to reduce the energy needed to cool its data centres by 40%. AI can do the same for the energy usage of other buildings, which contribute a substantial share of carbon emissions in Europe. One trial in Sweden even managed to reduce the energy used by two large apartment buildings by 20%.
Next to reducing energy usage, battery technology is key for a successful green transition. AI has the potential to help find more efficient materials for production, improve battery maintenance for longer battery life, and assess operational risk in real time. In the longer-term, AI-powered self-driving cars could facilitate car sharing, reducing the total number of cars needed (since they mostly stand idle), and in turn reducing the environmental burden of car production.
Overcoming common challenges
The examples above illustrate why AI and the green transition in fact don’t have to be competitors. Rather, they can be companions that can help and feed off each other. Indeed, the main clash – the energy demand of data centres – isn’t intrinsic. Competitive sources of (nearly) carbon-free electricity generation already exist. Moreover, the main roadblock to both building new data centres and greening electricity generation is the problems surrounding authorisations and grid connections. Addressing these bottlenecks is vital if Europe wants to achieve the productivity gains offered by AI and avoid the productivity losses of further climate change.
There are further common challenges both AI adoption and the green transition face, particularly in Europe. Both transitions require substantial investment. In this regard, Europe is falling behind, particularly on spending on research and development. The green transition will require from 2.7% to 3.7% of EU GDP in additional investment every year until 2030. The European Investment Bank has identified an AI investment gap of around 5-10 billion euro per year. Around 40% of small and medium-sized companies cite a lack of investor willingness to finance green investment as a very significant obstacle. Meanwhile, young European firms receive only a quarter of the scale-up finance enjoyed by US firms.
Recent ECB work has highlighted the need for strong institutions and governance to speed up the adoption of digital and green technologies. There is little incentive for firms to innovate if barriers to market entry and competition prevent them from reaping the benefits. Inefficient labour and product markets might make the needed reallocation of inputs across firms and sectors more difficult. Lack of skilled labour may also prove to be a hindrance, with the need to boost education in science, technology, engineering, and mathematics. Lifelong training and education are also necessary as AI and the green transition result in shifts in employment.
In light of these challenges, national environmental and AI policies would benefit from more concerted integration to harness the synergies between these twin transitions. There is a role, too, for policies and funding at European level, particularly where the environmental benefits of AI may be widespread but have fewer commercial applications. Taken together, these can help ensure that the common challenges to both transitions can be overcome and to ensure that – rather than clash – both are successfully delivered.
The views expressed in each blog entry are those of the author(s) and do not necessarily represent the views of the European Central Bank and the Eurosystem.
Check out The ECB Blog and subscribe for future posts.
For topics relating to banking supervision, why not have a look at The Supervision Blog?
De Vries, A. (2023), “The growing energy footprint of artificial intelligence”, Joule, Vol. 7, no 10, pp 2191-2194.