The sustainability challenge: rising demand for energy, water, and materials

AI system training and inference – alongside continued growth in digitalization – increase the energy demand for computing, processing, storing, and data exchange. At the global level, data centers consumed 415 TWh in 2024, roughly equivalent to the combined electricity consumption of Italy and the Netherlands. The United States accounts for almost half of the global volume, followed by China (25%) and Eastern and Western Europe (15%).¹

The power demand of data centers is expected to grow as they continue to expand, resulting in a 70% increase in Western Europe by 2030 compared to 2023 and exceeding 80 TWh, according to Deloitte research (Figure 1). This projection is based on a scenario that assumes the continued rapid growth of AI server deployment (high adoption). It also aligns with the European Union’s (EU) ambition to triple its data-center capacity over the next five to seven years, as announced in April 2025.² In contrast and in a baseline scenario in which AI adoption continues at a more moderate pace, electricity consumption is expected to rise 40% by 2030.³

High water consumption at data centers is another concern as large amounts of water are required for cooling. As data centers expand, management of water resources will remain a critical challenge, especially in areas where water supply is already under stress. The construction and operation of data centers also require significant amounts of raw materials. Examples include arsenic and beryllium for semiconductors, and rare earth metals such as neodymium and dysprosium, whose mining and processing impact the environment, in addition to posing supply-chain risks.^4,5

The policy environment: incentivizing sustainable value chains in the EU

Sustainability targets remain a central priority for the EU, despite the current loss of momentum. In fact, geoeconomic and political events have undermined progress on the implementation of the Green Deal, Europe’s growth strategy for transforming the economy into a net-zero, circular system by 2050.⁶ Yet efforts to enhance the sustainability of data centers remain relevant, even in light of the concessions proposed under the EU Omnibus packages. For example, the EU Ecodesign regulation specifies requirements that information and communication technologies are expected to apply as of 2027/2028.⁷

The growing emphasis on building sovereign and globally competitive AI and digital value chains at both the EU level and among several member states is expected to increase environmental pressure.^8,9 These concerns are addressed in the AI Continent Action Plan, published in April 2025, which outlines a set of measures to position the EU as a global leader in AI. Several of the proposed initiatives explicitly incorporate sustainability objectives, such as enhanced water and energy efficiency, reused waste heat from data centers, and leveraged AI to improve environmental and climate outcomes (AI for green) (Figure 2).

The solution: key measures and innovation

Beyond regulatory pressure, elevated energy and resource demands carry strategic, financial, operational, and resilience implications for companies in the sector. Access to energy is becoming a key constraint on the expansion of computing infrastructure and services. At the same time, competitive pricing for digital services is driving efforts to improve energy efficiency.

Moreover, addressing sustainability challenges remains critical for data-center companies to meet their internal environmental targets. While many have significantly improved their sustainability performance in recent years, the recent surge in AI adoption poses a risk that could undermine this progress.¹⁰ As a result, all elements along the data-center value chain are actively developing and implementing sustainability measures (Figure 3).

The reuse of hardware components and recycled materials is vital to reducing raw material use, not only softening environmental impact but enhancing supply-chain resilience.¹¹ The latter is increasingly critical, particularly in the light of pressures regarding consistent supply of key components, such as graphics processing units (GPUs) and semiconductors.

Several strategies can reduce the energy consumption of AI-based applications at data centers:

• selection of the most suitable AI model for specific AI-based applications, e.g. by shifting towards smaller, less energy-intensive models
• more efficient deployment of applications, such as training models, at data centers by optimizing hardware for specific use cases, and using higher quality data for training to reduce computational requirements¹²
• development of applications that take data-center efficiency requirements into account, including a focus on data-center/cloud-native application architecture and green coding practices¹³

What’s more, advancements in smart energy use allow data centers to consider reallocating or shifting computing tasks on the basis of the availability of renewable energy. This is particularly suited to the training of AI systems, which is less sensitive to latency and does not require immediate results. Some hyperscalers are beginning to offer this feature through dynamic pricing models, with lower rates when renewable power is plentiful. They are also enhancing transparency around energy and water usage so that customers can monitor and optimize their resource consumption and costs. This increased visibility is crucial to fostering behavioral change in the consumption of digital resources.

Innovation: the window of opportunity for sustainable data centers in the EU

Data centers are the backbone of the (digital) economy, but their operations carry an environmental footprint. Addressing this impact is increasingly important, as Europe scales up its digital infrastructure. By developing and implementing advanced technologies, such as AI-driven energy management, efficient cooling systems, and circular resource strategies, data-center operators can reduce their environmental impact while unlocking cost savings and improving resilience.  

Authors and contributors:

Ines Österle, Manager | Economic Research 
Katrin Grashof, Consultant | Consulting
Asad Ali, Senior Consultant | Consulting
Tobias Koppe, Partner | Consulting
Carolin Holle, Junior Staff | Consulting
Mosche Orth, Senior Manager | EU Policy Centre

We are grateful to Oracle for sharing their perspectives on sustainability challenges of data centers in Germany.