AI is increasingly recognized as a crucial element for enhancing operational efficiency within data centers, according to a recent report from the agency managing much of the European Union’s digital framework.
The "Energy-Efficient Data Centers" report by the EU Agency for the Operational Management of Large-Scale IT Systems in the Area of Freedom, Security and Justice (EU-LISA) delves into various topics, ranging from regulations to best practices currently shaping the industry.
The report explores innovations in leveraging AI for improved data center management and operations, emphasizing the role of generative AI as a significant workload. This shift is contributing to the adoption of advanced technologies such as liquid cooling and the utilization of graphics processing units (GPUs). However, it points out that other AI applications in data center operations have received comparatively less focus.
According to the findings, in expansive cloud and hyperscale environments, AI-fueled control systems can enhance infrastructure operations in real time. This leads to improvements in critical efficiency metrics like power usage effectiveness (PUE), enabling greener management of computing resources.
Nonetheless, EU-LISA, which oversees numerous large-scale IT initiatives, cautions that AI's potential in data center energy management is still emerging. The report indicates that most AI-based solutions have not yet matured sufficiently to be regarded as primary strategies for boosting energy efficiency in data centers.
As demand for data center capacity surges and physical limitations become pressing, stakeholders across the EU provided insights on the findings of the report, discussing the opportunities and hurdles AI presents for the governmental data center sector.
AI's Transformative Impact on Data Center Management
Research from the Uptime Institute, a data center certification authority, aligns with EU-LISA’s observations. The Institute has noted that while AI can significantly impact data center operations, there remains a notable hesitancy toward widespread implementation.
The Uptime Institute's 2025 Global Data Center Survey acknowledges the availability of AI tools for managing data centers, but observes that the industry is in a phase of cautious testing and validation. The organization highlights that data centers are typically slow to adopt new technologies, making AI no exception.
Tor Björn Minde, director of the ICE Data Center unit at RISE Research Institutes of Sweden, which contributed to the EU-LISA report, noted that some AI technologies, like machine learning, could enhance the management of individual pieces of equipment or specific actions. Examples include optimizing settings or conducting predictive maintenance.
However, the complexity of managing an entire data center presents a more substantial challenge, likely exceeding the capabilities of current AI solutions. Minde emphasized that data centers represent intricate environments where risks must be minimized.
One of the challenges involves training AI tools to respond to failures—a task complicated by a lack of available data on failures or non-optimized operations. Minde posed the question: "Without data on failures, how can the model adapt?"
AI Innovations Shaping Data Centers' Future
Despite concerns regarding maturity, the EU report recognizes advancements made by hyperscalers in implementing AI to manage data centers. Some noteworthy examples include:
- Google, which used DeepMind technology to optimize HVAC settings, achieving a 40% reduction in cooling energy consumption and a 15% increase in PUE. - IBM Watson employs real-time data analysis to foresee failures in aging equipment such as fans and uninterruptible power supplies (UPS). - Meta has utilized AI to organize non-essential workloads based on renewable energy availability, enhancing energy efficiency from solar and wind sources.
The report also presents other cutting-edge technologies:
- High voltage direct current (HVDC) systems, which minimize energy loss during power distribution by bypassing several conversion steps, are backed by suppliers like Nvidia for AI-focused data centers; however, widespread adoption of this new standard is expected to take time. - Cooling chillers with magnetic bearing compressors are designed to enhance energy efficiency but involve higher initial costs and longer payback periods. - Hydrogen production and storage are mentioned as potential components of future energy grids, although they carry concerns regarding carbon intensity and storage expenses.
Insights from EU-LISA on Data Center Management
Javier Galbally, a research and innovation officer at EU-LISA, explained that the organization manages three critical data centers across Europe, with its primary facility located in Strasbourg, supported by additional sites in Austria and Estonia. Regulations stipulate that EU-LISA must maintain its on-premises data centers and cannot utilize third-party service providers.
Galbally indicated that the impetus behind the report stems from the capacity limits experienced at their main site, highlighting the importance of understanding best practices for boosting data center efficiency to share those insights with other organizations.
“We are nearing our physical space capacity for adding more racks,” he explained. “Expansion will be necessary at some point.”
While AI promises significant potential to revolutionize data center operations—improving efficiency, minimizing energy consumption, and advancing sustainability—experts from EU-LISA and other sectors emphasize that the technology is still developing. Consequently, organizations must pursue thoughtful validation and adoption strategies.
As the hunger for data center capacity intensifies, entities like EU-LISA are diligently exploring innovative pathways while balancing operational stability and safety. Although the transition to fully AI-optimized data centers may be slow, the advancements made by hyperscalers and emerging technologies forecast a positive trajectory for the industry.
