The European Union has unveiled a $30 billion initiative aimed at constructing high-capacity AI data centers. This project seeks to bolster the EU’s standing in the global AI market, which currently lags behind the United States and China. The initiative’s goal is to establish a network of data centers capable of supporting millions of AI GPUs.
To date, the EU has committed €10 billion, approximately $11.8 billion, for the establishment of 13 AI data centers. Concurrently, an additional €20 billion has been allocated as initial funding for a network of gigawatt-class AI facilities. According to CNBC, this project has garnered significant interest, attracting 76 expressions of interest from 16 member states, encompassing a total of 60 potential locations for these facilities.
The initial phase of the project is already underway, with the first AI factory anticipated to become operational in the coming weeks. Furthermore, a large-scale project is scheduled to commence in Munich early in September, marking a significant step forward in the EU’s AI infrastructure development. These initial projects represent a concrete manifestation of the EU’s commitment to strengthening its AI capabilities.
Each gigawatt datacenter is projected to cost between €3 and €5 billion. UBS estimates, as reported by CNBC, these facilities will deliver computational power significantly exceeding that of existing AI data centers. This enhanced capability could potentially support over 100,000 advanced AI GPUs per site. The scale of these facilities is designed to provide the EU with a substantial boost in AI processing capacity.
Temu is now in hot water with European regulators
To provide context, xAI’s Colossus super cluster consumes approximately 150 MW of power when equipped with 100,000 H100 GPUs. Given this benchmark, a gigawatt facility would possess the capacity to host a considerably larger number of GPUs, potentially accommodating around 300,000 Blackwell Ultra processors. This illustrates the immense computational potential of these planned data centers.
If fully realized, the EU’s initiative stands as one of the world’s largest publicly funded endeavors in the field of artificial intelligence. While the investment may fall short of the total expenditure by Chinese authorities at both the federal and local levels, it surpasses the AI investment efforts of other major economies. This underscores the EU’s commitment to closing the gap with global AI leaders.
Henna Virkkunen, European Commission executive vice president for technology policy, emphasized that Europe possesses a strong talent pool. She stated that Europe has “reportedly 30% more AI researchers per capita than the U.S.” However, she noted that their development has been hindered by “limited access to computing.” The establishment of these massive AI data centers is intended to address this limitation, stimulating growth across the EU’s AI sector.
Despite the considerable interest expressed in the project, concerns persist regarding its overall scale and long-term sustainability. Bertin Martens of Bruegel raised questions regarding the level of public sector investment, emphasizing that while the EU has allocated taxpayer funding, the precise amount of public investment remains unclear. In addition, the specific technical specifications of the data centers have not yet been fully defined.
Martens also noted that access to advanced AI hardware, such as Nvidia GPUs, is only the first step. He highlighted the necessity of developing robust business models, thorough planning, and securing interest from private companies to ensure the profitable operation of these data centers and the creation of world-class AI products. These factors are crucial for translating the investment into tangible AI advancements.
Power supply represents a significant concern for the project. A gigawatt-class data center demands an enormous amount of power, and establishing such generation capacity is a time-consuming process. Martens cautioned that Europe’s existing grid infrastructure may lack the capacity to support such high loads. Consequently, substantial upgrades to the grid infrastructure are likely to be required, potentially affecting the project’s timeline.
