EU launches AI super factory competition to reduce reliance on U.S. cloud

EU launches AI Gigabit factory bidding plan

The EU has launched a bidding process for up to seven AI "Gigabit factories" as part of its procurement plan. The plan aims to mobilize more than 30 billion euros in public and private investment to enable European companies to gain access to the computing power they need to develop cutting-edge AI models.

Brussels wants more autonomy in the hardware needed for future AI technology, rather than relying on cloud services from U.S. companies. The decision comes at a time when computing power has become a major bottleneck in the industry. Advanced AI models can only be trained through clusters containing tens of thousands of top accelerators, and European officials know that existing public infrastructure is overwhelmed.

The plan is part of the European Commission's larger strategy called "AI Continent" to achieve sovereign AI computing capabilities. In addition, this move confirms the trend revealed in Stanford University's 2026 AI Index report: Countries around the world are beginning to view AI infrastructure as a strategic national asset as important as semiconductors, energy and data.

Why owning hardware is key

In the words of a senior European Commission official, Europe's problems can be summarized in one sentence: "Current infrastructure needs are saturated." This means that the biggest challenge facing European AI today is not the lack of research talent, but the lack of computing power. This assertion strengthens Stanford University's conclusion that the development of cutting-edge AI technology often benefits entities with huge budgets to build and operate large GPU clusters.

Under the AI Gigabit Factory plan, each site will have the ability to install up to 100,000 cutting-edge AI processors and be equipped with ultra-modern networks, cloud software, reliable power supplies, and state-of-the-art cooling technologies for training and running large models. Europe has built 19 AI factories connected to public supercomputers. The goal of these gigabit factories is to provide an alternative to existing facilities by building larger, privately led computing facilities.

The purpose is to ensure that AI development takes place within the EU's existing regulatory and data governance frameworks. The plan also echoes the European Union's Artificial Intelligence Coordination Plan, implemented since 2018, which aims to integrate investment from member states rather than separate them.

How 30 billion euros constitutes

This financing plan uses state funds to attract more private sector investment. Overall, funding is expected to include € 10 billion from countries and the EU, which will leverage € 20 billion from private investors, and public financing typically does not exceed 35% of project spending. Part of the funding will still depend on the outcome of future EU budget negotiations. The European High Performance Computing Consortium (EuroHPC JU) will be responsible for the competitive selection process and maintain the EU's strategy of attracting private sector investment in key technologies with public funds.

Two bid sections and massive chips

The bidding is divided into two categories: one category supports up to four medium-sized Gigabit factories, and the other category supports up to three large facilities. Depending on the project, the amount of funds provided by the EU and governments ranges from 1 billion to 2 billion euros. Medium-sized sites should be able to efficiently deploy 25,000 to 75,000 AI processors, while large sites can exceed 100,000. Countries such as Germany, Italy, Greece, Portugal and Spain have shown interest in operating large facilities.

A cluster of this size would place the EU's gigabit factory among the largest AI training facilities operated by large U.S. companies today, but it is still insufficient compared to the hundreds of thousands of GPU clusters planned by companies such as xAI. The comparison highlights Brussels 'goal: to build technologies that can train cutting-edge AI models, rather than ordinary scientific supercomputers.

Europe's dependence that cannot be freed from through design

Simply building an independent AI infrastructure will not reduce Europe's dependence on foreign semiconductor manufacturers. After the EU-US trade agreement was reached in July 2026, the European Commission has signed letters of intent with AMD, Nvidia and Qualcomm to increase access to hardware related to gigabit factory projects. However, officials are very realistic about this.

Due to intensified competition, future hardware upgrades are expected to involve more suppliers, but chips alone cannot solve the problem. The International Energy Agency also pointed out that AI data centers are exacerbating global power demand growth, so in addition to advanced processors, attention must also be paid to reliable power supply, sufficient grid capacity and cooling systems.

Time is running out

The application deadline is November 12, 2026. After evaluation by EuroHPC, the awards are expected to be announced in early 2027. The selected projects are expected to be put into operation within approximately 18 months after the contract is signed. Henna Verkunen, Executive Vice President of the European Commission for Technical Sovereignty, Security and Democracy, declared the launch a "milestone" in the EU's AI vision.

The EU's interest in gigabit factories is not limited to increasing the cloud capacity of economies, but also because these facilities may provide much-needed infrastructure for faster deployment of AI in areas such as manufacturing, medical care, transportation and public services. However, OECD analysis shows that the success of AI application across EU countries has been uneven, which means that the 30 billion euros of additional computing power is just a small piece of the puzzle in the important battle of economic competitiveness.