Key Takeaways
1. The European Commission has allocated €10 billion for the first network of 13 regional AI factories and plans to invest an additional €20 billion for larger gigafactories.
2. Each gigafactory is expected to cost between €3 and €5 billion and will house at least 100,000 high-performance GPUs, significantly increasing Europe’s AI computing power.
3. The initiative aims to enhance Europe’s sovereignty in AI by providing centralized infrastructure for start-ups and SMEs to access affordable resources for training and inference.
4. Securing adequate electricity supply poses a challenge, as a single gigafactory may consume up to one gigawatt of power, necessitating upgrades to power grids and generation capacity.
5. Initial pilot factories are already operational, with projects like Telenor’s AI factory in Norway and a major facility opening in Munich, aiming to close the trans-Atlantic AI gap.
Europe’s “AI factory” initiative has transitioned from being just an idea to actual building, with the European Commission allocating €10 billion (≈ US$11.6 billion) for the first network of 13 regional AI factories. Additionally, they plan to invest another €20 billion (≈ US$23.2 billion) to kickstart a second phase of much larger gigafactories.
Ambitious Scale of Gigafactories
These gigawatt-class facilities are quite ambitious: each is expected to cost between €3 and €5 billion (≈ US$3.5–5.8 billion) and will contain at least 100,000 high-performance GPUs. They are projected to provide significantly more computing power than the largest current AI clusters in Europe. According to UBS modeling, if 10 to 15 of these plants are fully constructed, they could contribute around 1.5 to 2 GW of IT load—approximately 15% of Europe’s existing data center capacity.
Aiming for Sovereignty
Brussels presents this project as a matter of sovereignty. The European Union has around 30% more AI researchers per capita compared to the United States, yet it lacks sufficient computing resources to meet their goals. By bringing together hardware, data, and expertise in centralized “one-stop shops,” the Commission aims to provide start-ups and small-to-medium enterprises (SMEs) with affordable access to large-scale infrastructure for training and inference.
Electricity Supply Challenges
However, securing enough electricity might be a significant challenge. Experts warn that a single gigafactory could consume as much as one gigawatt of power, which is similar to the output of a mid-size nuclear power plant. Upgrading the existing power grids and increasing generation capacity will likely require more time than constructing the data centers themselves. The role of private investment remains uncertain; officials acknowledge that public funding alone won’t bridge the financial gap, and think tanks like Bruegel have raised concerns about whether subsidies are the right approach for such large-scale projects.
Progress on Pilot Sites
Some initial pilot factories are already operational. Telenor’s small AI factory in Norway has started testing language services for public-sector clients, while the first major EU facility is set to open in Munich in early September. If the challenges related to power and financing are addressed, Brussels is optimistic that this computing infrastructure will help close the trans-Atlantic AI gap and foster a new wave of European models and applications.
Source:
Link