The European Union is the world’s second-largest economy in terms of GDP, but when it comes to its place on the AI market, its position is by far not as strong. To catch up with the U.S. and China, the bloc is launching a $30 billion initiative to build a network of high-capacity data centers that can host millions of AI GPUs, reports CNBC. If successful, the EU will have gigawatt-class datacenters with performance akin to that owned by leading U.S. companies.
To date, the European Union has allocated €10 billion (approximately $11.8 billion) to establish 13 AI data centers, alongside an additional €20 billion earmarked as initial funding for a network of gigawatt-class AI facilities. So far, the project has attracted 76 expressions of interest from 16 member states, covering a total of 60 potential locations, according to CNBC. Initial launches are underway, with the first AI factory expected to go live in the coming weeks and a large-scale project in Munich planned for early September.
Each gigawatt datacenter is expected to require €3 to €5 billion and deliver a level of computational power far greater than existing AI data centers, potentially supporting over 100,000 advanced AI GPUs per site, according to estimates by UBS cited by CNBC. xAI’s Colossus cuper cluster consumes about 150 MW of power when equipped with 100,000 H100 GPUs, so a gigawatt facility will probably be able to host many more GPUs. Perhaps, 300,000 Blackwell Ultra processors.
The EU’s effort, if realized, is probably one of the world’s largest publicly funded initiatives in artificial intelligence, probably well below what Chinese authorities (both federal and local) have invested in AI data centers, but well ahead of what other big economies invest in their AI efforts.
Henna Virkkunen, European Commission executive vice president for technology policy, told CNBC that while Europe has a strong talent base — reportedly 30% more AI researchers per capita than the U.S. — their limited access to computing has held back development. Building massive AI data centers is designed to solve this problem and kick-start the AI sector across the EU.
Despite strong public interest, the scale and sustainability of the project remain in question. Bertin Martens of Bruegel noted that while the EU has committed taxpayer funding, it is unclear how much the public sector will invest in the project. Also, the specifications of the upcoming data centers are unclear. While the EU has access to Nvidia GPUs and other advanced AI accelerators developed in America through a trade agreement with the U.S., Martens pointed out that acquiring hardware is only the beginning.
Also, running these data centers profitably and using them to create world-class AI products requires business model development, thorough planning, and interest from private companies, Martens noted. As a result, the AI buildout in Europe may be very gradual to avoid mistakes made by Chinese authorities.
But perhaps power supply is the biggest concern. A single gigawatt-class data center requires an enormous amount of power, but building such generation capacity takes years. Martens cautioned that Europe’s current grid infrastructure might not be ready to support such high loads, and massive upgrades will likely be needed, which will stretch schedules of the project.
Follow Tom’s Hardware on Google News to get our up-to-date news, analysis, and reviews in your feeds. Make sure to click the Follow button.
