EU’s Ambitious AI Gigafactory Plan Faces Challenges
The European Commission plans to invest $20 billion in building four “AI gigafactories” as part of its strategy to compete with the United States and China in the artificial intelligence sector. However, industry experts are questioning the practicality of the project, given the existing challenges.
The plan, unveiled last month by European Commission President Ursula von der Leyen, involves constructing large public access data centers. Experts believe the project will face difficulties, especially in obtaining vital components such as chips. Other concerns include finding appropriate sites and ensuring a sufficient power supply.
“Even if we would build such a big computing factory in Europe, and even if we would train a model on that infrastructure, once it’s ready, what do we do with it?” asked Bertin Martens of the economic think tank Bruegel, highlighting the problem of creating AI models within Europe.
The Commission hopes these new AI infrastructure investments will encourage the growth of local firms, similar to France’s Nvidia-backed Mistral start-up. The aim is to create AI models that meet EU standards for safety and data protection, which are stricter than those in the U.S. or China. However, the absence of major European cloud service companies like Google and Amazon, or firms with millions of paying customers such as OpenAI, poses substantial risks for such a large-scale hardware venture.
Europe’s Response to the ‘Stargate’ Plan
This gigafactory plan is part of Europe’s broader response to the Draghi report on competitiveness. The report advocated for significant investments and a more proactive industrial policy. Von der Leyen unveiled details during the February 11 AI summit in Paris, as part of InvestAI, Europe’s 200 billion euro ($216.92 billion) response to the $500 billion U.S. Stargate plan. Von der Leyen views the gigafactories as a “public-private partnership … (that) will enable all our scientists and companies – not just the biggest – to develop the most advanced very large models needed to make Europe an AI continent.”
Financing will come from a new 20 billion-euro fund, drawing money from existing EU programs and member states, with the European Investment Bank also participating. Von der Leyen stated that each gigafactory would house 100,000 “cutting-edge” chips, making them more than four times larger than the EU’s largest supercomputer currently under construction: the Jupiter project in Germany. Nvidia, the leading U.S. chipmaker, sells the critical GPU chips for AI training at approximately $40,000 each, implying a multi-billion euro price tag for each gigafactory.
Despite the substantial investment, the plan falls short of projects announced by U.S. firms. Meta, the parent company of Facebook, is spending $10 billion to build a 1.3 million GPU facility in Louisiana, powered by 1.5 gigawatts of electricity.
Key Hurdles: Chip Supply and Electricity
Kevin Restivo, a data center expert at real estate consultancy CBRE, noted that gigafactories in Europe face challenges similar to those of private projects: difficulty in securing scarce Nvidia chips and the need for large-scale electricity. Moreover, the U.S. government, under former President Joe Biden, restricted access to AI chips to prevent the construction of gigafactories in specific countries. It’s not clear whether the new administration will maintain this restriction.
Martens expressed doubt about the wisdom of using public money to enter an AI spending race, considering the relatively short lifespan of data centers. He pointed out, “The lifetime of such factories, before you have to write it off and buy new Nvidia chips, is about … a year and a half.”
Shifting Trends in AI Development
Meanwhile, the emergence of Chinese AI model Deepseek raises questions about the trend of using less computing power to train AI models. Some experts believe that investments should be directed toward AI applications, which require distinct types of chips. Previous technology infrastructure support programs, such as the 2023 Chips Act, failed to meet the goals of bringing cutting-edge chip manufacturing to Europe or reach 20% of global production, although it did lead to investments in new factories for automotive chips.
Alongside the gigafactory plan, the Commission is also upgrading 12 scientific supercomputer centers to transform them into AI factories. Kimmo Koski, managing director of Finland’s LUMI supercomputer, said that it is not yet clear how AI gigafactories will differ other than in size. He added, “In my understanding, it relates to pushing industry use further. That would be “an innovation in Europe, a very welcome event of course.”
Supercomputers like LUMI are already used for diverse projects, from machine learning to climate modeling. Koski highlighted the example of Silo AI, a Finnish firm that used LUMI to develop large language AI models before it was acquired by U.S. chipmaker AMD last year for $665 million. Potential beneficiaries of the supercomputer expansion include European chipmakers that make non-GPU chips: Infineon of Germany and ST Microelectronics of France, as well as start-ups like SiPearl of France and AxeleraAI of the Netherlands.
