The European Commission is raising $20 billion to build four “AI gigafactories” (large-scale factories that produce high-tech products, especially electronic components and microchips), as part of a strategy to help Europe catch up with the US and China in the field of artificial intelligence (AI). However, some industry experts have expressed concerns about whether building these facilities really makes sense.

The plan to build these massive data centers, first announced by European Commission President Ursula von der Leyen last month, will face many challenges, from procuring the necessary microchips to finding suitable locations and power sources.

“ Even if we build a large-scale technology factory in Europe and we train a model on that infrastructure, when the model is ready, what do we do with it ?” said Bertin Martens, from the Bruegel economic think tank.

The planned gigafactories will be financed through a €20 billion fund, with funding drawn from existing EU and member state programmes. Illustration photo

It’s a chicken-or-egg problem. The hope is that local startups like French AI developer Mistral, backed by Nvidia, will develop and use these facilities to create AI models that comply with Europe’s stricter data protection and safety rules than those in the US and China.

However, in a European context lacking large cloud service businesses like Google and Amazon, or companies with millions of paying customers like OpenAI (developer of the ChatGPT tool ), building such large-scale hardware is a risky project.

The plan to build gigafactories is Europe’s response to the Draghi Report, which recommended strong investments and a more proactive industrial policy. Ursula von der Leyen detailed the plan for the first time at the AI ​​Summit in Paris on February 11, as part of InvestAI, an EU initiative to boost investment in artificial intelligence and digital technologies in Europe, with a budget of €200 billion ($216.92 billion).

Ursula von der Leyen said the gigafactories would be public-private partnerships that would enable all scientists and companies – not just the biggest ones – to develop very large and advanced AI models, making Europe a leading continent in AI. The gigafactories would be financed through a new €20 billion fund, with funding from existing EU and member state programmes. The European Investment Bank would be involved in the financing.

Each gigafactory will contain 100,000 advanced chips, making them four times larger than the largest supercomputer currently under construction in the EU, Ursula von der Leyen said. The advanced GPU chips, supplied by US manufacturer Nvidia, cost around $40,000 each, meaning each gigafactory could be worth several billion euros.

Gigafactories will face similar problems to private projects in Europe, including difficulties in procuring Nvidia chips and a lack of power sources to meet the scale required, said Kevin Restivo, a partner at data center consultancy CBRE.