Europe’s Troubled Bet on AI Factories

It has become the latest gold rush in Europe: mobilizing public funds to build AI factories and gigafactories. These data centers will feature high-performance servers and AI chips, cooled by industrial water systems and powered by nearby energy grids.

Even as proponents present AI factories as the continent’s best hope of catching up in the crucial AI race, the history of large state-backed European industrial projects raises doubts. Galileothe satellite navigation system, a competitor to American GPS, arrived late and beyond cost, and Gaia-Xthe offer of a European cloud, has struggled to reduce the influence of the American cloud giants.

For critics, and there are many, AI Factories fail to address the main weaknesses of AI in Europe: high energy costs, strict copyright rules and strict AI regulation. They increase potential supply but fail to stimulate what is most needed: demand. Electricity powers the chips and data centers needed to run AI models. costs up to three times US levels in Europe. Access to high-quality data is crucial for creating AI models. In the United States, fair use provides access to a vast treasure trove. In Europe, rights holders demand high payments, increasing the potential cost of data.

Europe AI Law also generates backflow. While Brussels policymakers are proud to have passed the world’s first binding legislation to mitigate the potential dangers of the new technology, AI developers are calling the effort premature — and dangerous. The new law creates uncertainty. No one really knows which applications will be labeled “high risk” and subject to significant liability risks. A movement is growing to “pause” implementation for at least a few years, but no decision has been made so far.

Despite doubts, governments and businesses across the continent are mobilizing to bid for the 20 billion euros in European funds. AI Factories target startups, researchers and enterprises to harness the three ingredients needed to create advanced AI models: computing power, data and skilled talent.

As European universities struggle to fund AI, supporters of the project highlight the potential link with research institutes. The planned AI factories will be clustered around Europe’s leading supercomputing centers – LUMI in Finland, BSC in Barcelona, ​​Cineca in Bologna and Jülich in Germany – all linked by ultra-fast networks.

It’s no surprise that businesses and national governments are rushing to profit. Who wouldn’t accept free government funding to buy chips and servers and build data centers? The European Commission said it had received a total of 76 expressions of interest across 60 different sites in 17 of the 27 EU member states.

European leaders present AI factories as symbols of digital sovereignty. Announcing the project, European Commission President Ursula von der Leyen called them proof of the continent’s determination to “become a leader in AI innovation.” Between 2018 and the third quarter of 2023, investors paid almost 32.5 billion euros in European AI companies, compared to more than 120 billion euros for American AI companies. Brussels is betting that AI Factories can fill this funding gap.

But critics remain skeptical. The AI ​​Factory location selection process led to bitter political infighting. In central and eastern Europe, the Poland-based AI Chamber and other groups have urged governments to unite behind a single regional project rather than divide resources.

The only AI chips on the market are American-made, Nvidia and AMD. European data centers are struggling to provide the cutting-edge development tools offered by American hyperscalers. Interoperability issues hamper efficiency. Each AI factory could have its own software stack, making it difficult to move workloads between them.

“Business applications require the flexibility that cloud infrastructure already provides,” says Alexandru Voica, public policy specialist at Synthesisa UK-based AI video generator startup. “With these supercomputers, you quickly reach the ceiling and you still end up depending on American cloud providers. »

The European AI factory initiative pales in comparison to the sums invested in AI infrastructure by China and the United States. Even if 20 billion euros seems like a considerable sum, the American company OpenAI has mobilized half a trillion dollars over the next four years for its Stargate project.

Europe could do better by focusing on the growing demand for AI products. Only 13% of European companies say AI is core to their business, compared to 49% in the US and 83% in China. If more Europeans used AI, then European companies could naturally arise to respond.

The European Commission spear an AI Apply strategy this month “to accelerate the use of AI in key industries and the public sector”. He plans to provide a budget of 3.5 billion euros per year. This could be more effective than investing more than five times as much funds in AI gigafactories.

Dr Anda Bologais a senior researcher in the technology policy program at the Center for European Policy Analysis (CEPA).

Bandwidth is CEPA’s online journal dedicated to advancing transatlantic technology policy cooperation. All opinions expressed on Bandwidth are those of the author alone and may not represent those of the institutions they represent or the Center for European Policy Analysis. CEPA maintains a strict policy of intellectual independence in all its projects and publications.