Infrastructure: AI’s new strategic asset in Europe

Whilst ChatGPT, Claude and Mistral are grabbing all the attention, another battle is being fought away from the spotlight: the battle over the infrastructure that makes artificial intelligence possible.

For behind every language model lies a far more tangible reality: graphics processing units (GPUs), data centres, electricity grids capable of powering thousands of servers, and colossal investments.

“There’s a lot of talk about language models, but the real strategic asset today is computing power,” says Ibrahima Sissoko, founder of the Café Crème Group. “Without computing power, even the best models are useless.”

This idea is gradually gaining ground within the industry. Innovation no longer relies solely on algorithms. It now depends on the infrastructure capable of training them, running them and meeting skyrocketing demand.

For the past two years, the European debate on AI has focused primarily on generative models and their regulation. However, the main limiting factor now lies elsewhere.

Each new generation of models requires greater computing power. Storage requirements are skyrocketing, whilst infrastructure must be upgraded at an ever-faster pace in response to rapid advances in semiconductor technology.

This development means that computing power has become a strategic resource which, in the view of some observers, is comparable to the energy networks or telecommunications infrastructure of previous industrial revolutions.

Europe is investing... but the race is global

In response to these challenges, Brussels is stepping up its efforts. AI Factories, EuroHPC, supercomputers, investment in data centres: the European Union is launching a series of initiatives to reduce its technological dependence and strengthen its digital sovereignty.

For Ibrahima Sissoko, Europe has identified the right areas for investment. “We are heading in the right direction. Investment in infrastructure is essential.” The challenge, however, is no longer merely financial. “The real shortage today is of engineers capable of building this infrastructure. Europe produces excellent graduates, but many of them then go abroad.”

In other words, digital sovereignty is not simply a matter of building more data centres. It also involves retaining the talent needed to design and operate them.

Ibrahima Sissoko, CEO, Café Crème Group

Energy: the other side of sovereignty

As AI advances, another issue comes to the fore: that of electricity. With data centres consuming ever-increasing amounts of energy, the competitiveness of future infrastructure will also depend on access to abundant and stable electricity supply.

“The foundation of artificial intelligence is computing power. And the foundation of computing power is energy,” says Ibrahima Sissoko. In his view, France has an advantage thanks to its nuclear power stations, whilst Spain could also play an increasingly important role in the development of major digital infrastructure.

Beyond the debate on technology, AI is thus becoming an energy policy issue.

Should investors switch to a different software programme?

This transformation also raises a key question for the financial markets. For more than a decade, European venture capital has focused primarily on software companies. The business models were not particularly capital-intensive and allowed for relatively rapid exits. Infrastructure tells a different story.

Building computing capacity or funding data centres involves substantial investment and much longer timeframes before a return on investment. “We’ve been funding the software for 15 years. Now we need to fund the infrastructure,” argues Ibrahima Sissoko.

This development could lead investors to regard digital infrastructure as a new asset class, on a par with energy or transport infrastructure.

Is there a role for Luxembourg’s financial centre to play?

This line of thinking naturally opens up opportunities for Luxembourg. As Europe’s leading centre for investment funds, the Grand Duchy already possesses recognised expertise in infrastructure financing, the energy transition and sustainable bonds.

Could the next step be to fund the digital infrastructure needed for the AI economy? For Ibrahima Sissoko, the issue goes beyond simply providing capital. He also calls for the development of new financial instruments tailored to these long-term projects, capable of supporting companies that build the infrastructure rather than those that solely develop software.

This conviction also guides the strategy of Café Crème, which is currently engaged in a series of acquisitions aimed at consolidating its growth and preparing for its listing on Euronext.

After the models, the foundations

Artificial intelligence is often portrayed as a software revolution. It is also, and perhaps above all, an industrial revolution. As models become more powerful, the real competition could shift towards those who control the infrastructure that powers them: computing power, energy, networks and financing capabilities. For Europe, the challenge is no longer simply to develop AI leaders.

The task now is to lay the foundations that will enable these champions to thrive in the long term.