Issue #26 - August 14, 2026
Europe launches AI Gigafactories and data center sovereignty rules to stay in the race
Sébastien Louradour
8/14/20263 min read
Europe launches AI Gigafactories and data center sovereignty rules to stay in the race
Over the past few years, the level of investments led by US tech companies has been astonishing. Most of these investments are intended to fund data centers to anticipate a growing demand in use of GenAI over the next decade, and ultimately a potential revolution in the knowledge worker economy. In a recent FT article by Robin Wigglesworth, in Alphaville, the journalist has listed the purchase commitments of the main US hyperscalers for Q2 2026, and has already counted it for $1.5tn. This includes commitments to secure chips and power for the data centres that the company is building. In the meantime, Europe has started to adopt AI policies that are both meant to promote sovereign use of data in Europe and promote the launch of AI gigafactories through a call for proposal with a €10bn envelope, and a target to reach €30bn with private capital added. Other commitments have been announced over the past few months, such as Softbank committing to invest €75bn in France, or another in Sweden led by a Business Consortium (Wallenberg Investments, AstraZeneca, Ericsson, Saab, SEB) .
Brussels has been active to position AI at the core of its digital agenda to demonstrate willingness to position Europe as an AI leader, and able to compete with less reliance on foreign technology. The EU Cloud and AI Development Act (CADA) introduced two months ago by the European Commission aims at setting up a tiered-level approach for cloud providers. Each cloud provider would be given a level of sovereignty from 1 (the lowest) to 4 (the highest) and each member state would determine what level of sovereignty is required for specific public sector activities. Private entities engaged in sensitive activities could also use this classification to pick its cloud providers. By doing so, the EU is creating a EU sovereignty stamp aimed at promoting its own players. Overall, the tiered categorisation of CADA creates uncertainty for non-EU players, and might impact their willingness to invest in an environment where hostility against giant US players is growing. CADA also aims at encouraging the deployment of large data-centers across the EU, through the creation of data center acceleration zones, and the designation of strategic data center projects. That leaves the door open to EU players to invest in the continent. But then comes the question of capital access to do so.
The sense of urgency has been repeated multiple times by Mistral’s CEO Arthur Mensch, who has been worried that Europe goes on too little and too late. Mistral has advocated for further foreign partnerships in Europe to ultimately strengthen its sovereignty, and has recently signed a deal with Microsoft where the US company commits to use Mistral’s data centers to run its models. But the ongoing policy agenda in Brussels suggests his fears may prove right already. The call for gigafactory consortia seems to emerge from the idea that the EU needs to activate the pump itself to prevent a form of market failure. In comparison, US private companies have decided alone to commit hundreds of billions of dollars, without any form of Federal government involvement. What is to fear is a very slow and bureaucratic process to launch AI gigafactories while they are already popping-up in the US.
At the current pace of AI progress, what EU tech companies need too is the capacity to train models and keep up with fierce global competition. To enable this, the EU needs to clarify the regime to use data for model training, which would in exchange further incentivise investments in Europe. This was the initial idea of the digital omnibus, but the conversation has stalled with the question of data privacy remaining central to EU policymakers. This debate happens in a context where limiting access to data for model training could lead to further gaps with US labs, and expose the hard choices policymakers are unwilling to compromise on : trading privacy over competitiveness. But one element policymakers should also acknowledge is competitiveness is fundamental to maintain Europe’s welfare model, with AI being the best opportunity to improve competitiveness in a knowledge-worker based economy. Not acting now could ultimately weigh on Europe’s capacity to sustain its generous retirement and social security models moving forward.