28 French companies including Iliad, EDF, Orange, Hugging Face and Kyutai announce the AION consortium to build AI gigafactories in Europe. Objective: 10 billion euros of investment and 1 gigawatt of electrical power to guarantee France's artificial sovereignty.
The European Commission is launching a call for projects to create AI gigafactories on the continent. AION responds with the largest industrial coalition ever gathered in France around artificial intelligence. "Every company is going to use AI, the question is: where do we build it?" sums up Damien Lucas, CEO of Scaleway (an Iliad subsidiary).
The context
Generative AI relies on ever-larger models, which require massive computing resources. These resources are today controlled by a few American giants: Microsoft (Azure), Amazon (AWS), Google (Google Cloud). Europe therefore depends on these infrastructures to run its own models and its own AI applications.
The European Commission has decided to act by launching an AI gigafactory program on the continent. The idea: to create a sovereign infrastructure that allows European companies to access the power they need without going through American servers. The budget planned for the whole of Europe is counted in billions of euros. France, with AION, intends to be among the selected countries.
The current problem for French SMBs is threefold: the cost of AI computing power, dependence on American infrastructure, and the risk of a technological gap with American companies that have easier access to these resources. AION aims to address these three points.
The AION consortium in detail
The AION consortium brings together 28 French companies representing the entire artificial intelligence value chain:
Telecoms and cloud
Iliad (Free), Orange, Scaleway (Iliad's cloud subsidiary) — these companies bring the network infrastructure and data center expertise. Scaleway coordinates the project with Damien Lucas. Planned investment: 4 billion euros for Iliad alone.
Energy
EDF — guarantees the electricity supply of these ultra-consuming sites. Without accessible, low-cost energy, there is no viable gigafactory.
Consulting and integration
Artefact, Capgemini — bring deployment and consulting expertise to help companies use these infrastructures.
Hardware
Bull (Atos group) — designs and delivers the supercomputers needed to train the models.
Finance
Ardian, Crédit Agricole — the investment funds mobilized to finance the project. In total, the investment exceeds 10 billion euros over several years.
European AI start-ups
Hugging Face, Kyutai, LightOn — these flagships of French and European AI bring their expertise in models, training and deployment. They will be the first users of this infrastructure.
The concrete stakes for French SMBs
For a French SMB that wants to integrate AI into its processes, the problem is not only the cost of SaaS subscriptions. It is also dependence on American infrastructure for hosting sensitive data, the cost of computing power for custom models, and the risk that the AI tools available in Europe remain behind the American versions.
With AION, French SMBs could use AI services hosted on European territory, without transferring data outside Europe. Important for regulated sectors (healthcare, finance, defense).
A European gigafactory means increased competition with American cloud providers, which can drive prices down for all companies. Lower cost for custom models.
AION brings together AI start-ups, large groups and investment funds. SMBs that join this ecosystem can benefit from collaborations, technology transfers and visibility with major accounts.
Beyond the technology, the AION project will generate thousands of direct jobs (data center, maintenance, supervision) and indirect ones (integration, development, consulting) on French territory.
The challenges to overcome
The energy question
An AI gigafactory consumes as much electricity as a medium-sized city. Finding sites with access to power of 100 megawatts to 1 gigawatt, with suitable connections, is a massive infrastructure challenge. EDF is involved to address this point, but the connection lead time can take years.
Construction timelines
Building a gigafactory takes time. The European call for projects must deliver its decision in the coming months, but building a data center of this scale takes between 2 and 4 years. Europe is behind the United States and China in terms of installed capacity.
International competition
The other European consortia (Germany, Netherlands, Finland) are also in the running. The European Commission will not finance every project. France must prove that its consortium is the most credible and the most able to deliver an operational infrastructure quickly.
Frequently asked questions
What is an AI gigafactory?
An AI gigafactory is a very large capacity data center, designed to train and run artificial intelligence models at industrial scale. Unlike classic data centers, it requires massive electrical power (up to 1 gigawatt) and extreme cooling infrastructure.
Why should French SMBs care about this consortium?
AION is not limited to large companies. The consortium includes AI start-ups (Hugging Face, Kyutai), investment funds and service providers. French SMBs will be able to access this infrastructure through sovereign cloud offerings, creating an ecosystem where smaller players too can benefit from the computing power reserved today for the American giants.
What is the timeline for this project?
The European Commission's call for projects is due to select the winning consortia in the coming months. AION aims to deploy an initial capacity of 100 megawatts as early as 2027, with gradual scaling up to 1 gigawatt. France intends to be selected to host at least one gigafactory on its territory.
What impact for companies working with AI day to day?
In the short term, little direct change. Current AI tools (ChatGPT, Claude, Gemini) will continue to work. In the medium term (2027-2030), the emergence of a sovereign infrastructure means GDPR-compliant AI offerings, hosted in France, with potentially lower costs thanks to competition. This is good news for SMBs that hesitate to adopt AI for compliance reasons.