GENCI (Grand Équipement National de Calcul Intensif) has expanded its mission beyond traditional numerical simulations to include big data processing and quantum computing. In 2025, the organization supported over 2,400 research projects, with 1,700 of them focused on artificial intelligence. As a public institution, GENCI provides its computing resources free of charge to researchers in academia and industry, as long as the results are shared through open research practices. This approach supports scientific innovation while ensuring that France remains independent in its technological choices. Stéphane Requena, a representative of GENCI, explains that the organization’s goal is to support both French and European research by offering a strong alternative to foreign commercial computing infrastructures. This helps maintain technological independence and reduce reliance on foreign systems. GENCI operates through three computing centers. The CEA’s Very Large Computing Center (TGCC) focuses on traditional simulations using central processing units (CPUs), while also incorporating quantum computing capabilities from French companies like Pasqal and Quandela. At IDRIS (CNRS), the Jean Zay supercomputer uses 4,000 Nvidia GPUs, while the Adastra center (CINES) in Montpellier uses 2,000 AMD GPUs. This mix of American suppliers helps avoid dependency on any single provider and strengthens resilience in the face of global supply chain issues. Technically, GENCI’s computing systems are designed to handle the different precision needs of various tasks. For example, material simulations require high precision (FP64), while AI models can function with lower precision. The Montpellier site recently received a security certification for handling health data, allowing it to process sensitive information within national borders. The Alice Recoque supercomputer, set to be launched at the TGCC next year, will be the most powerful in Europe for research. It is part of the EuroHPC initiative, which includes 36 European countries. The project represents a major investment, shifting from a typical budget of 100 million euros to 500 million euros, reflecting the growing importance of AI in national strategies. Alice Recoque will be built using a European technology stack aimed at achieving self-sufficiency. It will include new AMD MI300 GPUs and processors developed by the French company SiPearl. The system will be connected using the BullSequana interconnection technology, produced by the French manufacturer Bull (now part of Atos). According to a GENCI expert, this leap in computing power is essential for creating high-resolution digital models of Earth or training complex language models that go beyond current capabilities. The AI Factory France initiative aims to transform computing power into a comprehensive service ecosystem, offering a "One-Stop Shop" for AI development. In partnership with Inria, French Tech, Station F, and the Hub France IA, the project provides access to critical datasets, expert guidance on optimizing system architecture, and a catalog of 800 open-source models. The initiative encourages collaboration between public and private sectors, allowing startups to use GENCI resources for free during the research and development phase, then transition to commercial infrastructure for deployment. This transition is supported by partners like Scaleway and OVHcloud. The IA Cluster plan aims to train 100,000 people to address the shortage of specialists in deploying large-scale AI models. GENCI highlights the startup Pleias, which provides services to the RATP (Paris public transport authority), as an example of the practical benefits of its infrastructure. The company developed a model for identifying incidents in the metro using only data from Île-de-France. Initially using a Mistral model, Pleias found it only moderately effective for their needs. By switching to GENCI’s infrastructure, they were able to create a more efficient solution. The deployment of the AI model was done quickly, with the CTO of GENCI noting that the model was deployed "by pressing a button" to the Scaleway platform during the summer. This approach is now being replicated across 12 different sectors, including finance, health, aerospace, and earth sciences. Handling sensitive data is a key aspect of maintaining technological sovereignty. At CINES in Montpellier, strict isolation of computing processes, reinforced by HDS certification, ensures that patient data can be processed without the risk of leaking to non-European systems. However, the success of the AI Factory will depend on its ability to raise awareness among entrepreneurs. Stéphane Requena acknowledges that many startup founders are still unaware of the availability of these free resources. To address this, GENCI is building a network of ambassadors within the French Tech and competitiveness clusters. Events like Big Data Paris are also used to increase the initiative’s visibility and reach.