Scaleway Expands in Scandinavia, Mistral Unveils ML4

Scaleway is preparing a GPU Zone in Copenhagen with Nvidia and a supercomputer with AMD near Stockholm, claiming a 100% European offering. Mistral, for its part, is highlighting Mistral Large 4, a natively multimodal model with 1 trillion parameters offered in open-weight.

Scaleway Equips Copenhagen and Prepares a Supercomputer in Stockholm

Scaleway plans to install the NVIDIA HGX B300 platform, designed for training large language models, in its new GPU Zone in Copenhagen. This center is set to provide massive GPU resources for training and inference of artificial intelligence models. In addition to Nvidia, Scaleway also announces a partnership with AMD to establish a supercomputer specialized in AI in the Stockholm region of Sweden, based on AMD Instinct MI355X GPUs. According to Lisa Su, this first supercomputer from Scaleway outside of France will provide AI companies and startups with the computing power necessary to execute complex tasks and support their growth. With this new cluster, Scaleway aims to meet the growing demand from European businesses. Damien Lucas, CEO of Scaleway, believes that access to high-performance computing for AI is becoming a strategic priority for European organizations, which want these capabilities to be provided by trusted players.

Positioning as 100% European and Avoiding Extraterritorial Jurisdictions

Scaleway claims to structure its operations to strengthen its positioning as a 100% European cloud provider and to escape extraterritorial jurisdictions, particularly the U.S. Cloud Act. Northern Europe is seen as a favorable environment for establishing infrastructure, thanks to a cooler climate that reduces cooling needs, direct access to low-carbon electricity, and a technology ecosystem located within the European Union, ensuring legal compliance. These advantages make the region a preferred destination for European AI players. In this context, Mistral announced earlier this year a €1.2 billion investment in a data center in Sweden, which is expected to provide an additional capacity of 23 megawatts. These initiatives respond to the growing demand from European businesses for AI.

Mistral's Technical Bet: ML4 in Open-Weight and 1 Trillion Parameters

Mistral Large 4 is a multimodal model from its inception, comprising 1 trillion parameters. Distributed in open-weight, it offers companies the ability to download the model weights and run it on their own servers, without disclosing the entire dataset or the training method. Mistral claims it is the most performant open-weight model developed in Europe or the United States, and synthetic rankings place it among the most advanced open models globally, although the majority of Chinese labs remain at the forefront. This model is particularly valued for applications in cybersecurity, the financial sector, and industrial production. It was trained on 3,800 Nvidia Grace Blackwell GPUs in Mistral's European data centers. "We wanted to put all our science into this model," said Pierre Stock, head of scientific research and one of Mistral's first employees, during his presentation alongside Aude Durand. The model is nicknamed "Le Chonk."

An ai-PULSE Focused on Open Source and Sovereignty

The 2026 edition of ai-PULSE, held at Station F, placed open source at the center of discussions, advocating for European sovereignty. The event took place without the presence of figures like Xavier Niel, Eric Schmidt, or Yann LeCun, but gathered players from the AI ecosystem advocating for an approach different from those of the United States and China, while acknowledging the difficulty of breaking free from certain American giants in the semiconductor field. Mistral presented its new model there, detailed by Pierre Stock, head of scientific research and one of the company's first employees. In a highly competitive sector, Scaleway will need to continue strengthening its computing capabilities.