OVHcloud teams up with Qwant and Shadow for a Europe-hosted AI offering
Announced on 19 November 2026 in Roubaix, during the OVHcloud AI Summit 2026, the agreement brings together four links of the artificial intelligence chain: the web index, the models, the compute infrastructure and the uses.
The four companies present this set as an “end-to-end” AI offering, fully hosted and operated in Europe.
A European web index feeds OVHcloud's models and agents
The press release describes the operation as a “technology and commercial partnership”. It brings together OVHcloud, the search engine Qwant, Shadow, which operates the Blade GPU platform, and European Search Perspective (EUSP), a joint venture created in 2024 by Qwant and Ecosia to develop a web index.
The latter serves as the foundation for Staan, Qwant's web search API. The OVHai LLM models, accessible via AI Endpoints, and the agents built on AI Agents, presented as OVHcloud's new agentic platform, rely on this index to query the web and provide sourced, up-to-date answers.
Staan comes in three APIs (Web Search, Web Search for AI and AI Answer), complemented by the Fetch API, which retrieves content intended for RAG processing pipelines and agents.
The partnership links companies with common executives. Boris Lecoeur serves as chief executive of Shadow, Qwant and EUSP, while Shadow belongs to the Synfonium group, founded by Octave and Miroslaw Klaba. Octave Klaba is also the chief executive of OVHcloud.
Model catalogues shared between OVHcloud and Blade
The commercial component relies on cross-distribution. Blade's Models and AI Fabric offerings now include the large open-weight models from OVHcloud's AI Endpoints (more than 70 billion parameters), alongside the Olala models (general-purpose, translation and summarisation) and Gladia's transcription models.
Conversely, AI Endpoints hosts compact open-weight models running on Shadow's GPUs, intended for low-cost inference.
Each customer keeps its usual provider as its single point of entry, while the models remain hosted and run in OVHcloud's European data centres. The mechanism thus widens each partner's offering without moving the commercial relationship with the customer.
Qwant's flash answer, the first quantified use of the alliance
Displayed at the top of Qwant's results, the flash answer now runs on Olala mini 7A1B, the summarisation model of OVHai LLM, executed on Blade. According to the press release, this model handles 3 million requests per day with three times fewer resources than the previous model, for answers 20% faster.
OVHai LLM states that it relied on Blade AI Fabric, Blade's serverless offering, for the reinforcement learning phase of its new Olala model family. Blade also hosts Olala's public trial space, which allows the model to be tested in a browser before moving to the API.
For Boris Lecoeur, “Europe has the index, the models, the GPUs and the uses. What it lacked was connecting them”. Olivier Debeugny, chief executive and founder of OVHai LLM, stresses that “reinforcement learning is the step that takes an LLM from good to useful”.