Digital Independence: French OVHcloud to Release Sovereign MoE AI Model

Europe is accelerating the creation of an independent sovereign artificial intelligence stack. On October 20, 2026, OVHcloud CEO Octave Klaba announced plans to release the company's proprietary Mixture of Experts (MoE) architecture in early 2027. The model will be trained on the state-of-the-art Jupiter supercomputer.

This move symbolizes a major European business refusing the role of mere "hosting" for American neural networks. Possessing 110 MW of available cloud capacity, OVHcloud is integrating its own model directly into the OVHai Workspace platform for secure processing of corporate email and documents. The B2G sector and European enterprise are acquiring a closed ecosystem physically isolated from US jurisdiction (Cloud Act). OVHcloud's strategy proves: in the AI macroeconomy, the winners are vertically integrated providers capable of guaranteeing zero data leakage at both the hardware and algorithmic levels.

Source: OVHcloud / Reuters