Key facts
- OVHcloud plans to develop frontier AI models, aiming to compete with Mistral.
- The company's CEO cited shifting economics and the need for technological self-sufficiency.
- The development is seen as a response to US restrictions on AI model access and Europe's reliance on foreign tech.
- OVHcloud intends to launch a family of models and eventually open-source them.
- A model has been pre-trained using Jupiter, Europe's fastest supercomputer.
OVHcloud, Europe's largest cloud provider, plans to develop its own frontier AI models, positioning itself as a potential European challenger to firms like Mistral. CEO Octave Klaba stated that mastering this technology is crucial for the company's future, especially amid increasing calls for European AI sovereignty and recent US restrictions on foreign access to advanced AI models.
Klaba indicated that the economics of developing such cutting-edge systems have shifted, with projects that once cost around 1 billion euros now potentially achievable for 150 million to 200 million euros. He described the industry as entering a "second wave" and confirmed that OVHcloud would not use client data for training its models. The company intends to launch a family of models, as no single system can perform all tasks, and aims to open-source them once they reach sufficient performance.
Pre-training has been completed on a model using Jupiter, Europe's fastest supercomputer, though detailed performance claims are not yet available. This move by OVHcloud aligns with broader European efforts, highlighted at the VivaTech conference by France and Germany, to build independent technological capabilities and reduce reliance on US and Chinese AI systems.