France’s OVHcloud is preparing to launch a new suite of frontier large language models, signaling its ambition to become Europe’s next significant player in AI development alongside startups like Mistral.
- OVHcloud plans multiple specialized frontier AI models rather than one general system.
- Development costs have fallen significantly, enabling new entrants like OVHcloud to compete.
- Models will be open-sourced after reaching sufficient performance levels.
What happened
OVHcloud revealed plans to train and develop frontier AI models, the most advanced category of large language models built from the ground up using extensive datasets and computing power. This marks a strategic pivot for the company as it seeks to position itself as a challenger to existing European AI startups such as Mistral and compete with dominant US and Chinese technology providers.
CEO Octave Klaba shared these plans at the VivaTech conference, explaining that mastering AI technology is essential for OVHcloud’s future viability. He noted the significantly reduced costs for building such models due to improvements in chips, training methodologies, and synthetic data generation, bringing development budgets down from around 1 billion euros to closer to 150-200 million euros.
Why it matters
OVHcloud's investment in frontier AI models addresses growing European demand for alternatives to dominant US and Chinese systems amid concerns about supply chain sovereignty and data privacy. The company’s commitment to not using client data for model training and plans to open-source these new models underscore a transparent approach that aligns with European values.
This development signals the emergence of a 'second wave' in the AI model industry, where new players build on foundational work by firms like OpenAI and Anthropic. OVHcloud's use of Jupiter, Europe’s fastest supercomputer, and the acquisition of AI startup DragonLLM suggest strong technical foundations for these efforts.
What to watch next
OVHcloud is currently completing pre-training phases and aims to roll out a family of models targeting different applications rather than a single all-purpose system. Observers should monitor performance benchmarks and the timeline for when these models become open-source, as this will influence adoption and further innovation within Europe’s AI ecosystem.
Competitive responses from other European AI startups and cloud providers will also be key to watch, as OVHcloud’s entry may accelerate efforts to establish European AI model independence and diversify the global AI landscape.