OVHcloud plans frontier AI models as Europe seeks alternatives to US and Chinese systems

OVHcloud Enters Frontier AI Development Race

France's OVHcloud is making a strategic pivot from its core cloud infrastructure business into frontier AI model development, a move that positions the European cloud provider as a potential challenger to Mistral AI and marks a significant step in Europe's quest for AI sovereignty 1

1

. The company plans to train a family of advanced large language models from scratch and eventually release them as open-source AI models once performance targets are met, CEO Octave Klaba announced at the VivaTech conference 2

2

.

Dramatic Cost Reduction Makes AI Model Development Accessible

The economics of developing OVHcloud frontier AI models have shifted dramatically, making the venture financially viable for Europe's second LLM player. Klaba told Reuters that what might once have cost approximately 1 billion euros can now be attempted for 150 million to 200 million euros, thanks to advances in chips, training techniques, and synthetic data 2

2

. This cost reduction represents a fundamental change in the accessibility of frontier AI technology for companies seeking alternatives to US models and US and Chinese AI systems.

Strategic Shift Driven by Existential Concerns

The move reflects deeper concerns about technological independence and long-term viability in the AI era. "It became quite clear to us that if we don't master this technology, we can't guarantee our future," Klaba stated, highlighting the existential nature of AI model development for cloud providers 2

2

. The urgency has intensified following recent disruptions, including Anthropic's abrupt switch-off of top-tier models, which underscored the risks of dependence on external AI technology 2

2

.

Multi-Model Strategy and Current Progress

OVHcloud plans to launch a family of models rather than a single system, recognizing that specialized models serve different purposes. "We can clearly see that the major players release multiple models, because each model is built for something specific," Klaba explained. "There's no one model that does all the magic alone" 2

2

. The company has already completed pre-training on a model using Jupiter, Europe's fastest supercomputer, leveraging technology from DragonLLM, a recently acquired startup 2

2

. Importantly, Klaba emphasized that OVHcloud would not use client data to train its models, addressing privacy concerns central to European AI development 2

2

.

European AI Landscape and Competitive Positioning

This initiative puts OVHcloud in direct comparison with Mistral AI, the Paris-based model developer that has emerged as Europe's most visible challenger to US AI labs 1

1

. Klaba described the industry as entering a "second wave," with new entrants building on groundwork laid by firms such as OpenAI, Anthropic, and Mistral 2

2

. However, analysts suggest that while training costs have decreased, the bigger challenge may be keeping these systems updated without undermining sovereignty or commercial viability 1

1

. As governments and companies across the continent seek alternatives to US and Chinese AI systems, OVHcloud's success could determine whether European AI can establish genuine independence in this critical technology domain.