Nvidia and artificial intelligence search company Perplexity have announced a collaboration with more than a dozen European and Middle Eastern organisations to develop sovereign large language models tailored to local languages and cultures.
The chip giant is working with model builders including Barcelona Supercomputing Center, Bielik.AI, H Company, LightOn, and the Technology Innovation Institute, among others, to optimise their AI models using NVIDIA Nemotron techniques. The partnership aims to reduce operational costs whilst improving accuracy for enterprise AI applications.
“Europe’s diversity is its superpower — an engine of creativity and innovation,” said Jensen Huang, founder and chief executive officer of NVIDIA. “Together with Europe’s model builders and cloud providers, we’re building an AI ecosystem where intelligence is developed and served locally to provide a foundation for Europe to thrive in the age of AI — transforming every industry across the region.”
The initiative will support the European Union’s 24 official languages, with several models specialising in national languages and cultures. Partners include H Company and LightOn in France, Dicta in Israel, Domyn in Italy, Bielik.AI in Poland, and organisations in Spain, Sweden, the United Arab Emirates, and the United Kingdom.
These optimised models will be integrated into Perplexity’s AI-powered answer engine, which currently handles over 150 million questions per week. The platform will allow European enterprises to access locally-developed AI models that reflect regional linguistic and cultural nuances.
“Perplexity’s goal is to provide accurate, trustworthy answers to any question from any person, wherever they are,” said Aravind Srinivas, co-founder and chief executive officer of Perplexity. “Bringing Nvidia-optimised sovereign AI models to Perplexity empowers innovation in Europe with AI built and running in the region.”
The models will be deployed through Nvidia’s DGX Cloud Lepton marketplace, running on infrastructure provided by European cloud partners including Nebius, Nscale, and Fluidstack. Companies will also be able to fine-tune the sovereign models through a new Hugging Face integration with DGX Cloud Lepton.
Kari Briski, Nvidia vice president of generative AI software for enterprise, explained the company’s approach to addressing language barriers in AI development. “We’re doing a lot of synthetic data generation to bring to these low-resource languages and translating our reasoning data so that they can train on it,” she said.
The announcement comes as Nvidia continues expanding its international partnerships, including recent deals in the United Arab Emirates and Taiwan. For Perplexity, which was recently valued at $14 billion, the collaboration represents part of its strategy to become a model-agnostic platform serving diverse global markets.
The first distilled models from Europe’s model builders are expected to become available later this year.
