NVIDIA CEO Jensen Huang has begun a tour of Europe, announcing on Tuesday plans to invest an additional $2 billion in the region, alongside new funding for four French startups. During his visit, he met with French President Emmanuel Macron to discuss NVIDIA’s investment strategy in France, as well as topics such as open-source software and AI safety.
Huang noted that this $2 billion is in addition to the $8 billion NVIDIA has already invested in Europe over the past year. Paris served as the first stop on his European tour, with subsequent visits planned for London and Berlin. In Paris, he remarked, “Europe must join this race.”
As the world’s largest manufacturer of AI chips, NVIDIA has committed hundreds of billions of dollars globally, aiming to accelerate the adoption of AI and drive demand for its own products. However, there is growing pushback against AI—ranging from concerns about models spiraling out of control to the environmental impact of energy-intensive data centers—leading to increased calls to curb the development of the technology.
Huang opposes slowing down AI development, arguing that the technology can both fuel economic growth and be advanced safely.
In contrast, following a series of high-profile safety incidents involving AI agents, leaders of major AI companies—such as Anthropic CEO Dario Amodei and OpenAI CEO Sam Altman—are currently weighing how best to manage the pace of AI development.
Huang is also advocating for open-source models, which can be more cost-effective and easier to customize. He hopes this will help Europe catch up with the more advanced AI technology found in the United States, while also reducing reliance on products controlled by a handful of large corporations.
“Open models are important because they can be adapted to different languages and specific sovereign requirements,” he said.
NVIDIA has invested in the French AI startup Mistral AI and supplies the AI accelerator chips required for its data centers. Mistral develops “open-weight” models—making its model parameters publicly available—which allows other developers to use and modify them freely. The company’s latest model was trained using 4,000 Grace Blackwell chips.