At GTC Paris, NVIDIA CEO Jensen Huang outlined the company’s strategy for expanding AI infrastructure across Europe. Addressing an audience both online and at the Dôme de Paris, Huang described how NVIDIA is collaborating with European governments, telecommunications companies, and cloud providers to establish AI systems built on its Blackwell architecture, alongside developments in quantum computing and robotics.
Huang presented the GB200 NVL72 platform, described as NVIDIA’s most powerful AI system to date, now in mass production. According to Huang, approximately 1,000 units are being produced weekly. The platform supports a range of uses, including sovereign AI models and quantum computing. Smaller systems, such as the DGX Spark and RTX PRO Servers, were also showcased.
NVIDIA’s approach includes the construction of both AI infrastructure for external use and proprietary AI factories for enterprise applications. The company is expanding its technology centers across Europe, with new hubs in Finland, Germany, Spain, Italy, and the United Kingdom. These centers aim to enhance AI skills and support the region’s quantum computing initiatives.
In Denmark, the CUDA-Q platform is operational on the Gefion supercomputer, enabling hybrid AI and quantum applications. CUDA-Q is also being integrated into NVIDIA’s Grace Blackwell systems, facilitating collaborative research in quantum error correction and hybrid computing.
Huang also highlighted NVIDIA Nemotron, a platform intended to assist in the creation of localized large language models. These models will be available via Perplexity, a reasoning-based search engine designed for multilingual and culturally specific AI interactions. Additional tools introduced include the NeMo Agent toolkit and a new AI Blueprint for building adaptive AI agents, with deployment support via the DGX Cloud Lepton platform.
In the industrial sector, NVIDIA plans to establish what it describes as the world’s first industrial AI cloud in Germany. Built on the Omniverse platform, this cloud is intended to support manufacturers in simulation, automation, and optimization tasks. The company is also advancing its automotive AI stack, NVIDIA DRIVE, to support autonomous vehicle deployment.
NVIDIA’s broader vision encompasses a rapid rise in inference workloads, which Huang claimed have increased from 8 million to 800 million users within two years. This growth, he said, underscores the need for specialized computing systems capable of reasoning and planning, roles NVIDIA assigns to its Blackwell-powered technologies. These systems, referred to as AI factories, are designed to produce data tokens for AI agents and robots. The event concluded with a demonstration involving a robot named Grek and a discussion of upcoming AI and robotics collaborations, including projects with DeepMind and Disney.
Photo: NVIDIA
