Highlights from NVIDIA Keynote and GTC Paris 2025

AI Technology & Infrastructure - NVIDIA is accelerating AI across Europe, focusing on agentic AI that can understand, reason, and act, envisioning AI factories as engines of transformation [2] - NVIDIA partners with European technology leaders to increase AI infrastructure investment by 10x over the next two years, supporting digital sovereignty, economic growth, and industrial competitiveness [3] - NVIDIA's Blackwell platform is in full production, with new benchmarks showcasing GB200 integrated with cuQuantum libraries and CUDA-Q platform to accelerate quantum algorithms [4] - Grace Blackwell MBL72 is designed to meet the computing demands of advanced reasoning AI models [5] - New RTX Pro servers, with integrated ConnectX-8 networking and optimized for NVIDIA AI software, offer a foundation for enterprise-scale generative AI and industrial digitalization [5] - NVIDIA introduced DGX Cloud Lepton, a unified AI platform designed to connect developers to global accelerated compute with multicloud, multi-tenant, and global scaling capabilities [6] AI Model Development & Deployment - NVIDIA supports Europe's model builders and cloud providers in developing sovereign LLMs using NVIDIA and NeMoTron, optimized for cost and cultural relevance [7] - These models will soon be available on Perplexity and deployable via NVIDIA NIM's microservices [7] Training & Innovation - New NVIDIA AI technology centers are training talent, supporting researchers, and seeding innovation [4] - Thousands of attendees at GTC Paris gained practical tools and a vision of what's next through sessions, demos, and technical workshops on NIM, NeMo, and TensorRT [8]