At CES 2026, NVIDIA founder and CEO Jensen Huang opened the event at the Fontainebleau Las Vegas with a powerful vision of AI transforming every industry and device. “Computing has been fundamentally reshaped by accelerated computing and artificial intelligence,” Huang said. “Over the past decade, about $10 trillion in computing infrastructure is now being modernized to this new paradigm.” Huang unveiled the NVIDIA Rubin platform, the company’s first extreme-codesigned, six-chip AI system now in full production. Named after pioneering astronomer Vera Rubin, Rubin is the successor to the Blackwell architecture and represents a major leap in AI scalability. Designed from the data center outward, the platform integrates chips, trays, racks, networking, storage, and software in a tightly coordinated system to eliminate bottlenecks and slash training and inference costs. A key innovation is the NVIDIA Inference Context Memory Storage Platform — an AI-native key-value cache tier that delivers five times higher tokens per second, five times better performance per dollar, and five times greater power efficiency. Together, these advancements enable AI inference at one-tenth the cost, accelerating the pace of innovation and time to market. Huang emphasized that NVIDIA is now a frontier AI model builder, training open models on its own supercomputers to power breakthroughs across healthcare, climate science, robotics, and autonomous driving. The company’s open model portfolio spans six domains: Clara for healthcare, Earth-2 for climate science, Nemotron for reasoning and multimodal AI, Cosmos for robotics and simulation, GR00T for embodied intelligence, and Alpamayo for autonomous driving. “Every six months, a new model emerges — smarter, faster, more capable,” Huang said. “Downloads are exploding, and the world is building on open models.” He highlighted that 80% of AI startups now use open models, underscoring the importance of open collaboration. These models are fully accessible for creation, evaluation, safety alignment, and deployment. Huang also introduced AI’s personal future, demonstrating a local AI agent running on the NVIDIA DGX Spark desktop supercomputer, interacting with a Reachy Mini robot using Hugging Face models. “Just a few years ago, this would have been unimaginable,” he said. The DGX Spark now delivers up to 2.6x performance for large models, with support for Lightricks LTX-2 and FLUX image models, and upcoming NVIDIA AI Enterprise availability. In the physical AI space, Huang showcased how robots are trained in photorealistic simulations using NVIDIA Isaac Sim and Isaac Lab before operating in the real world. He announced Alpamayo — an open suite of reasoning, vision, language, and action models, along with simulation blueprints and datasets — designed to enable level 4 autonomy. A video demonstrated a vehicle using Alpamayo navigating complex San Francisco traffic, making real-time decisions based on sensor input and internal reasoning. The first passenger car featuring Alpamayo, built on the NVIDIA DRIVE full-stack platform, will debut in the new Mercedes-Benz CLA, with AI-defined driving launching in the U.S. this year, following the car’s EuroNCAP five-star safety rating. Huang also highlighted the growing adoption of DRIVE Hyperion, the open, modular, level-4-ready platform used by automakers, suppliers, and robotaxi providers worldwide. He was joined on stage by small, animated robots, illustrating how NVIDIA’s full-stack approach is fueling a global physical AI ecosystem. Partnerships with Siemens, Synopsys, Cadence, Boston Dynamics, and Franka were showcased, demonstrating how AI is integrated from design and simulation to production. “These manufacturing plants will become giant robots,” Huang said. Huang concluded by reaffirming NVIDIA’s mission: to build the entire stack so others can create transformative applications. “Our job is to create the foundation so you can build the future,” he said.