NVIDIA (NVDA) CES 2026 Keynote summary

Industry platform shifts and AI transformation

• Two simultaneous platform shifts are occurring: the rise of AI as a foundational layer and a reinvention of the entire computing stack, moving from traditional programming to training software and leveraging GPUs over CPUs.

• Hundreds of billions in VC funding and a significant portion of global R&D budgets are shifting toward AI modernization, impacting a $100 trillion industry.

• Open models are rapidly reaching the frontier, with significant releases from 2023 to 2026, and 80% of startups are building on open models.

• Model size and parameter count are growing 10x per year, with token costs dropping 10x annually, driving increased adoption and capability.

• The proliferation of agentic systems—AI models capable of reasoning, planning, and tool use—has begun to solve complex problems and is expected to drive the next wave of AI adoption.

AI infrastructure, open models, and enterprise integration

• Proprietary AI supercomputers (DGX Clouds) and the Vera Rubin platform, featuring six co-designed chips, are used to develop frontier models in domains like digital biology, weather prediction, and robotics.

• Libraries such as NeMo, Physics NeMo, Clara NeMo, and BioNeMo provide lifecycle management for AI, from data processing to deployment, all open-sourced for broad industry participation.

• AI models top global leaderboards in multimodal document understanding, speech recognition, and semantic search, enabling rapid development of AI agents.

• The agentic AI framework allows for multimodal, multi-cloud, and hybrid cloud applications, with customizable skills and intent-based model routing.

• Major enterprise platforms (Palantir, ServiceNow, Snowflake, NetApp) are integrating agentic AI systems as user interfaces, revolutionizing enterprise workflows.

Physical AI, robotics, and industrial transformation

• Physical AI requires three computing layers: training, inference (robotics), and simulation, with Omniverse and Cosmos providing digital twin and world modeling capabilities.

• Synthetic data generation, grounded in physics, enables scalable training for physical AI, with Cosmos as a foundation model for reasoning and scenario generation.

• The Alpamayo autonomous vehicle AI, trained with human and synthetic data, is being deployed in Mercedes-Benz vehicles, with global rollout planned for 2024.

• The AV stack features dual redundant software systems for safety, with full traceability and open ecosystem participation for L4 and robo-taxi development.

• Robotics platforms (Isaac Sim, Isaac Lab) and partnerships with companies like LG, Caterpillar, and Boston Dynamics are advancing the next era of robotic systems.