NVIDIA• Mar 16, 2026• 2:20:10Interview

NVIDIA CEO Jensen Huang GTC 2026 Full Keynote

From NVIDIA

Jensen Huang•CEO, NVIDIA

Executive Summary

NVIDIA announced its next-generation AI supercomputer platform, Vera Rubin, designed for the era of agentic AI, featuring new GPUs, CPUs (Vera), and networking (Spectrum X with CPO).
The company is framing the future of data centers as 'AI Factories' optimized for generating 'tokens' (intelligence) at the lowest possible cost, emphasizing performance-per-watt as a key metric.
NVIDIA is making a major push into physical AI and robotics, releasing a suite of open-source platforms (Isaac Lab, Newton, Groot) and announcing major automotive partnerships for its robo-taxi platform.
The CEO predicts a fundamental shift in the software industry from SaaS (Software as a Service) to 'GaaS' (Generative Agent as a Service), driven by the rise of autonomous AI agents.

12 quotes

Concerns Raised

Meeting the exponential (1 million-fold in 2 years) growth in compute demand.
Scaling manufacturing and supply chains to deliver complex, rack-scale systems in volume.
The immense power and infrastructure requirements for building out global 'AI Factories'.

Opportunities Identified

Establishing the Vera Rubin platform as the standard for the next wave of agentic AI.
Capturing the emerging multi-trillion dollar market for physical AI and robotics.
Driving the industry-wide transition from SaaS to 'GaaS' (Generative Agent as a Service).
Becoming the core infrastructure for reinvented industries, from automotive to telecommunications.

Key Themes

The Rise of the AI Factory

NVIDIA is repositioning the data center from a place of information storage to a 'factory' for producing intelligence in the form of tokens. This new paradigm prioritizes throughput, energy efficiency, and the total cost of token generation over traditional metrics. The entire Vera Rubin platform is architected to maximize the output of these factories.

This shift reframes the entire economic model of cloud computing and enterprise IT, making performance-per-watt and operational efficiency critical for competitive advantage in the AI era.

Agentic AI as the Next Computing Wave

The presentation heavily emphasizes that the next frontier is 'agentic AI'—autonomous systems that can reason, plan, and act. The Vera Rubin platform is explicitly designed for the diverse workloads of these agents, which involve not just LLM inference ('thinking') but also data retrieval and tool usage. The CEO predicts this will transform every SaaS company into a 'GaaS' (Generative Agent as a Service) provider.

This signals a move beyond simple chatbots to complex, autonomous AI systems, creating exponential demand for more powerful, versatile, and efficient computing infrastructure.

Extreme Co-Design and Full-Stack Dominance

NVIDIA is not just a chip company; it's a systems company. The announcements detail a vertically integrated stack, from new GPUs and CPUs (Vera) to sixth-generation NVLink interconnects, Spectrum X networking with co-packaged optics, and a vast software ecosystem (CUDA, NemoClaw). This 'extreme co-design' allows for massive generational performance leaps, such as the 35-50x performance-per-watt gain with Grace Blackwell.

This full-stack approach creates a powerful competitive moat, making it difficult for competitors to match performance. It also means customers are buying into an entire ecosystem, not just a component, increasing lock-in and driving down TCO.

Physical AI: From Simulation to the Real World

NVIDIA is aggressively expanding into robotics and physical AI. The company is releasing a comprehensive suite of open-source tools like Isaac Lab for training and Newton for physics simulation to close the 'sim-to-real' gap. This is complemented by major partnerships in the automotive sector (BYD, Hyundai, Uber) for its robo-taxi platform.

This marks a significant expansion of AI's total addressable market from the digital realm to the physical world, unlocking new applications in manufacturing, logistics, healthcare, and transportation.

Get started free

Topics

NVIDIA GTC AI Factories Agentic AI Vera Rubin Platform Grace Blackwell Architecture Token Generation Robotics Physical AI Isaac Lab CUDA NVLink Generative AI Performance per Watt Data Center Architecture Co-packaged Optics (CPO)Spectrum X GaaS (Generative Agent as a Service)Autonomous Vehicles Robo-taxi

Processed Apr 2, 2026 yt-dlp + mlx-whisper + Gemini