We’re bringing you live updates from San Jose throughout the week, covering Nvidia CEO Jensen Huang’s keynote as it unfolds, along with breakout sessions, live demos, and on-the-ground highlights through March 19.
Before the keynote, the SAP Center filled up with attendees anticipating the main event.
The keynote opened with a video introducing the token as the basic unit of modern AI, the building block behind systems for science, discovery, virtual worlds, and real-world machines.
NVIDIA founder and CEO Jensen Huang walked onto the stage to loud applause from the audience.
He started by thanking the pre-show hosts and commending the partners over 450 sponsors, 1,000 sessions, and 2,000 speakers who were involved in the event.
This conference will cover a single layer of the five-layer framework of artificial intelligence, Huang said.
After outlining the event, Huang celebrated the 20th anniversary of CUDA, Nvidia’s parallel computing platform and programming model, calling it the flywheel behind accelerated computing and the platform that supports every single phase of the AI life cycle.
Huang then discussed Nvidia’s GeForce, calling it the foundation of the company’s efforts to bring CUDA to the world and connecting its history to AI and DLSS 5. A video demonstrated 3D-guided neural synthesis delivering real-time, photoreal 4K performance on local hardware. More details are available in the press release.
Moving from product highlights to industry partners, Huang explained how data processing is accelerating in the AI era. He talked about working with IBM, Dell, Google Cloud, AWS, Microsoft Azure, Oracle, and CoreWeave to help their customers.
Hu Wang also gave an interview on the accelerated computing ecosystem, which includes industries like automotive, financial services, healthcare, industrial media, quantum, retail, robotics, and telecom.
All of these different areas of AI have platforms that Nvidia provides, Huang said, pointing out the company’s wide range of CUDA-X libraries collections of software tools built to help developers use CUDA efficiently which he called the crown jewels of Nvidia.
Huang talked about the rise of AI natives, new companies like OpenAI and Anthropic, and others still emerging. This last year, it just skyrocketed, he said, noting $150,000,000,000 invested in startups and reviewing the technologies that sparked the newest tech boom.
Because of this boom, demand for NBA GPUs is off the charts, he said. I believe computing demand has increased by a factor of 1,000,000 over the last few years.
Huang said that as a result, he expects at least $1,000,000,000,000 in revenue from 2025 to 2027.
Vera Rubin and Beyond — A Generational Leap in Computing
Huang pointed out that NVDIA token cost – the computational cost to process a piece of data in an AI model – is the lowest in the world. Thanks to extremecodesign. He enjoyed hearing one analyst call Nvidia the inference king. This is the incredible power of extreme codesign, Huang said, referring to the design of software and silicon (hardware chips) together.
The next step is Nvidia Vera Rubin, a new full-stack computing platform that includes every component from hardware to software with seven chips, five rack-scale systems, and one supercomputer for Agentic AI (AI that can act independently). The platform features the new Nvidia Vera CPU (central processing unit) and BlueField 4 STX storage architecture (the design for organizing and accessing stored data).
When we think of Vera Rubin, we think of a complete, vertically integrated system with software extended end-to-end and optimized as one giant system. Huang said as he showed the audience the inner workings of these new technologies.
Looking forward, NVIDIA’s next major architecture is called Feynman.
This architecture will feature a new CPU called Nvidia Rosa. It is named after Rosalind Franklin, whose X-ray crystallography revealed the structure of DNA and changed modern biology. Just as Franklin uncovered life’s hidden structure, Rosa is designed to move data and tokens fluently across agency AI infrastructure.
Rosa is at the center of a new platform that combines LP40, Nvidia, Next Generation LPU with Nvidia, BlueField 5, and NP. CX10. These are connected using NVIDIA Kyber for both copper and co-packaged optics. The system also includes Nvidia, Spectrum Class, and Optical Scale Out. Huang said the Feynman generation improves every part of the AI factory: compute, memory, storage, networking, and security.
To further accelerate the growth of new AI capacity, Huang announced the Nvidia Vera Rubin DSX AI factory reference design — a model plan for building AI infrastructure — and the Nvidia Omniverse DSX blueprint, which provides guidelines for designing AI workspaces. DSX Air is part of the larger DSX platform and enables companies to simulate AI factories in software before building them in the real world.
Expanding the reach of NVIDIA’s technology. Huang then announced that NVIDIA is heading to space. The new Vera Rubin architecture is named after the astronomer who studied dark matter. Future systems such as NVIDIA Space One and Vera Rubin are being developed to bring AI data centers into orbit. This expands accelerated computing beyond Earth.
NVIDIA NemoClaw For OpenClaw Nemotron Coalition
Huang highlighted OpenClaw, an open-source project by developer Peter Steinberger, which he called the most popular open-source project in the history of humanity.
OpenClaw has open-sourced the operating system of agentic computers. Now OpenClaw has made it possible for us to create personal agents. Huang said:
With just one command, developers can download the OpenClaw setup and AI agent and start adding tools and context. NVIDIA is now supporting OpenCL across its platform, making it easier for developers to safely build, deploy, and speed up AI agents on NMID-powered infrastructure. No company in the world today has to have an open cloud strategy, Huang said.
To ensure this technology is secure for businesses, Huang introduced the NVIDIA OpenShell runtime and the NVIDIA NemoClaw stack. These combine policy enforcement network guardrails and privacy routing. Huang said this could become the policy engine for all SaaS companies worldwide. NVIDIA is also growing its open model ecosystem with the new Nematron Coalition. This brings partners together around six leading model families: Nvidia, Nemotron. (language and reasoning) n media cosmos (world and vision), Nvidia ISAC GR00T (general purpose robotics), Nvidia Alpamayo (autonomous driving), Nvidia Bio-Nemo (biology and chemistry), and Nvidia Earth 2 (weather and climate).
Physical AI
Extending AI’s influence beyond digital agents, NVIDIA is now moving AI into the physical world, enabling it to operate there.
Huang said that Nvidia’s Robotaxi Ready platform is attracting new automaker partners like BYD, Hyundai, Nissan, and Geely.
He also mentioned a partnership with Uber to add these vehicles to its ride-hailing network.
Beyond automakers, Nvidia is also teaming up with industrial software leaders and robotics companies like ABB, Universal Robots, and KUKA to integrate its physical AI models and simulation tools. This will help deploy smarter robots on manufacturing lines. NVIDIA is working with telecom providers like T-Mobile as base stations become edge AI platforms.
That’s a Wrap
Huang ended his keynote with a surprise: Olaf, the showman from Disney’s Frozen, seemed to walk right off a digital screen and onto the stage.
Ladies and gentlemen, Olaf Huang announced as the character waddled out, powered by Nvidia’s physical AI stack, the Newton physics engine, and N Media Omniverse Simulation. Laf—how are you? I know because I gave you your computer, Jetson. Huang joked.
When Olaf asked what that was, Huang answered, “Well, it’s in your tummy, and you learned how to walk inside Omniverse.”
The demo highlighted Huan’s main point: everything shown, from humanoid robots to animated characters, was simulated in real time, not pre-rendered. He closed by recapping the themes inference, the AI factory, the open claw, physical AI, and robotics then handed the stage to a musical ensemble: singing robots, a digital Jensen avatar, and an animated lobster performing a campfire song.
All right, have a great GTC, Huang said as he left the stage. Olaf stayed behind, entertaining the crowd before vanishing beneath the stage through a trap door.
