Monday, March 16, 11:00 a.m. PT 🔗
Stay Updates From the GTC Keynote
Welcome to GTC 2026
A capability crowd on the SAP Heart, all ready for a similar factor.
The keynote opened with a video framing the token as the essential unit of contemporary AI — the constructing block behind methods used for scientific discovery, digital worlds and machines working within the bodily world.
NVIDIA founder and CEO Jensen Huang then took the stage to raucous applause from the group.
He opened by thanking the pregame present hosts and highlighting the companions taking part within the present, together with the 450+ sponsors, 1,000 classes and a pair of,000 audio system.
“This convention goes to cowl each single layer of the five-layer cake of synthetic intelligence,” Huang mentioned.
He marked the twentieth anniversary of CUDA — describing it because the “flywheel” driving accelerated computing and the platform that helps “each single part of the AI lifecycle.”
Huang turned to GeForce, describing NVIDIA as “the home that GeForce made,” the platform that introduced CUDA to the world. He walked via GeForce’s historical past, tying all of it again to AI, and launched DLSS 5, launching a video exhibiting how 3D-guided neural rendering permits real-time, photoreal 4K efficiency on native {hardware}. Be taught extra within the press launch.
Subsequent, Huang walked via the sector of knowledge processing and the way it’s being accelerated for the period of AI. He detailed work with IBM, Dell, Google Cloud, AWS, Microsoft Azure, Oracle and CoreWeave, to serve their clients.
Huang then surveyed the accelerated computing ecosystem — automotive, monetary providers, healthcare, industrial, media, quantum, retail, robotics and telecom.
“All of those totally different vectors of AI have platforms that NVIDIA gives,” Huang mentioned, highlighting NVIDIA’s broad vary of CUDA-X libraries, which he described because the “crown jewels” of the corporate.
Huang highlighted the rise of “AI natives” — brand-new firms, some well-known, reminiscent of OpenAI and Anthropic, and a few nonetheless rising. “This final yr, it simply skyrocketed,” Huang mentioned, citing $150 billion of funding into enterprise startups and strolling via the historical past of the applied sciences that sparked the newest expertise growth.
On account of this growth, the computing demand for NVIDIA GPUs is “off the charts,” he mentioned. “I imagine computing demand has elevated by 1 million instances over the previous couple of years.”
Consequently, Huang mentioned he now sees at the least $1 trillion in income from 2025 via 2027.
Vera Rubin and Past — A Generational Leap in Computing
Huang famous that NVIDIA’s token price is one of the best on the earth, due to excessive codesign, reveling in a single analyst description of NVIDIA as “the inference king.” “That is the unbelievable energy of maximum codesign,” Huang mentioned, referencing a course of the place software program and silicon are designed in tandem.
The following step: NVIDIA Vera Rubin, a brand new full-stack computing platform comprising seven chips, 5 rack-scale methods and one supercomputer for agentic AI. The platform consists of the brand new NVIDIA Vera CPU and BlueField-4 STX storage structure.
“After we assume Vera Rubin, we expect your complete system, vertically built-in, full with software program, prolonged finish to finish, optimized as one big system,” Huang mentioned, strolling the viewers via the insides of latest methods constructed on these applied sciences.
Wanting past Vera Rubin, NVIDIA’s subsequent main structure is Feynman.
It should embrace a brand new CPU, NVIDIA Rosa, named for Rosalind Franklin, Huang mentioned, whose X‑ray crystallography revealed the construction of DNA and reshaped fashionable biology. As Franklin uncovered the hidden structure of life, Rosa is constructed to maneuver information, instruments and tokens effectively throughout the total stack of agentic AI infrastructure.
Rosa anchors a brand new platform that pairs LP40, NVIDIA’s subsequent‑technology LPU, with NVIDIA BlueField‑5 and CX10, linked via NVIDIA Kyber for each copper and co‑packaged optics scale‑up, and NVIDIA Spectrum‑class optical scale‑out, Huang mentioned. Collectively, the Feynman technology advances each pillar of the AI manufacturing facility: compute, reminiscence, storage, networking and safety.
And to assist speed up the scale-out of latest AI capability, Huang introduced the NVIDIA Vera Rubin DSX AI Manufacturing unit reference design and the NVIDIA Omniverse DSX Blueprint. DSX Air, a part of the broader DSX platform, lets firms simulate AI factories in software program earlier than constructing them within the bodily world.
Lastly, Huang introduced NVIDIA goes to house. Its new Vera Rubin structure honors the astronomer whose work revealed darkish matter, and future methods like NVIDIA House-1 Vera Rubin are being designed to carry AI information facilities into orbit, extending accelerated computing from Earth to house.
NVIDIA NemoClaw for OpenClaw, Nemotron Coalition
Huang spotlighted OpenClaw, an open supply undertaking from developer Peter Steinberger that he referred to as “the most well-liked open supply undertaking within the historical past of humanity.”
“OpenClaw has open sourced the working system of agentic computer systems … Now, OpenClaw has made it potential for us to create private brokers,” Huang mentioned.
With a single command, builders can pull down OpenClaw, arise an AI agent and start extending it with instruments and context. NVIDIA is saying help for OpenClaw throughout its platform, making it simpler for builders to soundly construct, deploy and speed up AI brokers on NVIDIA‑powered infrastructure.
Each single firm on the earth as we speak has to have an OpenClaw technique, Huang mentioned.
To make sure this expertise might be deployed securely inside enterprises, Huang launched the NVIDIA OpenShell runtime and the NVIDIA NemoClaw stack — combining coverage enforcement, community guardrails and privateness routing. These applied sciences can function “the coverage engine of all of the SaaS firms on the earth,” Huang mentioned.
As well as, NVIDIA is increasing its open mannequin ecosystem with a brand new Nemotron Coalition, rallying companions round six frontier mannequin households: NVIDIA Nemotron (language and reasoning), NVIDIA Cosmos (world and imaginative and prescient), NVIDIA Isaac GR00T (common‑objective robotics), NVIDIA Alpaymayo (autonomous driving), NVIDIA BioNeMo (biology and chemistry) and NVIDIA Earth‑2 (climate and local weather).
Bodily AI
NVIDIA is extending AI from digital brokers into bodily AI that may navigate the actual world.
Huang mentioned NVIDIA’s robotaxi‑prepared platform is drawing new automaker companions, together with BYD, Hyundai, Nissan and Geely.
He additionally highlighted a partnership with Uber to deploy these automobiles into its journey‑hailing community.
Past automakers, NVIDIA is working with industrial software program giants and robotics leaders reminiscent of ABB, Common Robots and KUKA to combine its bodily AI fashions and simulation instruments, enabling deployment of smarter robots on manufacturing strains, and with telecom suppliers like T‑Cellular as base stations evolve into edge AI platforms.
That’s a Wrap
Huang capped off his keynote with a shock go to from Olaf, the snowman from Disney’s Frozen, who appeared to stroll straight off a digital display screen and onto the stage.
“Women and gents, Olaf,” Huang mentioned, because the character waddled out, pushed by NVIDIA’s bodily AI stack, the Newton physics engine and NVIDIA Omniverse-powered simulation.
“Olaf, how are you? I do know as a result of I gave you your pc — Jetson,” Huang joked.
When Olaf requested what that was, Huang replied, “Nicely, it’s in your tummy … and also you realized how you can stroll inside Omniverse.”
The demo underscored Huang’s broader level: every part on show — from humanoid robots to animated characters — had been simulated, not pre-rendered.
Huang closed by recapping the themes — inference, the AI manufacturing facility, the OpenClaw, bodily AI and robotics — then handed the stage to a musical ensemble: singing robots, a digital Jensen avatar and an animated lobster, performing a campfire music.
“All proper, have an awesome GTC,” Huang mentioned, exiting stage left as Olaf lingered, hamming it up for the group as he sank again beneath the stage via a lure door.
Learn all NVIDIA information from GTC on the net press package.
Construct an agent of your personal at NVIDIA’s build-a-claw occasion within the GTC Park, March 16-19 — 1-5 p.m. on Monday, and eight a.m.-5 p.m. on Tuesday via Thursday — to customise and deploy a proactive, always-on AI assistant.
