Jensen Huang Brings the Thunder at GTC 2026

In front of 30,000 attendees filling the SAP Center in San Jose, California — fans who traveled from 190 countries — Nvidia CEO Jensen Huang took the stage in his trademark leather jacket on Monday, March 16, to deliver what many industry observers are calling the most consequential tech keynote of the decade. The message was simple: the era of agentic AI has arrived, and Nvidia intends to power all of it.

Vera Rubin: A Generational Leap

The centerpiece of the event was the unveiling of the Nvidia Vera Rubin platform, named after the astronomer whose work revealed the existence of dark matter. Comprising seven breakthrough chips, five rack-scale systems, and one revolutionary supercomputer architecture, Vera Rubin represents what Huang called "a generational leap in full-stack computing."

At the heart of the platform is the new Nvidia Vera CPU — purpose-built for agentic AI workloads — paired with the Rubin GPU and the BlueField-4 STX storage architecture. The entire system is vertically integrated and software-optimized end-to-end, designed to handle the explosive growth of inference workloads driven by agentic applications.

"When we think Vera Rubin, we think the entire system, vertically integrated, complete with software, extended end to end, optimized as one giant system," Huang said.

Vera Rubin will deliver 10 times more performance per watt than its predecessor, Grace Blackwell — a critical advance at a moment when energy consumption is one of the most pressing constraints on AI infrastructure buildout.

The $1 Trillion Opportunity

Huang didn't mince words about the financial stakes. Last year, Nvidia saw a $500 billion revenue opportunity across Blackwell and Rubin. This year, with AI adoption accelerating from chatbots to full agentic systems, he revised that figure dramatically upward.

"I believe computing demand has increased by 1 million times over the last few years," Huang said. "I see at least $1 trillion in revenue from 2025 through 2027."

Nvidia shares rose roughly 2% during the keynote. The company has now reported eleven consecutive quarters of revenue growth above 55%, and its February earnings showed year-over-year revenue surging 77% to approximately $78 billion for the current quarter.

The Groq 3 LPU and the Inference Inflection Point

In another major reveal, Huang introduced the Nvidia Groq 3 Language Processing Unit (LPU) — the first chip from the AI startup Nvidia acquired in a $20 billion deal last December, the company's largest acquisition ever. The Groq 3 is built to complement the Vera Rubin GPU, with one core optimized for low-latency inference to offset the GPU's high-throughput focus.

The Groq LPX rack holds 256 LPUs and sits alongside the Vera Rubin rack-scale system, with Huang claiming it can increase tokens-per-watt performance of Rubin GPUs by 35 times. The first units are expected to ship in the third quarter of 2026.

Central to the keynote's narrative was what Huang called "the inflection point of inference." As AI systems shift from training to deployment — and especially as agentic AI requires models to constantly process new information and take real-world actions — the demand for fast, efficient inference has overtaken training as the primary driver of compute demand.

"Finally, AI is able to do productive work, and therefore the inflection point of inference has arrived. AI now has to think. In order to think, it has to inference. AI now has to do. In order to do, it has to inference."

Looking Beyond Rubin: The Feynman Generation

Huang also offered a preview of what comes after Vera Rubin. The next major architecture, codenamed Feynman, will pair a new CPU called Rosa — named for Rosalind Franklin, whose X-ray crystallography revealed the structure of DNA — with the LP40 LPU, BlueField-5, and NVIDIA's Kyber interconnect. Kyber will support co-packaged optics for scale-up and Spectrum-class optical scale-out for the full AI factory stack.

NemoClaw, OpenClaw, and Going to Space

In a moment that surprised even longtime GTC watchers, Huang gave extended airtime to OpenClaw, the open-source AI agent framework created by developer Peter Steinberger (who joined OpenAI in February). Huang called it "the most popular open source project in the history of humanity" and announced NemoClaw — an Nvidia reference stack built on OpenClaw that packages policy enforcement, network guardrails, and privacy routing for enterprise deployment.

"Every single company in the world today has to have an OpenClaw strategy," Huang said.

Finally, Huang announced that Nvidia is going to space. Future systems in the Vera Rubin lineage — called NVIDIA Space-1 Vera Rubin — are being designed to bring AI data centers into orbit, extending accelerated computing from Earth to space.

What It Means

GTC 2026 confirmed what analysts have been watching unfold for the past two years: Nvidia is no longer just a chip company. It is the infrastructure layer of the AI economy. From chips and CPUs to software stacks, agent frameworks, and orbital data centers, Huang's vision is of a vertically integrated AI ecosystem that no competitor comes close to matching. The question is no longer whether Nvidia dominates the AI era — it's how long that dominance can last.