NVIDIA partners with BYD again! Jensen Huang officially announces

On March 16 local time, the annual developer conference (GTC 2026) of NVIDIA, dubbed the “Spring Festival Gala of AI,” kicked off in San Jose, California, USA. At 11 a.m. that day (approximately 2 a.m. Beijing time on March 17), NVIDIA founder and CEO Jensen Huang, dressed in his signature black leather jacket, took the stage at the SAP Center in San Jose to deliver a keynote speech, introducing NVIDIA’s plans for the upcoming year to developers, analysts, and the media. It is reported that GTC 2026 will run from March 16 to 19.

As one of the most important technology conferences in the global AI industry, this GTC conference not only marks the official launch of the fully mass-produced Vera Rubin Super AI platform but also reveals NVIDIA’s strategic layout through the acquisition and integration of the Groq team to introduce the third-generation LPU, as well as the release of the enterprise-level agent operating system NemoClaw, from which many future indicators of AI industry development can be gleaned.

The AI industry has transcended the “Generative AI” stage

In his speech, Jensen Huang pointed out that the artificial intelligence industry has transcended the “Generative AI” phase and entered the “Inference Inflection Point” and “Agentic Era.” The conference announced the fully mass-produced Vera Rubin Super AI platform, the third-generation LPU, and the enterprise-level agent operating system NemoClaw.

Among them, the officially produced Vera Rubin platform is the highlight of this conference. Jensen Huang emphasized that this is no longer an iteration of a single chip but a super system composed of seven vertically integrated chips, specifically designed for Agentic AI. Compared to the Hopper architecture, the Vera Rubin platform has improved its token generation capability per watt by 35 times. In a 1GW data center, the token generation rate has increased from 2 million per second to 700 million, a 350-fold increase, achieving a leap in performance.

Jensen Huang also revealed NVIDIA’s latest strategic integration dynamics during his speech. To address the contradiction between high throughput and low latency, NVIDIA has acquired the Groq chip team and obtained its technology license, officially incorporating Groq LPU 30 into NVIDIA’s product lineup. It is reported that with its deterministic data flow architecture and massive SRAM, Groq LPU excels in low-latency decoding, complementing the Rubin GPU, which is adept at high throughput pre-filling.

Jensen Huang suggested that for high-throughput workloads, Vera Rubin can be utilized 100%. For high-value, low-latency code generation or interactive scenarios, 25% of Groq LPU nodes can be configured in the data center to achieve optimal performance and cost solutions.

“Lobster” accelerates the revolution of open-source software ecosystem

The recently popular open-source project “Lobster” OpenClaw also appeared in Jensen Huang’s speech. When launching the agent operating system, he highly praised OpenClaw, stating that its adoption speed surpasses that of Linux and is “the operating system for agentic computers.”

With the maturation of AI agent technology, Jensen Huang asserted that every SaaS company will become an AI company, or more accurately, an AaaS (Agentic as a Service) company. “The future software is designed for AI agents, not humans.”

Regarding enterprises’ concerns about accessing sensitive data, executing code, and external communication when using OpenClaw, NVIDIA has launched the enterprise-level reference architecture NemoClaw. Through built-in policy engines, network barriers, and privacy routers, it ensures that agents operate securely within enterprises. It supports enterprises in fine-tuning NVIDIA’s open models such as Nemotron, Cosmos (physical world model), and GR00T (robot model) to build sovereign AI.

Trillion-dollar orders and the realization of physical AI

As a “shovel seller” in the AI era, NVIDIA’s future development is under close scrutiny. In his speech, Jensen Huang made an optimistic prediction about the company’s future revenue trends. “The demand for computing will be even higher, and we will face shortages.” He predicts that NVIDIA’s revenue will reach $1 trillion by 2027.

Jensen Huang stated that data centers have transformed into “Token factories.” Under power constraints, the number of tokens produced per watt directly determines revenue. NVIDIA’s system efficiency highlights its advantages in token production costs, “even if competitors offer their architecture for free, they cannot compare.” Jensen Huang said.

The physical AI that Jensen Huang frequently mentioned in January at the 2026 International Consumer Electronics Show (CES 2026) also had more application cases during this speech.

Jensen Huang introduced that in the field of autonomous driving, NVIDIA announced collaborations with BYD (002594), Hyundai, Nissan, Geely, and Uber to deploy a Robotaxi network, promoting the “ChatGPT moment” of autonomous driving; in the robotics field, NVIDIA partnered with ABB, KUKA, Disney, and others to train humanoid robots using Isaac Lab, the Newton physics engine, and the Cosmos world model; in the realm of space computing, NVIDIA launched the Vera Rubin Space One program, which will build data centers in orbit to address cooling challenges using radiation heat dissipation technology.

The industrial revolution of artificial intelligence welcomes its groundbreaking ceremony

Similar to the scene at NVIDIA’s launch event held in a hotel theater in Las Vegas, USA, in January, where a white-haired Jensen Huang watched desert exploration videos on a big screen alongside a pair of adorable BDX robots, resembling a kindly old father. On the stage of GTC 2026, Jensen Huang continued to invite robots as his dialogue partners and engaged in an amusing interactive Q&A with Disney’s robot Olaf, who looked like a small snowman with a carrot nose.

“The factory is powered on, agents are learning to drive, all modules are decoupling, reorganizing, and accelerating. Inference has become a reality. Robots are being born for AI,” Jensen Huang said.

This over two-hour-long speech felt like a lively party. At the end of the speech, the organizers even created a roughly four-minute song using AI, and Jensen Huang transformed into an animated character, sitting around a campfire in the forest, showcasing the industry’s hot topics such as GTC, AI Factories, OPEN SOURCE, and Physical AI through music and song with robot musicians.

As Jensen Huang said in his speech, “This is not just a conference - this is the groundbreaking ceremony of the artificial intelligence industrial revolution.”