Nvidia CEO Jensen Huang speaks during the keynote address at Nvidia’s GTC conference on March 16, 2026 in San Jose, California. Nvidia’s GTC conference focuses on recent developments and future uses of AI.
Benjamin Fanjoy | Getty Images
in Nvidia’s At the company’s annual developer conference on Monday, CEO Jensen Huang told a packed room that he expects orders between Blackwell and Vera Rubin to reach $1 trillion by 2027.
Last year, the company predicted that the two chip technologies would represent a $500 billion revenue opportunity. Following Nvidia’s earnings report last month, finance chief Colette Kress said the company expects growth this year to exceed what was included in its forecast.
Huang said demand is surging from both startups and large corporations. Nvidia stock rose about 2% on Monday.
“If we can increase capacity, we will be able to generate more tokens and the revenue will increase,” Huang said at GTC in San Jose, California.
Nvidia’s artificial intelligence graphics processing units have made the brand a household name and the world’s most valuable publicly traded company, worth about $4.5 trillion. As mass adoption of AI moves from chatbots to agent apps that launch other agents to perform tasks, the number of tokens being generated will explode, further increasing the need to perform inference faster.
In February, the company announced that its sales for the current quarter would be about $78 billion, up about 77% from a year ago. The company has reported 11 consecutive quarters of revenue growth of more than 55%.
Nvidia plans to roll out Vera Rubin later this year. The system is made up of 1.3 million components and the company claims it delivers 10x better performance per watt than the previous generation Grace Blackwell. This is an important advance when energy consumption is one of the most important issues facing building AI.
Also on Monday, Huang announced the Nvidia Groq 3 language processing unit (LPU). This is the company’s first chip from the startup, which it acquired most of through a $20 billion asset acquisition in December, the company’s largest-ever deal. It is scheduled to ship in the third quarter.
Nvidia CEO Jensen Huang delivers the keynote address at the company’s annual GTC developer conference at SAP Center in San Jose, California on March 16, 2026.
Josh Edelson | AFP | Getty Images
Groq is Google It is an in-house tensor processing unit that has gained attention in recent years as a competitor to Nvidia’s graphics processing units. The Groq 3 LPU is built to power technology and features one core optimized for GPU acceleration.
Huang installed a full rack dedicated to housing the new Groq accelerator.
The Groq 3 LPX rack has 256 LPUs and will be installed next to a Vera Rubin rackscale system that will ship to customers later this year. Huang said the Groq LPX rack can improve Rubin GPU’s token-per-watt performance by 35x.
“We took two radically different processors and integrated them, one with high throughput and one with low latency. That doesn’t change the fact that you need a lot of memory,” Huang says. “So we’re going to add a bunch of Groq chips to expand the amount of memory.”
Huang also showed off a prototype of Kyber, Nvidia’s next big leap forward in rack architecture after Rubin. To increase density and reduce latency, 144 GPUs are integrated into a vertically oriented compute tray instead of horizontally. Kyber’s design will be available in Nvidia’s next rack-scale system, the Vera Rubin Ultra, scheduled to ship in 2027.
About two hours after the keynote, Huang turned to the OpenClaw phenomenon, which Austrian software developer Peter Steinberger announced in January. Thanks in part to social media attention, their popularity has skyrocketed as consumers and businesses flock to products that can autonomously complete tasks, make decisions and take actions on your behalf without continuous human guidance.
Steinberger joined OpenAI last month, and CEO Sam Altman said OpenClaw “will remain within the foundation as an open source project that OpenAI will continue to support.”
Huang highlighted new developer toolkits that use Nvidia hardware to help you build and experiment with what’s possible in new AI areas. He specifically introduced a so-called reference stack named NemoClaw for OpenClaw and helped make it “enterprise-ready.”
“Find OpenClaw and download it. Build an AI agent,” Huang says.
In the automotive field, Mr. Huang has previously announced Uberannounced that it will launch a ride-hailing service powered by Nvidia’s Drive AV software in 28 cities on four continents by 2028, starting with Los Angeles and San Francisco next year.
Huang announced that Nissan, BYD, Geely, Isuzu, and Hyundai Motors are developing Level 4 self-driving cars based on Nvidia’s Drive Hyperion program. Isuzu and China’s Tier IV are also building self-driving buses using the platform, with the help of Nvidia’s AGX Thor robotic system chip.
—CNBC’s Jordan Novet contributed to this report.
Watch: Inside Nvidia’s Vera Rubin AI system
