Patrick T. Fallon / AFP
- Nvidia’s big AI conference, GTC, kicked off in full on Tuesday.
- CEO Jensen Huang gave an update on Nvidia’s chips roadmap, including Blackwell Ultra and Vera Rubin launch timing.
- He announced a Nvidia’s Dynamo Open-Source system, a GM partnership, and more.
Nvidia’s “Super Bowl of AI” is here.
CEO Jensen Huang took the stage in San Jose on Tuesday to give the main keynote address at Nvidia’s GTC 2025 AI conference.
“What an amazing year it was, and we have a lot of incredible things to talk about,” Huang said. “And I just want you to know that I’m up here without a net. There are no scripts, there’s no teleprompter, and I’ve got a lot of things to cover. So let’s get started.”
Huang said GTC used to be compared to Woodstock, but now it’s being compared to the Super Bowl.
“The only difference is, everybody wins at this Super Bowl,” Huang said onstage.
During the Nvidia CEO’s highly technical presentation — which you can watch below — Huang talked about Nvidia’s upcoming AI chipsets and architectures, including the next-generation Blackwell Ultra and AI superchip Vera Rubin platform, as well as the chipmaker’s work in robotics and autonomous driving.
Nvidia’s roadmap for next-generation AI chips
The biggest thing Wall Street was looking for on Tuesday was an update on Nvidia’s launch pipeline for its cutting-edge AI chips, used by many Big Tech and AI startups to develop frontier AI models.
Nvidia is preparing for the transition from Blackwell to Blackwell Ultra, which is expected to launch later this year. During Nvidia’s most recent earnings call, Huang said that Blackwell demand has been “extraordinary” after a “hiccup” in early productionand and he expects the transition to Blackwell Ultra to go more smoothly.
The company is also readying its AI superchip platform, Vera Rubin, which Huang said the company named after the astronomer who discovered dark matter. Huang first unveiled the Rubin platform last year at Computex.
The Nvidia CEO gave updates to both launch timelines as well as what’s coming next.
Huang said Vera Rubin will be coming in the second half of 2026.
Nvidia
The platform has a new CPU and networking architecture, and twice the performance of Hopper and more memory.
“Basically everything is brand new, except for the chassis,” Huang said.
Huang said Vera Rubin Ultra, the next generation of Vera Rubin, will be available the second half of 2027.
Huang announces Dynamo open-source system, GM partnership, and more
Huang also announced Nvidia Dynamo, an open-source inference software to help accelerate and scale AI reasoning models.
The Nvidia CEO referred to Dynamo as “essentially the operating system of an AI factory.” The system is named after the first instrument that started the last Industrial Revolution, he added.
Huang added that Perplexity, one of his “favorite partners” is working with Nvidia on Dyanmo.
“Love them so much because the revolutionary work that they do, and also because Aravind such a great guy,” Huang said about Perplexity and its CEO Aravind Srinivas.
Nvidia also announced it’s partnering with General Motors to “build custom AI systems using Nvidia accelerated compute platforms,” including vehicles, factories, and robots.
Nvidia’s GPUs have powered the AI gold rush, with the company’s first-generation AI chip, Hopper, reportedly selling for upwards of $40,000 and quickly becoming a hot commodity. However, recent advancements like Chinese startup DeepSeek’s lower-cost AI model have raised questions about the level of infrastructure investment needed to drive frontier LLM development.
On stage, Huang told the audience that he expects growing demand for compute. In an example Huang presented, he said a reasoning model like Deepseek’s R1 required 20 times more tokens to make a wedding seating chart than a traditional LLM model.
Nvidia’s stock was trading down more than 2% as of 2:30 p.m. in New York.
This is a developing post…