Dive Brief:
- Nvidia CEO Jensen Huang unveiled new chip designs, artificial intelligence innovations, corporate partnerships and robotics breakthroughs at the company’s 2025 GTC event in San Jose, California, on Tuesday.
- During the two-hour keynote presentation, Huang introduced Nvidia’s next family of graphics processing units, named after astronomer Vera Rubin. The GPUs are set for release in the second half of 2026 and are expected to have about three times the performance of the company’s latest Blackwell chips.
- Nvidia also unveiled an open-source, humanoid foundational model that companies can use to help speed up the development of robots for their own operations.
Dive Insight:
As companies look to adopt and bolster AI systems for a range of applications, Nvidia is leading the way with a host of computing infrastructures to meet growing demand for data centers, robotics and physical AI, such as for autonomous vehicles and surgical tools.
So far this year, Nvidia has sold more than 3.6 million Blackwell GPUs to the top four U.S. cloud-service providers alone, according to Huang’s presentation slides. That is nearly triple the amount of Hopper chips sold in 2024, underscoring that computing demand is here to stay. Hopper is the predecessor to Blackwell.
“AI is going through an inflection point,” Huang said.
As AI continues to advance and become more useful to consumers and customers, the amount of computation necessary to train and infer those models will also grow, Huang said. For example, he expects data center buildout to reach $1 trillion by 2028 to support a widespread shift in AI computing.
As the industry reshapes, Nvidia is releasing new superchips about every 12 to 18 months. In the second half of 2025, the company is set to start production of its Blackwell Ultra GPU, about 1.5 times faster than the base model. It also unveiled its Rubin Ultra GPU, expected to be more than 14 times faster than Blackwell and slated for the second half of 2027.
“This isn’t like buying a laptop,” Huang said, underscoring that it takes years of planning and coordination to ensure there is land and electrical power to accommodate the data centers needed to support AI infrastructure.
Beyond chip innovations, Nvidia is advancing AI through a number of partnerships. The company has agreed to work with General Motors to build its next generation of self-driving cars. This includes partnering with the car giant on AI infrastructure for manufacturing, enterprise systems and the cars themselves, Huang said.
Nvidia is also taking steps to advance robotics across industries to address labor issues. By the end of this decade, Huang estimated a global shortage of at least 50 million workers. Another estimate has the worldwide talent deficit at closer to 85 million, according to management consulting firm Korn Ferry.
To accelerate robot development, Nvidia is making available its Isaac GROOT N1, which the company claims is the world’s first open-sourced, customizable foundational model for general humanoid reasoning and skills. This can be translated to a number of use cases such as material handling, packaging and inspection.
“This is going to be a very, very large industry,” Huang said about the potential of AI in robotics.
