Nvidia announced a series of groundbreaking advancements at its annual GTC conference on Tuesday, marking a significant leap forward in artificial intelligence technology. The company's sales have soared, increasing more than sixfold since the release of OpenAI's ChatGPT in late 2022, transforming Nvidia's business landscape. At the event, Nvidia introduced new versions of its Blackwell family of chips, named Blackwell Ultra, which are designed to enhance reasoning models. Additionally, the company revealed plans for its next-generation GPU, Vera Rubin, set to ship in 2026.
The announcement highlighted Nvidia's dominance in the AI sector, particularly in the realm of "big GPUs" used for developing advanced AI through a process called training. Nvidia's CEO, Jensen Huang, emphasized the significant computational demands and the rapid acceleration of AI technologies.
"This last year is where almost the entire world got involved. The computational requirement, the scaling law of AI, is more resilient, and in fact, is hyper-accelerated," said Jensen Huang.
Nvidia's introduction of Blackwell Ultra chips marks a pivotal enhancement in capabilities, especially for reasoning models. Furthermore, Nvidia is redefining what it terms as a GPU with the announcement of its Vera Rubin graphics processing unit. Vera Rubin is unique in that it comprises two GPUs and will be housed in a rack called Vera Rubin NVL144. The company also disclosed plans for a subsequent release, the "Rubin Next" chip, anticipated in the second half of 2027. This chip will integrate four dies into a single unit, effectively doubling Rubin's speed.
The company's strategic updates extend beyond chips. Nvidia unveiled enhancements to its networking components designed to interconnect hundreds or thousands of GPUs, enabling them to function as a cohesive unit. This innovation is expected to significantly boost performance and efficiency in AI model training and deployment.
In addition to hardware improvements, Nvidia introduced Dynamo, a software package aimed at maximizing the performance of its chips. Dynamo is expected to offer users enhanced capabilities and efficiencies in utilizing Nvidia's hardware.
The advancements in Nvidia's custom design are equally noteworthy. The Vera design promises to be twice as fast as the CPU employed in last year's Grace Blackwell chips. This leap in processing power underscores Nvidia's commitment to maintaining its leadership position in AI technology.
"In the last 2 to 3 years, a major breakthrough happened, a fundamental advance in artificial intelligence happened. We call it agentic AI," Jensen Huang remarked during the conference.
Looking ahead, Nvidia anticipates shipping systems based on its next-generation GPU family in the latter half of 2026. This timeline sets the stage for continued innovation and expansion in AI capabilities.
The GTC conference also provided a platform for Nvidia to showcase its extensive array of products and services beyond the newly announced chips. These demonstrations further highlighted Nvidia's comprehensive approach to advancing AI technology and supporting its widespread adoption.