San Jose, United States (Reuters)-NVIDIA announced on Tuesday new processors, including a family of them that combines GPU and CPU functions, and its executive president, Jensen Huang, said the company is well positioned to navigate the evolution of artificial intelligence technology.
Huang, speaking at the company’s annual software developer conference at San Jose, California, defended the company’s leading AI chips, which was recently questioned by investors after Chinese Deepseek developed a competitive chatbot that allegedly requires less data processing power and, consequently, fewer AI chips.
“Almost the whole world got it wrong,” Huang said on the conference stage, dressed in his usual black leather jacket and jeans. He called the conference “The AI Super Bowl.”
“The amount of computing we need as a result of AI agent, as a result of reasoning, is easily 100 times larger than we thought we needed last year,” he said, referring to AI autonomous agents that require little human intervention for routine tasks.
But its presentation positioning the company’s current chips as well suitable for the changes that sweep the AI market could not reassure investors and the company’s shares fell 2.5% this afternoon.
Nvidia’s value of actions has more than quadrupled in the last three years, as the company has boosted the emergence of advanced AI systems such as ChatgPT, Claude and many others.
Continues after advertising
Much of Nvidia’s success has resulted from the decade that the company based in Santa Clara, California, has been building software tools to attract AI researchers and developers – but it was Nvidia’s data center chips, which are sold by tens of thousands of dollars each, which represented most of their $ 130.5 billion in sales last year.
But Nvidia’s lucrative chips face the pressure of technological changes as IA markets change “training”, where AI models like chatbots are developed using large amounts of data to make it “intelligent” to “inference”, which is when the model uses their “intelligence” to produce responses to users.
New chips
Continues after advertising
Huang has announced new chips, including the upcoming GPU processor, Blackwell Ultra, which will be available in the second half of this year.
The executive said Nvidia’s chips have two main goals: help AI systems to respond intelligently to a large number of users, but also give these answers as soon as possible. Huang argued that Nvidia processors are the only ones who can do both.
“If you take too long to answer a question, the customer won’t come back. This is like a web search,” he said.
Continues after advertising
Huang has announced a new software called Dynamo, which Nvidia has launched for free and aims to accelerate the reasoning process of models. The executive also announced that the automaker General Motors chose Nvidia as a partnership in the production of a fleet of autonomous cars.
The executive hinted last year that the company’s new flagship will be called Rubin and will consist of a chip family-including a Graphic Processing Unit (GPU), a Central Processing Unit (CPU) and network chips-all designed to work in huge AI-training data centers. Analysts expect chips to go into production this year and will be released in large volumes from next year.
Nvidia is trying to establish a new pattern of launching a main chip each year, but so far has found internal and external obstacles.
Continues after advertising
The company’s current main chip, called Blackwell, is reaching the market more slowly than expected after a project failure caused manufacturing problems.
Also at the conference, Huang announced what it called the first open model for humanoid robots and simulation patterns to accelerate robot development. The model is called Isaac Gr00t N1.
The company also launched a personal computer and announced that the next generation of company microprocessors, Feyman, will hit the market in 2028.
