By Stephen Nellis and Max A. Cherney
SAN JOSE, California (Reuters) – Nvidia CEO Jensen Huang is expected on Tuesday to reveal fresh details about the company’s newest artificial intelligence chip at its annual software developer conference.
Nvidia stock has more than quadrupled in value over the past three years as the company powered the rise of advanced AI systems such as ChatGPT, Claude and many others.
Much of that success stemmed from the decade that the Santa Clara, California-based company spent building software tools to woo AI researchers and developers – but it was Nvidia’s data center chips, which sell for tens of thousands of dollars each, that accounted for the bulk of its $130.5 billion in sales last year.
Huang hinted last year the new flagship offering will be named Rubin and consist of a family of chips – including a graphics processing unit, a central processing unit and networking chips – all designed to work in huge data centers that train AI systems. Analysts expect the chips to go into production this year and roll out in high volumes starting next year.
Nvidia is trying to establish a new pattern of introducing a flagship chip every year, but has so far hit both internal and external obstacles.
The company’s current flagship chip, called Blackwell, is coming to market slower than expected after a design flaw caused manufacturing problems. The broader AI industry last year grappled with delays in which the prior methods of feeding expanding troves of data into ever-larger data centers full of Nvidia chips had started to show diminishing returns.
Nvidia shares tumbled this year when Chinese startup DeepSeek alleged it could produce a competitive AI chatbot with far less computing power – and thus fewer Nvidia chips – than earlier generations of the model. Huang has fired back that newer AI models that spend more time thinking through their answers will make Nvidia’s chips even more important, because they are the fastest at generating “tokens,” the fundamental unit of AI programs.
“When ChatGPT first came out, the token generation rate only had to be about as fast as you can read,” Huang told Reuters last month. “However, the token generation rate now is how fast the AI can read itself, because it’s thinking to itself. And the AI can think to itself a lot faster than you and I can read and because it has to generate so many future possibilities before it presents the right answer to you.”
(Reporting by Stephen Nellis and Max Cherney in San Jose, California; Editing by Rod Nickel)
Comments