SAN JOSE >> In 2009, when Nvidia held its first developer conference, the event was something of a science fair. Dozens of academics filled a San Jose hotel decorated with white poster boards of computer research. Jensen Huang, the chipmaker’s CEO, roamed the floor like a judge.

This year, Nvidia’s developer conference is far different.

More than 25,000 people were expected to congregate Tuesday around the event, known as Nvidia GTC. The crowds filled a National Hockey League arena to hear a speech about the future of artificial intelligence from Huang, who has been nicknamed “AI Jesus.” Nvidia, the world’s leading developer of AI chips, has also wrapped San Jose in the company’s neon green and black colors, shutting down city streets and sending hotel prices soaring as high as $1,800 a night.

A who’s who of industry leaders attended, including Michael Dell, CEO of Dell Technologies; Jeffrey Katzenberg, the co-founder of DreamWorks and WndrCo, a venture capital firm; and Bill McDermott, CEO of ServiceNow.

“GTC is jam-packed,” Huang said, as he kicked off the conference Tuesday morning. “The only way to hold more people at GTC is we’re going to have to grow San Jose.”

The transformation of Nvidia’s conference from an academic event to the Super Bowl of AI — a weeklong showcase of robots, large language models and autonomous cars — is symbolic of the company’s metamorphosis. As AI has gone mainstream, customers have clamored for Nvidia’s graphics processing units, or GPUs, the powerful chips that help create the technology. That has propelled the chipmaker to a nearly $3 trillion valuation, up from $8 billion in 2009.

Yet Nvidia’s ascent has raised questions. Generative AI, which can answer questions, create images and write code, has been celebrated for its potential to improve businesses and create trillions of dollars in economic value. Microsoft, Amazon, Google, Meta and others are spending hundreds of billions of dollars to make that idea a reality.

But the spending has prompted concerns across Wall Street and Silicon Valley about whether AI will make enough money to justify its staggering costs. And the technology’s trajectory can be upended by new entrants such as DeepSeek, a small Chinese company that made a cutting-edge AI system with a small fraction of the Nvidia chips that other companies used. (In January, when investors realized what DeepSeek had done, Nvidia lost $600 billion in value on a single day.)

At Nvidia GTC, Huang sought to reassure people that AI will deliver on its potential. He talked about how AI systems are providing more accurate answers by doing thousands more computations for each request. The result is AI systems that are more useful and capable of providing services that people will want to pay for, like AI agents, that can autonomously perform tasks such as shopping for groceries.

But those computations mean that the demand will increase for more powerful computers. Enter Nvidia’s next line of chips.

In late 2026, Huang said, Nvidia will release its next generation of GPUs, called Rubin. The chips can be packaged into a supercomputer that includes four times as many GPUs as today’s model. The system will be 14 times more powerful than today’s supercomputer, and consume less power.

Huang is betting that those chips will account for a sizable share of the more than $1 trillion that analysts predict will be spent on data centers annually in 2028. “The more you buy the more you save,” Huang said.

The Rubin chip is crucial to Nvidia’s staying at the forefront of AI. The company faces challenges as its customers, including Amazon, Google and Meta, make their own AI chips. And Nvidia’s chips also have to change as AI companies try to get better performance out of their AI models.

“The gravy train comes to a screeching halt if cloud companies stop spending,” said Patrick Moorhead, founder of Moor Insights & Strategy, a tech research firm. He said Huang’s product road map “increased confidence” that Nvidia’s products will continue to be in high demand, so long as AI systems are creating business opportunities to justify their costs.

In addition to outlining Nvidia’s coming products, Huang said General Motors has committed to use Nvidia’s AI tools to design cars and plan its automotive factories. He said that Nvidia is also working with Google DeepMind and Disney Research on a software system to improve the precision of robots, which he demonstrated onstage with a waist-high robot shaped like Pixar’s Wall-E.

Huang’s ability to command a crowd is reminiscent of Apple’s Steve Jobs. Before major company events, the Apple co-founder spent days rehearsing his speeches about a new iPod, iPhone or iPad, before taking the stage to thunderous applause and seeming to deliver his remarks as though they were unscripted.

Huang, 62, similarly prepares in great detail for Nvidia GTC. Two months before the event, he works with the company’s product divisions to identify what to announce, said Greg Estes, Nvidia’s vice president of corporate marketing. Huang also works with the marketing team to develop slides and demonstrations to show onstage, creating bullet points and checking facts that he may cite.

But Huang never writes a speech, Estes said. When he takes the stage in his trademark black leather jacket, he speaks extemporaneously. A speech scheduled for 90 minutes can run more than two hours.

“Sometimes a mistake will happen and he’ll say, ‘You know, we don’t rehearse,’” Estes said. “He’s not kidding. It is ‘grip it and rip it.’”

Nvidia GTC was formerly the GPU Technology Conference, named after the graphics processing units. The event, which was designed to encourage developers to use the company’s chips, included a research summit where academics put up poster boards detailing how they had used the components for computing research. Huang spoke to attendees about what they did with the chips and, over the years, often heard that they were using them to develop AI.

In 2014, Huang began devoting the majority of his speech at the conference to the way Nvidia chips could be used for machine learning and AI. Gaming developers, who used GPUs to render video game graphics and had long been the heart of the company’s business, were angered by the shift.

“They were like, ‘What the hell is this shiny new thing?’” said Naveen Rao, chief AI officer at Databricks, which provides software tools for storing and analyzing large amounts of data. “We were like: ‘No. No. This is the sea change.’”

Huang bet that AI would drive tech’s next big boom and that GPUs would be essential. In 2016, Nvidia developed a supercomputer packed with its chips and delivered it to OpenAI, an AI lab. A little over six years later, OpenAI released the ChatGPT chatbot, unleashing an AI frenzy.

(The New York Times has sued OpenAI and its partner, Microsoft, for copyright infringement of news content related to AI systems. OpenAI and Microsoft have denied the claims.)

Since then, Nvidia’s finances have soared. The company, which was founded in 1993, increased its annual profit more than 1,500% in a two-year period to $72.88 billion last year from $4.37 billion in fiscal 2023.