Nvidia stock rose Wednesday after Chief Executive Jensen Huang told financial analysts that estimates for the artificial intelligence infrastructure buildout are too low.
In a meeting with analysts at the GTC conference in San Jose, Calif., Huang said current forecasts typically only consider data center spending plans from cloud service providers. They fail to account for new data center builds from projects like Stargate and xAI and enterprises building AI factories for their own businesses.
Plus, current general-purpose computing systems will need to be upgraded to accelerated computing, he said.
"We are going to have to modernize the world's enterprise IT," Huang said.
For now, Nvidia has its hands full serving the needs of cloud data centers, he said. But the other opportunities are on the way.
Aside from underestimating the AI infrastructure opportunity, another area that people have gotten wrong was the impact of China's DeepSeek AI model, Huang said. Many worried that cloud service providers would cut their spending in response to DeepSeek's low-cost approach. That hasn't happened, Huang said.
He also downplayed the competitive threat from companies making application-specific integrated circuits, or ASICs. That's because Nvidia has a proven, full-stack technology approach to AI data centers, he said. Data centers need more than a chip. They need fully architected computer systems, networking and other infrastructure, he said.
"If you are building a data center today, there is only one option now," Huang said. "We're the company you can build your AI infrastructure on."
If a company is going to commit $100 billion to build a data center, it's not going to choose a riskier option, he said.
"Everybody's still trying to catch up to Hopper, and I haven't seen a competitor to Hopper yet," Huang said. Hopper was Nvidia's previous-generation graphics processing unit, or GPU. Blackwell is its current-generation product.
Plus, Nvidia has detailed its product roadmap for the next three years so its customers can plan accordingly, Huang said.
Nvidia Stock Rebounds
In other news, Huang said Nvidia is preparing to manufacture its processors in the U.S. Chip foundry Taiwan Semiconductor Manufacturing is now running production silicon for Nvidia at the TSMC plant in Arizona, Huang said.
On the stock market today, Nvidia stock rose 1.8% to close at 117.52.
On Tuesday, Nvidia dropped 3.4% to close at 115.43. That was after Huang gave a keynote speech at the company's GTC conference.
Wall Street analysts were generally upbeat about Nvidia's presentation, saying it reinforced bullish forecasts for data center spending.
"The demand for Nvidia GPUs remains extremely robust," Wedbush Securities analyst Daniel Ives said in a client note. "We believe demand is currently outstripping supply 15:1 with more enterprises waiting their turn in line to receive the most advanced AI chips on the market."
Nvidia is forecasting data center capex to increase from $500 billion in 2025 to $1 trillion by 2030.
Huang officially debuted Nvidia's next-generation AI processor, called Blackwell Ultra. He showed off Blackwell Ultra server racks, full computers and new networking technology along with AI inference software.
Blackwell Ultra-based products are expected to be available from hardware partners in the second half of 2025.
What's Next From Nvidia?
Huang also outlined Nvidia's processor roadmap. It will release the successor to Blackwell Ultra, called Rubin, in the second half of 2026. Rubin Ultra will come out in 2027, followed by Feynman in 2028.
"Once a year, like clock ticks," he said about the company's product cadence.
Rosenblatt Securities analyst Kevin Cassidy reiterated his buy rating on Nvidia stock with a price target of 220 after the keynote.
"We see this annual cadence as a formidable moat to competition including ASICs (application-specific integrated circuits) and expanding Nvidia's AI dominance," Cassidy said in a client note.
Huang discussed how artificial intelligence is evolving from generative AI to agentic AI. Systems are moving from generating educated guesses from user prompts to providing more accurate answers and working on a user's behalf.
Huang unveiled new inferencing software called Nvidia Dynamo for scaling AI reasoning models. The software promises faster response time while reducing operating costs. "It is essentially the operating system of an AI factory," Huang said.
Nvidia also introduced new networking switch technology that it claims will be a big energy saver for data centers. The products use silicon photonics technology.
Evercore ISI analyst Mark Lipacis maintained his outperform rating on Nvidia stock with a price target of 190.
"No one is investing at the pace or magnitude Nvidia is in building out a full-stack chip+hardware+networking+software ecosystem for the AI computing era," Lipacis said in a client note. His contacts at hyperscale cloud service providers tell him that Nvidia has an eight-year lead.
A Firehose Of News From GTC
Santa Clara, Calif.-based Nvidia issued 37 press releases and blog posts on Tuesday, detailing the company's GTC news for the day. And Nvidia's many partners posted their own announcements.
Nvidia touted several major partnerships at the show.
For starters, it is collaborating with General Motors on factory automation and next-generation vehicles, including self-driving cars.
Plus, Nvidia disclosed a partnership with GE HealthCare to develop autonomous imaging machines, focused on X-ray and ultrasound applications.
Nvidia unveiled a team-up with telecom industry leaders to develop AI-native wireless networks for upcoming 6G systems. Partners on that project include T-Mobile, Cisco Systems, Mitre, Booz Allen Hamilton and Cerberus Capital Management.
Nvidia Enters The Quantum Realm
Nvidia also made news on the quantum computing front. It announced that it is building a Boston-based research center to advance the promising technology.
The Nvidia Accelerated Quantum Research Center will integrate leading quantum hardware with AI supercomputers. It will focus on solving quantum computing's most challenging problems, ranging from tackling qubit noise to the transformation of experimental quantum processors into practical devices.
The research center is expected to begin operations later this year.
On Thursday at GTC, Nvidia is hosting discussions on progress and challenges for quantum computing.
Investors mostly shrugged at the GTC news on Tuesday. Year to date through Tuesday's close, Nvidia stock is down 14%.
"Nvidia's annual GTC AI conference showed few signs of the skepticism that pervades the tech tape," TD Cowen analyst Joshua Buchalter said in a client note. The company provided a "strong pushback against the DeepSeek bear theses," as it showed that AI reasoning will require more than 100 times more compute than traditional large language models, he said.
Buchalter kept his buy rating on Nvidia stock with a price target of 175.
View More Nvidia Stock News And Analysis
Follow Patrick Seitz on X, formerly Twitter, at @IBD_PSeitz for more stories on consumer technology, software and semiconductor stocks.
