Getty Images/iStockphoto
Nvidia reports strong Q4, prepares for new AI GPUs
Nvidia's revenue soared 78% to $39.3 billon, driven by strong AI GPU demand. With Blackwell's success, a more powerful variant launches later this year.
Nvidia's latest earnings report indicates demand for AI computing remains strong, with no signs of a slowdown as the company prepares to roll out a more powerful version of its Blackwell GPU later this year and a next-generation AI chip in 2026.
Blackwell, released in early December, contributed $11 billion in revenue in the fiscal fourth quarter ended Jan. 26, executives said after releasing earnings on Wednesday. This represents the fastest product ramp in the company's history.
Total revenue for the quarter rose 78% from a year ago to $39.3 billion, beating Wall Street analyst estimates. Most of the revenue, $35.6 billion, came from data center products.
Cloud service providers, including AWS, Google Cloud, Microsoft Azure and Oracle Cloud Infrastructure, accounted for half of Nvidia's data center revenue, executives said. The cloud providers deployed Nvidia's new GB200 systems in regions globally.
The GB200's core technology is a highly advanced processor that comprises Nvidia's Grace CPU and dual Blackwell B200 GPUs. The Grace CPU is named after American computer programming pioneer Grace Hopper, and the Blackwell GPU is named after American mathematician David Blackwell.
Analysts were watching Nvidia's earnings closely for signs of a slowdown in AI spending, particularly among cloud service providers that planned to spend tens of billions of dollars this year on data center infrastructure. Some analysts said a pullback was possible following Chinese startup DeepSeek's development of an advanced AI model using less advanced and cheaper technology than Nvidia's latest processors.
Jensen Huang,
Nvidia is on track to introduce Blackwell Ultra in the second half of the year, CEO Jensen Huang said during an earnings call with analysts. He added that he expected an easier transition from Blackwell to Blackwell Ultra than from Hopper to Blackwell.
A design flaw in moving to Blackwell initially caused low Blackwell chip production yields. Blackwell required a new chassis architecture and power delivery system that made the transition from the Hopper architecture difficult, Huang said.
"This was quite a challenging transition, but the next transition will slide right in," he said. However, Blackwell Ultra will have new networking, memory and processors.
"Whatever impact the Hopper/Blackwell transition had, it was clearly outweighed by the combination of sheer demand and [Nvidia chip manufacturer] TSMC's achievement in driving its capacity expansion forward through 2024 faster than anyone anticipated," said Alexander Harrowell, an analyst at Omdia, a division of Informa TechTarget.
In 2026, Nvidia will launch the Vera Rubin architecture, which comprises a Vera CPU and Rubin GPU, with the latter manufactured on Taiwan Semiconductor Manufacturing Co.'s 3-nanometer process. Vera Rubin combines the GPU and CPU into a single platform that succeeds the current Grace Blackwell offering.
"All of our partners are getting up to speed on the transition," Huang said. "And again, we're going to provide a big, big, huge step up."
Huang promised to release more details on Vera Rubin and Blackwell Ultra at Nvidia's annual GTC conference next month.
Tariffs, trade restrictions
One country unlikely to get Nvidia's most advanced technology is China. Nvidia's revenue from China has fallen by half since the U.S. started restricting chip exports to the country in 2022, according to Huang. Nvidia makes a less advanced China-specific processor called the H20.
The Trump administration's threatened tariffs on chips manufactured outside the U.S. could also affect Nvidia's future revenue. TSMC manufactures most of the company's processors.
"At this point, it's a little bit of an unknown," Nvidia CFO Colette Kress said. "It's an unknown until we understand further what the U.S. government's plan is."
Huang predicted that agentic AI, physical AI for robotics and sovereign AI will play major roles in the next AI wave, which will follow the current development of generative AI in business and consumer software. Agentic AI refers to systems that act autonomously; physical AI is used in robotics, and sovereign AI refers to a nation's development and control of AI.
"Each of these is barely off the ground, but we can see them," Huang said. "We can see them because we're in the center of much of this development, and we can see great activity happening in all these different places."
Antone Gonsalves is an editor at large for Informa TechTarget, reporting on industry trends critical to enterprise tech buyers. He has worked in tech journalism for 25 years and is based in San Francisco. Have a news tip? Please drop him an email.
