Nvidia’s latest quarter gave the market a clear answer to the question hanging over the AI boom: demand is still translating into very large revenue. The company reported record results, pointed to strength in its data center business, and issued a fourth-quarter forecast that suggests management expects the surge to continue.
The numbers do not settle every debate about AI spending. But they do show that, from Nvidia’s vantage point, customers are still buying the hardware needed to train, run, and expand AI systems at scale.
A quarter built on data center demand
Nvidia reported revenue of $57 billion in the third quarter, up 62% from the same quarter last year. Net income on a GAAP basis was $32 billion, up 65% year-over-year. Both figures came in above Wall Street expectations.
The company’s data center business was the main engine. It generated a record $51.2 billion in revenue, rising 25% from the previous quarter and 66% from a year ago. That means the overwhelming share of Nvidia’s quarterly revenue came from the segment most directly tied to AI infrastructure.
The rest of the business was much smaller by comparison. Nvidia’s remaining $5.8 billion in revenue came from gaming, professional visualization, and automotive. Gaming contributed $4.2 billion of that total.
That split matters because it shows how sharply Nvidia’s center of gravity has moved toward infrastructure. The company is still active across several markets, but the data center business is now the clearest reflection of how AI spending is showing up in its financial results.
Blackwell becomes the center of the story
Nvidia’s results were also tied to demand for its Blackwell products. Blackwell Ultra, a GPU unveiled in March and offered in several configurations, has become the leader within the company, according to Nvidia. Earlier versions of the Blackwell architecture also continued to see strong demand.
Jensen Huang described the strength in unusually direct terms, saying Blackwell GPU chip sales "are off the charts." In the company’s Q3 earnings statement, he also said cloud GPUs are sold out.
"Compute demand keeps accelerating and compounding across training and inference — each growing exponentially."
That statement frames Nvidia’s view of the market. The company is not describing AI demand as a one-time buying cycle. It is presenting training and inference as two expanding workloads that are both pushing customers toward more compute.
For Nvidia, that is the core business case. AI models require computing power to be built, improved, and used. If both training and inference continue to grow, the need for GPUs and related data center infrastructure grows with them.
AI factories, infrastructure projects, and broad demand
Nvidia CFO Colette Kress said in a shareholder statement that the data center business has been fueled by accelerated computing, powerful AI models, and agentic applications. On the company’s Q3 call, she said Nvidia announced AI factory and infrastructure projects in the quarter amounting to an aggregate of 5 million GPUs.
Kress said demand spans several categories of buyers, including CSPs, sovereigns, modern builders enterprises, and super computing centers. She also pointed to multiple landmark build outs.
Those details are important because they show that Nvidia is not presenting demand as coming from only one type of customer. The company is describing a wider buildout of AI infrastructure, with projects across markets and buyer groups.
That does not remove execution risk. Large infrastructure plans still have to become actual shipments, deployments, and ongoing use. But Nvidia’s third-quarter commentary indicates that the company sees demand as broad rather than isolated.
China remains a weak spot
The quarter was not entirely without pressure. Kress said shipments of H20, a data center GPU designed for generative AI and high-performance computing, were 50 million. She described that as disappointing because Nvidia was unable to sell to China.
On the earnings call, Kress said sizable purchase orders did not materialize in the quarter because of geopolitical issues and a more competitive market in China. She said Nvidia was disappointed by the current state that prevents the company from shipping more competitive data center compute products to China.
She also said Nvidia remains committed to engagement with the U.S. and China governments and will continue to advocate for America’s ability to compete around the world.
That makes China one of the clearest complications in an otherwise strong quarter. Nvidia’s results show booming demand in data center compute, but the company is still navigating limits and competition in a market it cannot fully address in the way it wants.
The forecast shifts the AI bubble debate
Nvidia is projecting $65 billion in revenue for the fourth quarter. That outlook helped push the company’s share price up more than 4% in after-hours trading.
The forecast is significant because it extends the story beyond a single strong quarter. Nvidia is telling investors that the next quarter should be even larger than the one it just reported.
Huang directly addressed the concern that AI spending may be inflated. During the earnings call, he said there had been a lot of talk about an AI bubble, then added, "From our vantage point, we see something very different."
The company’s numbers explain why he would take that position. Revenue and profit rose sharply from a year ago, the data center business set a record, Blackwell demand remained strong, and the fourth-quarter forecast points higher.
For now, Nvidia’s answer to AI bubble concerns is not a theory. It is a financial result built on $57 billion in quarterly revenue, a data center business producing $51.2 billion, and a forecast of $65 billion for the next quarter.