In This Story
With its highly-anticipated second-quarter earnings results, Nvidia NVDA beat expectations again, reporting record second-quarter revenue of $30 billion for fiscal year 2025 — up 122% from a year ago.
Nvidia was expected to report revenue of $28.7 billion for the second quarter — more than double its previous year revenue of $13.5 billion, according to analysts’ estimates compiled by FactSet FDS. The company was also expected to report earnings per share of $0.65, according to FactSet, but reported adjusted EPS of $0.68.
“Hopper demand remains strong, and the anticipation for Blackwell is incredible,” Jensen Huang, founder and chief executive of Nvidia, said in a statement. “Nvidia achieved record revenues as global data centers are in full throttle to modernize the entire computing stack with accelerated computing and generative AI.”
Despite the beat, investors seem to have been expecting a wider margin. The chipmaker set its third-quarter revenue guidance at $32.5 billion, plus or minus 2% — above the average of what analysts are expecting, but below top end estimates. Nvidia’s shares were down around 3.9% in pre-market trading on Thursday, and were down 3.7% during mid-day trading.
Here are four takeaways from Nvidia’s second-quarter earnings report and post-earnings call with analysts.
The Blackwell GPU got a design-change
To “improve production yield,” Nvidia made a change to its Blackwell GPU mask, Colette Kress, chief financial officer of Nvidia, said on the call with analysts. However, “there were no functional changes necessary,” Huang said on the call. The company expects to “ship several billion dollars in Blackwell revenue” in the fourth quarter, and demand “is well above supply,” it said. The Blackwell platform’s production ramp is set to start in the fourth quarter into fiscal year 2026.
During Nvidia’s first-quarter earnings call, Huang said its Blackwell AI platform would start shipping in the second quarter, ramp up in the third quarter, and be with customers in the fourth quarter. However, earlier this month, Nvidia’s shares fell after a report that Blackwell was delayed due to design flaws.
Meanwhile, demand for Nvidia’s Hopper chips remains strong while supply has improved, and the company expects shipments to increase in the second half of the fiscal year. More customer adoption of its Hopper architecture and sampling of its Blackwell AI platform are expected to drive third-quarter revenue, the company said.
Generative AI is driving data center growth
Nvidia reported data center revenue was a record $26.27 billion for the second quarter, up 16% from the previous quarter, and a year-over-year increase of 154%. Part of that growth was driven by demand for the company’s Hopper graphics processing units, or GPUs, which are used for training and inferencing some of the world’s most powerful generative artificial intelligence models, the company said.
Other drivers of growth included video, image, and text data pre- and post-processing with the company’s CUDA computing platform, and AI-powered recommender systems, Kress said. And growth “is expected to continue” as next-generation models will require more power to train with more data.
Business in China is expected to remain “very competitive”
Data center revenue in China also grew quarter-over-quarter, and was “a significant contributor to” the company’s data center revenue, Kress said. However, as a percentage, China “remains below levels seen prior to the imposition of export controls.”
Still, Nvidia continues “to expect the China market to be very competitive going forward,” Kress said.
Meanwhile, the U.S. is reportedly considering imposing tougher trade rules to curb shipments of advanced chipmaking equipment to China, and could ban Nvidia from selling its H20 chip — one of three it specifically designed to not require an export control license — according to Jefferies JEF analysts.
Nvidia is also reportedly working on a version of its Blackwell AI platform for the Chinese market that will comply with U.S. trade restrictions. The chipmaker will reportedly work with a local distribution partner, Inspur, to launch and sell the chip, tentatively called the “B20,” in China.
Countries are leaning more on AI
The company’s AI revenue from sovereign customers is expected to “continue to expand,” Nvidia said, “as countries recognize AI expertise and infrastructure as national imperatives for their society and industries.”
In Japan, for example, the National Institute of Advanced Industrial Science and Technology is working with Nvidia on building a supercomputer. This year, Nvidia said it expects sovereign AI revenue to “reach low double-digit billions.”
Meanwhile, enterprise AI is also expanding, and drove quarter-over-quarter revenue growth in the second quarter.