Updated:

Chipmaker Nvidia Corp. today signaled that it’s on course for yet more artificial intelligence-driven growth in 2025 after delivering a solid fourth-quarter earnings and revenue beat and offering strong guidance for the current quarter.

The company extended its run of successful quarters, reporting earnings before certain costs such as stock compensation of 89 cents per share, beating the analyst estimate of 84 cents by a comfortable margin. Revenue for the period came to $39.33 billion, up 78% from a year ago and surpassing Wall Street’s $38.05 billion forecast.

Nvidia’s profitability also improved, with the company delivering a net profit of $22.09 billion during the quarter, up from $12.29 billion in the year-ago period.

Looking ahead to the current quarter, Nvidia is forecasting sales of $43 billion, plus or minus 2%, which suggests growth of about 65% from a year earlier. The forecast is better than expected, with Wall Street analysts guiding for revenue of just $41.78 billion in the first quarter of fiscal 2026.

Investors appeared happy enough with what they saw at first, as Nvidia’s stock gained just over 1% in extended trading, adding to an anticipatory rise of more than 3% during the regular trading session. However, some will have noted that the company’s balance sheet wasn’t entirely spotless, for it posted a decline in one key metric: Its gross margin was 73%, down 3% compared to a year earlier.

Gross margin is a measure of the percentage of revenue that remains after deducting the direct costs of producing its high-powered computer chips. According to Nvidia, the decline was related to the higher costs and complexity of its newest data center products.

Update: That’s likely why investors turned a bit negative the next morning, as shares on Thursday fell more than 5%.

Nvidia has been growing at a rapid rate thanks to the unprecedented demand among technology companies for its powerful graphics processing units, which power the vast majority of AI workloads today. It has become one of the world’s most valuable publicly traded companies. As a result, it has fallen under intense scrutiny from investors, many of whom are betting that it will be able to extend its good run into the new fiscal year.

The company’s fortunes in fiscal 2026 are directly tied to the success of its next-generation GPU Blackwell, which began shipping late last year and accounted for more than $11 billion in sales in the last quarter. Nvidia Chief Executive Jensen Huang (pictured) said the company is seeing “amazing” demand for the new chips.

In a conference call with analysts, Nvidia Chief Financial Officer Colette Kress said she’s expecting a “significant ramp” in sales of Blackwell during the current quarter. “Blackwell sales were led by large cloud service providers, which represented approximately 50% of our data center revenue,” she added.

Sales of the Blackwell chips are reported in Nvidia’s data center business unit, along with those of its previous-generation Hopper AI chips. That segment accounted for 91% of the company’s total revenue during the quarter, up from 83% one year ago and just 60% in the fourth quarter of 2023. Data center revenue has risen almost tenfold in the last two years.

All told, data center revenue came to $35.6 billion in the quarter, increasing 93% on an annual basis, and coming in ahead of the Street’s forecast of $33.65 billion.

Third Bridge analyst Lucas Keh said Nvidia is continuing to show strong momentum thanks to its strong relationship with data center hyperscalers such as Amazon Web Services Inc. and Meta Platforms Inc., which make up just over 50% of its total data center revenue.

“It’s a very positive sign that Nvidia disclosed its Blackwell sales numbers during the quarter, which were nearly one-third of the total data center revenues,” Keh said. “This shows that the issues that led to the initial delays in Blackwell’s launch should largely be resolved.”

In the conference call, Huang said that whereas the company’s older chips were primarily used to train AI models, the new Blackwell chips are primarily driving AI inference, running AI models and applications in production. It’s a key distinction, Nvidia’s executives believe.

In recent weeks, some investors have raised concerns that demand for Nvidia’s most powerful chips might tail off with the rise of China’s DeepSeek, whose low-cost DeepSeek R1 reasoning model has taken the industry by storm after being developed for only a few million dollars. However, Kress sought to reassure those investors, saying that newer models designed to “think” more carefully about their responses will likely require much more computing power compared to earlier generative AI models.

“Long-thinking, reasoning AI can require 100 times more compute per task compared to one-shot inferences,” Kress told analysts.

Huang reiterated that, saying that the “vast majority of compute today is actually inference.” He believes that in the coming years, newer-generation AI models may even need “millions of times” the current amount of computing capacity that’s available to them.

Another fear investors have is that Nvidia’s status as the world’s top GPU maker could be threatened by rivals such as Amazon Web Services Inc., Google LLC and Microsoft Corp., which have all been working on their own, customized AI accelerators.

“Just because the chip is designed doesn’t mean it gets deployed,” Huang said, noting those rivals still have a long way to go to catch up.

According to Keh, some of those competitors are starting to take a toll on Nvidia’s position in the AI chip market, but their increasing competitiveness is not very material at this point. As such, he believes Nvidia should continue to deliver a double-digit sequential growth rate in its data center segment in the coming quarters, as Blackwell adoption heightens.

“The stock may be performing a bit poorly after this continued good news as the expectations continue to be extremely high for Nvidia, and as investors start to weigh and consider some downside risk,” Keh added.

Some of the other parts of Nvidia’s business didn’t fare so well. The company also sells networking equipment within its data center segment, which accounted for about $3 billion in sales, down 9% from a year earlier.

There’s also the company’s gaming business, where it makes GPUs for personal computers and games consoles to cater for graphics-intensive gaming. That segment’s revenue fell 11% from a year ago, to $2.5 billion in sales, well short of the Street’s forecast of $3.04 billion. Officials are hopeful of an improvement soon, though, as the company has recently announced its first graphics cards based on the Blackwell architecture.

The company also has a very small business selling chips for cars. It said automotive chip sales more than doubled from a year earlier, though the segment remains tiny next to the AI business, with revenue of just $570 million.

Holger Mueller of Constellation Research Inc. said all of Nvidia’s year-over-year comparisons are out of this world for such an established company, with double the revenue and operating and net income growing by 140%.

“The quarter-over-quarter growth rates were slower, which shows that Nvidia’s growth is settling back down into more realistic numbers, but importantly they’re also sustainable numbers, for now anyway,” the analyst said. “What’s also remarkable is Nvidia’s cost control, showing considerable thrift and frugality to make the company more profitable than it was a year ago. That’s never a bad trait in such a volatile chipmaking industry.”

Despite today’s gains, Nvidia’s stock is still down 2% in the year-to-date.

Photo: Robert Hof/SiliconANGLE