Nvidia tops Wall Street’s expectations again as data center revenue surges 75%

Nvidia Corp. sailed past analysts’ expectations again as it delivered its latest financial results today. Revenue in its core data center business jumped more than 75% from a year earlier, sending its stock higher in after-hours trading.

Once again, the company delivered solid fourth-quarter results across the board. Earnings before certain costs such as stock compensation came to $1.62 per share, easily beating Wall Street’s target of $1.53 per share. Meanwhile, overall revenue soared 73% to $68.13 billion, surpassing the $66.21 billion consensus.

Data center revenue came to $62.3 billion in the quarter, ahead of the Street’s $60.69 billion target. The segment, which manufactures chips for artificial intelligence workloads, now accounts for just over 91% of Nvidia’s total sales. The AI business is not only growing fast, but also proving to be extremely profitable for Nvidia, too. The chipmaker’s net income almost doubled to $43 billion, up from $22.1 billion in the year-ago quarter.

To cap it all off, Nvidia offered strong guidance for the coming quarter. It said it’s looking for revenue of around $78 billion in the first quarter of fiscal 2027, plus or minus 2%. Wall Street is looking for total sales of just $72.6 billion. Even better, Nvidia said it has discounted all possible revenue from China in its forecast.

Nvidia, led by its Chief Executive Jensen Huang (pictured), has outperformed all of its megacap peers so far this year as it continues to benefit from the rampant growth of AI. As of today’s market close, the shares have gained 5% this year, while the broader tech-heavy Nasdaq index is down 0.4%. Apple Inc. is the only other company with a trillion-dollar-plus market capitalization to show gains this year, and even then its stock has gained less than a percentage point.

Hyperscale dominance

Most analysts were expecting Nvidia to deliver, as it has done consistently since the AI boom first kicked off in late 2022. In the last couple of weeks, Alphabet Inc., Amazon.com Inc., Microsoft Corp. and Meta Platforms Inc. have all reported their latest financial results, and based on their capital expenditure forecasts, it was clear that demand for Nvidia’s chips is higher than ever. Those four companies are expecting to spend almost $700 billion on building out their AI data centers this year, and a good chunk of that change is earmarked for Nvidia’s graphics processing units.

Not surprisingly, Nvidia’s finance chief Collete Kress said that hyperscalers are still the company’s “largest customer category,” accounting for just over half of the revenue from its data center business. However, analysts told SiliconANGLE that there are valid reasons to wonder if that will always be the case.

Emarkater analyst Jacob Bourne told SiliconANGLE that hyperscaler demand for Nvidia’s chips remains as robust as ever, but questioned if it’s always going to be remain that way in future. “The competitive picture is slowly shifting as companies like Meta diversify towards AMD and the big cloud players invest more in custom silicon,” he said. “This puts the focus on Nvidia’s guidance for what the future holds in terms of maintaining its dominance as the AI buildout matures and questions around enterprise ROI intensify.”

Nvidia’s performance isn’t just about chips, though. Within the data center segment, the company sold more than $10.98 billion worth of networking gear, which is used to connect its GPUs into massive clusters to run large language models. All told, networking sales were up 263% from a year earlier. That reflects the strong adoption of its new NVLink technology and its Spectrum-X Ethernet switches, the company said.

The explosion of new revenue from its networking business is extremely eye-catching, said Fusion Collective analyst Yvette Schmitter. “It appears that the company is trying to lock-in its customers at scale,” she said. “Companies can’t buy GPUs without also buying the networking and software stack.”

However, Schmitter questioned why Nvidia only returned $4 billion to shareholders through dividends in the quarter, despite generating more than $35 billion in cash flow. “That’s only 12% of its profit, compared to the 54% returned to shareholders a year ago,” she said. “This is happening at a time when Nvidia claims that its Ampere chips being sold out is a good signal for demand. But why is it cutting buybacks by half?”

“The hyperscale growth is impressive, but it doesn’t mean Nvidia is bulletproof,” Schmitter continued. “With gross margins flat at 75% despite everything that’s been happening, we could be witnessing its plateau.”

Gaming revenue soars, but maybe not for long

It’s hard to believe nowadays that the gaming segment was once Nvidia’s largest. All the more so considering that it posted record quarterly growth today, with revenue up 47%. Even so, the unit’s sales “only” came to $3.7 billion overall. However, analysts have warned that the gaming business could suffer later this year, as Nvidia may decide not to launch a new gaming GPU due to the shortage of memory chips. There’s not enough memory to go around, and chipmakers like Nvidia are prioritizing deliveries to the AI industry first, which is likely to harm console sales this year.

Kress told analysts on a conference call that memory supply constraints will be a headwind for the gaming business beyond the first quarter of fiscal 2027. Fortunately for Nvidia, analysts are much more focused on AI, and there has been a lot of excitement around the upcoming launch of the company’s next-generation Vera Rubin rack-scale GPU systems. Vera Rubin is the successor to Grace Blackwell and it’s slated to launch later this year. According to Kress, the company has now shipped its first samples of the new chips to customers, and remains on track to start production shipments in the second half of the year.

Nvidia has said Vera Rubin will deliver 10-times more performance-per-watt than Grace Blackwell, which should prove to be extremely beneficial at a time when AI data centers are facing significant power constraints.

The chipmaker also provided an update on its much smaller businesses. The automotive segment, which makes chips for autonomous vehicles and robots, delivered sales of $604 million in the quarter, up 6% but below the expected $654.8 million. There’s also the professional virtualization segment, which generated revenue of $1.23 billion, far ahead of the Street’s $755.4 million target.

On the conference call, Huang discussed the company’s investments in large AI labs and other AI industry players, including the chipmaker Intel Corp. and numerous startups. The company spent more than $17.5 billion on such investments over the last year, with the bulk of that money used to “support early-stage startups.”

Huang also provided an update on the company’s ongoing negotiations with the AI industry’s leading model maker OpenAI Group PBC. “We continue to work with OpenAI towards a partnership agreement and believe we are close,” Huang said. The companies first announced they were discussing a $100 billion deal in September, but nothing has been finalized so far.

Photo: Nvidia