Nvidia Is Still Poised for a Blockbuster Year. And the Stock Remains Cheap.

Nvidia CEO Jensen Huang recently said that the company’s latest line of GPUs are in full production, with output ramping up after initial manufacturing obstacles. (PATRICK T. FALLON/AFP via Getty Images)

Investors shouldn’t get distracted by the noise about China. Nvidia is positioned to benefit from its most important and largest product ramp in its history. The signs are there.

Late Wednesday, the chip maker delivered another impressive, eye-watering report, with unprecedented growth rates for a company its size. Revenue for the July second fiscal quarter was up 56% year over year to $46.7 billion, ahead of expectations. Nvidia’s outlook was solid. For the current quarter ending in October, the company provided a revenue forecast range with a midpoint of $54 billion, which was above analysts’ consensus of $53.4 billion.

But Nvidia shares initially fell modestly after the results, which likely has more to do with some profit-taking after the stock rose about 35% over the prior three months. Uncertainty over its business in China may have been another factor.

To be sure, China remains Nvidia’s biggest problem. While some analysts had hoped for a quick rebound in sales of Nvidia’s China-specific H20 graphics processing unit, or GPU, it hasn’t happened. In fact, on Wednesday, Nvidia said there were no H20 sales from China-based customers in the second quarter at all. The company also hasn’t assumed any H20 shipments to China in its guidance for the current quarter.

“We continue to work through geopolitical issues,” Nvidia CFO Colette Kress said on the earnings call.

It’s worth noting that the company’s fiscal third-quarter outlook still beat Wall Street’s current estimates, even without the benefit of H20 sales.

Nvidia’s Chinese sales have become a casualty of the continuing back and forth between the U.S. and China. In April, the U.S. government moved to effectively ban sales of Nvidia’s H20 chips to China by tightening export licensing requirements. In July, the U.S. reversed course and said it would approve those H20 licenses. But China reportedly grew frustrated after disparaging remarks from U.S. Commerce Secretary Howard Lutnick and told companies not to order H20s due to security concerns.

Last week, Nvidia CEO Jensen Huang said he was surprised by the latest development given that China’s government had asked Nvidia to secure H20 export licenses from the U.S. government. “Hopefully, the response that we’ve given to the Chinese government will be sufficient. We are in discussions with them,” he said. Huang has denied the existence of any kind of security backdoor built into Nvidia’s H20 chips.

Despite all the noise, there’s a good chance things will get resolved given China had originally asked Nvidia for H20 licenses and with the high demand for Nvidia GPUs in the Asian country. The latest developments are likely more theatrics than a substantive change in views.

The H20 was designed to comply with restrictions that prohibit China’s access to the most advanced AI hardware. Despite those limitations, Chinese companies still prefer the H20 to run AI inference workloads reliably, as Nvidia’s software is more stable than anything created in China.

If the two economic superpowers come to terms, the benefit to Nvidia would be significant. Back in May, Nvidia said it could have sold $8 billion of H20 chips in the second quarter, absent regulatory hurdles.

Huang also said sales of a newer Blackwell chip for China is a “real possibility” in the future, but that it was too soon to know whether that would be approved.

This area of the company’s business is unpredictable and could change at any moment. The bottom line is that the China revenue for Nvidia is all upside from here.

Ultimately, Nvidia investors should focus on the company’s main AI chips, which will drive positive results no matter what happens with China. The company is still poised to benefit from its most important product releases in its history: the 72 GPU rack servers, called the GB200 NVL72 and the GB300 NVL72.

The NVL72 systems incorporate 72 GPUs, linked together inside one server rack, up from eight GPUs in the previous version. It’s an unprecedented density of computing power that is crucial to train the latest AI models.

Huang has confirmed that the new racks are in full production and that output is ramping up after initial manufacturing obstacles. Each rack costs several million dollars.

“It delivers an exceptional generational leap,” Huang said on the earnings call. Nvidia’s NVL72 “rack-scale computing is revolutionary, arriving just in time as reasoning AI models drive order of magnitude increases in training and inference performance requirements.”

There is evidence the rack server ramp is under way. Nvidia’s networking revenue segment surged 46% quarter on quarter in the July quarter. That’s a good sign, as NVL72 systems require more networking hardware per rack.

Over the course of the latest earnings season, we’ve learned that Big Tech firms continue to increase their capital expenditure budgets while emphasizing that demand for AI is outstripping capacity. Nvidia’s NVL72 rack servers are the product needed by start-ups and enterprises as they clamor for more AI computing resources.

With the production of Nvidia’s AI rack servers just ramping up, several quarters of strong financial performance should be on the way. While Nvidia stock has doubled from April’s lows, it isn’t overpriced. Shares trade at 32 times forward earnings—almost cheap, given the 50% earnings growth forecast by Wall Street analysts for the current fiscal year.

In this column, I have repeatedly said that Nvidia shareholders are best served by taking a long-term view on the stock and not selling in response to volatility unless there is a significant change in Nvidia’s long-term outlook. That advice has paid off, with Nvidia touting a world-topping $4.4 trillion market value.

Nvidia’s underlying fundamentals remain as strong as ever. Hold on.

Write to Tae Kim at tae.kim@barrons.com