Nvidia Up Next with New Chips, DeepSeek in Focus

Car enthusiasts used to eagerly await the new Oldsmobiles and Fords each year. Nvidia doesn't have showrooms, but if it did, tech investors might have their faces pressed to the glass to get a look at the Blackwells and Rubins.

Those are the newest AI chips Nvidia is expected to update investors on when it reports earnings after the close Wednesday. Blackwell's performance in its first quarter on the market, along with Nvidia's revenue guidance for coming quarters, could move shares of the AI giant.

Though AI chips can't be measured by horsepower or the time it takes to go zero to 60, customers clearly expect improved performance over past Nvidia chips, judging from spending plans recently shared by so-called hyperscalers like Amazon, Alphabet, and Meta. Though last month's news that China's DeepSeek might offer a cheaper AI alternative briefly sent Nvidia shares into a sell off, there's perhaps a sense that for now Nvidia may be the public's image of an AI chip leader.

That said, Nvidia's shares haven't really gone anywhere since last summer, leaving investors wondering what happened to the vigorous gains of 2022 and 2023. After topping $140 last July and briefly hitting $150 in early January, shares fell back into the $125 range by mid-February.

Recent concerns about flagging cloud growth and DeepSeek competition recently put a brake on Nvidia's stock, but shares lagged long before that news. There are also concerns about revenue guidance. Last time it reported, in November, Nvidia guided for fiscal fourth quarter revenue of $37.5 billion, up from $35.1 billion in the third quarter. That represents a 7% sequential increase, down from a 17% sequential increase achieved in the prior quarter and 15% the quarter before that.

"The key is Blackwell demand," said Nathan Peterson, director of derivatives analysis at the Schwab Center for Financial Research. "How demand is stacking up will be reflected in the guidance, but we've noticed the revenue beat on guidance, in recent history, had been getting smaller and smaller, and the stock price has reflected that deceleration in the guidance beat rate."

Another question entering earnings, besides any updates on Blackwell and Rubin, is whether CEO Jensen Huang still believes demand for AI chips is "insane," as he described it last October when discussing initial customer interest in Blackwell. Another question is whether the company is truly past some early manufacturing woes that temporarily slowed Blackwell production.

Blackwell, remember, is a chip expressly designed for AI use, one reason big expectations go with it. The other reason is planned spending from hyperscalers. The combination sounds potentially positive, but Nvidia's guidance on Blackwell is probably more meaningful from a market standpoint than any results from the fourth quarter when most Nvidia sales were of its older chips. There's a chance companies may have postponed purchases awaiting the new technology.

"Are we still 'insane' on the demand side? That's the question," Peterson said. "Some of the hyperscalers are developing their own chips and DeepSeek may have at least introduced the possibility that the level of compute needed for AI development is lower than initially anticipated. While we don't know whether either of those factors are enough to dent demand, Nvidia's guidance may be under higher scrutiny as a result. Also, 17% of Nvidia's revenue is from China, and that's caused some earlier volatility in shares."

China is a two-pronged threat, first in terms of DeepSeek potentially drawing domestic companies there away from U.S. chips, and second because of fresh tariffs and export controls leveled against China by both the Biden and Trump administrations. Nvidia criticized Biden's updated export rule last month that would cap the number of AI chips called GPUs (graphics processing units) that can be ordered by most countries without a special license.

"By attempting to rig market outcomes and stifle competition—the lifeblood of innovation—the Biden administration's new rule threatens to squander America's hard-won technological advantage," said Nvidia Vice President of Government Affairs Ned Finkle in a statement. It's unclear if Nvidia is lobbying the Trump administration to change the rule, but with Trump's early focus on tariffs against China, trade policy still looks like a possible headwind for Nvidia and other tech companies.

Additionally, the Trump administration, like the Biden administration before it, wants to bring chip manufacturing back to the United States, and recently re-emphasized that goal. Nvidia's chips are mainly manufactured by Taiwan Semiconductor Manufacturing (TSM). It's unclear how much traction the government might have in steering manufacturing of the most complex chips away from TSM, which dominates the field with its experience and facilities, but any change in manufacturing sites would likely complicate matters and be costly for Nvidia and other major chip firms.

Another hurdle is chip production. Earlier this month, TSM said it expects first quarter revenue to be near the lower end of its guidance due to an earthquake that slowed production in January. Nvidia and TSM have struggled to meet AI chip demand, Barron's reported, and that's possibly reflected in weaker-than-expected cloud performance from Microsoft (MSFT) and Alphabet (GOOGL). Microsoft, in its latest earnings call, said AI is "capacity constrained."

Though cloud growth disappointed pretty much across the industry last quarter and hurt shares of the Magnificent Seven, including Nvidia, it's a matter of semantics. The cloud business continued to grow around 20% to 30% annually for the biggest three in the business last quarter which is impressive considering robust growth the last few years.

This might bode well for Nvidia, but DeepSeek raised questions about where cloud providers might get their technology. Cheaper AI could pose another hurdle, and Nvidia executives will likely be asked about it on the earnings call.

"Nvidia is at the forefront of AI, but DeepSeek introduced the notion that Large Language Models (LLMs) are becoming commoditized, which presumably narrows the competitive gap between proprietary and open-source models. If less investment is needed to enter the AI space this invites more competition, and potentially faster adoption of AI. While Nvidia may still benefit from faster AI adoption, investors will want to be reassured of this when they report later this month," Peterson said.

Nvidia says that AI inference, in which pretrained AI models are deployed to generate new data and power innovation, is key to profitable AI. And the key to inference, Nvidia argues, is its own AI inference platform. This is one way it hopes to set itself apart from competition and help clients improve performance, which ultimately could translate to improved return on investment for those customers.

DeepSeek raised questions about return on investment and general demand for Nvidia's expensive AI platform. But after DeepSeek upended the market in mid-January, Nvidia was full of praise.

"DeepSeek is an excellent AI advancement and a perfect example of Test Time Scaling," an Nvidia spokesperson told CNBC. "DeepSeek's work illustrates how new models can be created using that technique, leveraging widely available models and compute that is fully export control compliant."

In other words, Nvidia seems to see DeepSeek as part of the anticipated evolution of AI. One argument bullish analysts make on Nvidia's behalf is that cheaper AI could raise AI demand, which ultimately would help hardware makers. The anti-Nvidia argument notes that during the internet boom years a quarter of a century ago, the hardware makers eventually lost profit ground to software makers as prices dropped.

One other thing to check when Nvidia reports is its latest gross margin, which dropped to 74.6% in the fiscal third quarter from 75.1% the prior quarter. Nvidia predicted another drop to 73% in the fourth fiscal quarter. Anything above or below that might grab attention, though all these numbers keep Nvidia quite profitable.

Third quarter data center revenue growth rose 17% from the prior quarter to $30.8 billion, up 112% from a year earlier. Some expect the yearly comparisons to get tougher and tougher as Nvidia laps its huge 2024 numbers, so sequential trends may be more important to watch.

Analysts expect Nvidia to report earnings per share of $0.84 when it shares results Wednesday, up from $0.81 in the company's third fiscal quarter and $0.52 a year ago, according to Schwab data. The average revenue estimate is $38.16 billion, according to Yahoo Finance.