Investors craving a boring backdrop for Nvidia earnings this week can just forget it. The main source of the hoopla is the emergent Chinese startup DeepSeek, which rattled the Club stock late last month with claims of more efficient AI models. Nvidia’s next-generation Blackwell chip platform has also faced rollout constraints. In other words, CEO Jensen Huang has plenty to talk about Wednesday evening when the dominant maker of artificial intelligence chips delivers fiscal fourth-quarter results. For the November to January period, Wall Street expects Nvidia to report earnings per share of 84 cents on revenues of $38.05 billion, implying year-over-year growth of 61.5% and 72%, respectively. Nvidia is also expected for the first time to provide a look at fiscal 2026 numbers. The consensus estimate for revenue in the company’s fiscal first quarter, which ends in April, is roughly $42 billion. The initial market panic over DeepSeek has calmed, and Nvidia shares have erased most of their 18% tumble in the first week of the fallout, trading at roughly $135 each Monday. The stock fell from $142.62 on Jan. 24 before closing as low as $116.66 on Feb. 3. Still, the overall tenor of the AI conversation feels different, and questions circling Monday about Microsoft’s data-center lease plans underscore the uneasiness that exists among some AI investors. In about 48 hours, though, Wall Street analysts will finally get a chance to ask Huang, widely regarded as an AI visionary, questions about DeepSeek’s implications for Nvidia’s business. The reason Nvidia shares plunged following DeepSeek’s arrival was a concern that its claims of efficient, lower-cost AI models — though probably not as low cost as initially assumed — will sharply reduce the number of advanced chips needed in the future to train and run AI applications on a day-to-day basis, a process known as inference . The counterargument was AI models that are cheaper to develop and deploy should make AI more accessible and boost adoption by companies and consumers alike — and, ultimately, require significantly more computing power in aggregate. NVDA 1Y mountain Nvidia’s stock performance over the past 12 months. Huang’s only public comments on DeepSeek came to light last week in a taped conversation with DDN chief executive Alex Bouzari that aired during a virtual event held by that data-storage company. The comments may foretell the arguments Huang makes on Wednesday’s call, so it’s worth taking a closer look at them. In the DDN interview, Huang suggested that the stock market reaction to DeepSeek was based on an incomplete understanding of AI model training, a compute-heavy process in which models are fed massive amounts of data — such as text, photos and video — to learn patterns within the dataset . The models rely on those learned patterns to generate outputs during inference. Huang said that many investors failed to recognize that training is a multistep process, with an initial “pretraining” phase followed by “post-training,” where the model is fine-tuned and optimized before being deployed into inference. Post-training is “where you learn to solve problems,” Huang said. “There’s a whole bunch of different learning paradigms that are associated with post-training, and in this paradigm, the technology has evolved tremendously in the last five years and the computing needs [are] intensive,” he added. With some of DeepSeek’s innovations, “people thought, ‘Oh my gosh, pretraining is a lot less.’ They forgot that post-training is really quite intense,” he said. Huang also said that DeepSeek’s top-end model, known as R1, is considered a reasoning model, which breaks down user prompts into smaller pieces and spends more time “thinking” before generating an answer. In general, these types of models are capable of handling more advanced tasks than the one that powered OpenAI’s ChatGPT when it first launched in late 2022 and sparked the AI boom. OpenAI, for its part, launched its reasoning model late last year and released a new version in January. “Reasoning is a fairly compute-intensive part,” Huang said. “And so, I think the market responded to R1 as an, ‘Oh my gosh, AI is finished. … We don’t need to do any computing anymore.’ It’s exactly the opposite.” Nvidia’s launch of Blackwell also will command considerable focus on Wednesday evening — and as far as the short-term impact on the financials goes, it holds a lot more weight than DeepSeek. Baked into Nvidia’s fiscal Q4 revenue guidance of roughly $37.5 billion, which is a bit lower than the current consensus, was at least “several billion dollars” of Blackwell sales. Succeeding Nvidia’s blockbuster Hopper lineup, Blackwell entered full production in the fall — and by early January, Huang said at the CES trade show in Las Vegas that “every single cloud service provider” had Blackwell-based systems “up and running.” However, the process of manufacturing and installing the top-of-the-line Blackwell product — a full server rack version known as the GB200 NVL72 — has faced some challenges, analysts say. For example, server maker Hewlett Packard Enterprise announced its first GB200 shipment less than two weeks ago . But that doesn’t mean Nvidia’s quarterly numbers, or guidance for its fiscal 2026 first quarter ending in April, are completely at risk of coming up short. In a note last week, analysts at KeyBanc Capital Markets said that despite lower GB200 rack shipments, they expect a different, relatively simpler Blackwell product called the HGX B200, plus shipments of H20 chips to Chinese customers, to “drive upside.” In fact, KeyBanc said DeepSeek’s emergence has helped spark a “surge in demand” for H20s, which are throttled-back versions of Hopper chips intended to comply with U.S. trade restrictions on advanced AI chip exports to China. While there is a growing belief that the U.S. will tighten export controls in response to DeepSeek, in the near term, China appears to be a source of strength. The Hopper-to-Blackwell transition “has been a little bumpier than anticipated,” Melius Research analysts wrote to clients Monday. “However, sales prospects still seem great by F2Q26 – and there is already a lot of chatter about the Blackwell Ultra (GB300), likely to be unveiled at its GPU Technology Conference (GTC) in March and shipping by [calendar year-end].” GTC is Nvidia’s annual developer conference and product show and tell. Put it all together, and Melius suggested that Nvidia’s guidance for the April quarter may not exceed consensus by as much as investors have come to expect in recent years. But it won’t be long before “a really big bump” in revenue thanks to Blackwell materializes, analysts argued. Bottom line Nvidia investors have their hands full entering Wednesday evening. “The quarter feels poorly set up,” as Jim Cramer wrote in his Sunday column . To be sure, the DeepSeek saga in recent weeks tested Jim’s resolve, something he acknowledged during our Monthly Meeting last week . However, he said his faith in Huang’s stewardship of the company remains strong, and that’s why he designated Nvidia an “own it, don’t trade it” stock to begin with. The good news is that with so many questions surrounding Nvidia’s earnings report, investors are finally going to get answers directly from Huang and respected finance chief Colette Kress. We’ve been waiting for that. (Jim Cramer’s Charitable Trust is long NVDA, MSFT. See here for a full list of the stocks.) As a subscriber to the CNBC Investing Club with Jim Cramer, you will receive a trade alert before Jim makes a trade. Jim waits 45 minutes after sending a trade alert before buying or selling a stock in his charitable trust’s portfolio. If Jim has talked about a stock on CNBC TV, he waits 72 hours after issuing the trade alert before executing the trade. THE ABOVE INVESTING CLUB INFORMATION IS SUBJECT TO OUR TERMS AND CONDITIONS AND PRIVACY POLICY , TOGETHER WITH OUR DISCLAIMER . NO FIDUCIARY OBLIGATION OR DUTY EXISTS, OR IS CREATED, BY VIRTUE OF YOUR RECEIPT OF ANY INFORMATION PROVIDED IN CONNECTION WITH THE INVESTING CLUB. NO SPECIFIC OUTCOME OR PROFIT IS GUARANTEED.
Jensen Huang, co-founder and CEO of Nvidia Corp., holds up the company’s AI accelerator chips for data centers as he speaks during the Nvidia AI Summit Japan in Tokyo on Nov. 13, 2024.
Akio Kon | Bloomberg | Getty Images
Investors craving a boring backdrop for Nvidia earnings this week can just forget it.
