Nvidia’s strategic tie-up with fellow chipmaker Marvell Technology is yet another reason to stick with the AI giant’s sluggish stock. On Tuesday morning, the companies announced a partnership to integrate Marvell’s custom AI chips and networking products with Nvidia’s dominant AI computing ecosystem. In addition, Club name Nvidia is taking a $2 billion stake in Marvell, its latest investment into other technology companies using its cash windfall from the AI boom. For Nvidia investors, the deal is significant because it enhances Nvidia’s presence in the world of custom AI chips — widely considered the biggest threat to its crown as the king of AI computing. Nvidia’s largest customers, while still buying jaw-dropping amounts of compute from Nvidia, are investing in their own custom silicon with design help from the likes of Marvell and fellow Club holding Broadcom . Marvell’s customers include Amazon and Microsoft, while Alphabet and more recently Meta are known to be working with Broadcom. Shares of Nvidia jumped 5.6% Tuesday, on pace for their second-best day of the year. It is an encouraging reaction after the stock has languished for months despite a steady stream of good news — best exemplified by the frustratingly subdued market reaction to Nvidia’s impressive GTC conference in mid-March. To be sure, some of Nvidia’s recent weakness has stemmed from the broader market pullback on the Iran war. But no matter the cause, the end result was a stock that hasn’t been this cheap in over a decade. Now, with the Marvell arrangement, we have even more reason to believe Nvidia will continue delivering earnings growth in the years ahead. In effect, Nvidia has the opportunity to capture a larger share of the dollars spent on AI data centers if its other products are more compatible with custom silicon. While best known for its workhorse graphics processing units (GPUs), Nvidia has transformed into a one-stop-shop for AI computing, offering additional types of non-GPU processors and the networking technology that enables all the chips to talk to each other. Nvidia also has a vast software library that is core to its competitive standing, led by its CUDA (Compute Unified Device Architecture) platform. The deal with Marvell is focused on that networking side of its portfolio, specifically its NVLink technology. Last year, Nvidia announced the creation of NVLink Fusion to effectively open up its proprietary fabric to non-Nvidia processors — and that is what the companies will leverage in this strategic partnership. The news may have been foreshadowed in December, when we learned at Amazon’s cloud computing that Amazon would integrate NVLink Fusion with its own custom silicon solutions, including Tranium4 chip and its Graviton central processors (CPUs). In an interview Tuesday morning on CNBC, Nvidia CEO Jensen Huang explained the rationale behind the Marvell alliance. “All of the world’s data centers are going to be replaced with this new form of doing computing,” Jensen told Jim Cramer. “Of course, the vast majority of the data centers will be powered by Nvidia CUDA GPUs. However, for customers who like to also have specialized versions that they built themselves, we’re extending our architecture, starting with the networking architecture, basically the entire chassis of our Grace Blackwell and Vera Rubin systems. We’re going to extend that, through NVLink and connect it to Marvell. Together, we’ll be able to address the customers whether they would like to use all Nvidia gear or if they would like to augment our Nvidia gear with their specialized processors.” “Together, we are going to be able to address a much much larger [total addressable market, or TAM],” Jensen said. Grace Blackwell is Nvidia’s current generation computing platform, anchored by its Grace GPUs and Blackwell GPUs. It is on track to roll out its Vera CPUs and Rubin GPUs — under the Vera Rubin family of server racks — later this year. With NVLink Fusion, Jensen said a company that has designed its own custom AI chip can more easily link them to Nvidia products. He specifically called out its SpectrumX networking fabric, as well as Vera CPUs, its new Groq-infused inference chip designed for daily AI model usage, and its data-storage platform known as Bluefield. The Groq-infused chip and an updated version of Bluefield were among the bullish updates delivered at GTC. “All of this will be system architecture compatible. It’ll be ecosystem compatible… [and] makes it easier for them to interoperate and makes it fantastic for us as we expand out TAM,” Jensen said. Whether companies buy GPU compute platforms from Nvidia, or look to invest in custom silicon and partner with Marvell to do so, Nvidia stands to benefit. Also keep in mind: The $2 billion investment means Nvidia also benefits from any Marvell deals that don’t include any Nvidia components, given that it represents a roughly 2.5% ownership stake in the company. Nvidia has also used its massive cash pile to invest in startups like ChatGPT creator OpenAI, as well as suppliers such Lumentum and Coherent , both of which are in the optical technology space . Optics is a growing focus for Nvidia as the AI computing gets more intensive. Marvell has an optics business, and Tuesday’s announcement also called that out as an area of development. Corning is our bet on the growth of optical networking technology in the data center. Nvidia’s investment deals reinforce the notion that this is a strong company only getting stronger. That is precisely what has made the stock’s underperformance all the more bewildering and its valuation all the more shocking. Entering Tuesday’s session, Nvidia’s stock traded at roughly 20 times forward earnings, according to FactSet data — as cheap as it’s been in a decade. Roughly two years ago, the stock carried a forward price-to-earnings ratio of about 36. In August 2025, when shares began to stall out, its P/E was about 34. Over the past 10 years, the only two times we even got close to the current P/E of 20 was in 2018, as result of the Federal Reserve taper tantrum that crushed the market, and in the days following President Donald Trump’s reciprocal tariff announcement in April 2025. Crucially, Nvidia’s current valuation is before any uplift to earnings associated with the Marvell partnership are in the picture. So, that’s our starting point. Time will tell just how much the initiative accelerates Nvidia’s bottom-line growth this year, next year, and beyond. But the point is, the stock could end up proving cheaper than we thought Monday night. Bottom line At some point, something has to give. Even with Tuesday’s pop, the stock is still below where it was when the Iran war broke out. Sure, this could end up being a value trap, where earnings keep growing but nobody ever cares, resulting in persistent muted performance. However, we just don’t see that being the case. That leaves dueling outcomes where we one day find out the earnings estimates were way too high, or the price finally starts to reflect the fundamentals. Given what we heard at GTC and learned Tuesday, we struggle to see how estimates are too high — barring any dramatic intensification in Iran that plunges the economy into a recession and its customers are forced to reel in their AI spending. We think it’s more likely the estimates prove too low. When we step back and think about how fast the advancement and adoption of AI are proceeding, we simply struggle to see how Nvidia doesn’t start to work again once overall investor sentiment improves. Perhaps Tuesday’s rally ends up being the start of that (though we’re not exactly ready to ring the all-clear bell.) It is difficult to pound the table on a name that, up until Tuesday, did not seem capable of going higher, no matter how good the news got. But we are definitely screaming from the hilltops that Nvidia has given you one more reason to own the stock for the long term and not try to trade the name on every headline. (Jim Cramer’s Charitable Trust is long NVDA, AVGO, AMZN, MSFT, META, AVGO, GLW and GOOGL. 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.
Here is why Nvidia's partnership and investment in Marvell is such a big deal
view original post