Nvidia has taken a new step into its investment case, locating it far from its data centers. It’s on your desk in your office or home. At the influential Computex conference in Taiwan, CEO Jensen Huang focused the first half of his keynote on the greatness of Nvidia’s Vera computing platform for data center and agent AI workloads. It was familiar territory. Its dominance of the data center AI chip market has made it the world’s most valuable company. Huang then pivoted and introduced an entirely new product line for Nvidia, pitting it against Intel and AMD in the Windows personal computer market. Huang announced a suite of new laptops, desktops, and ruggedized workstations powered by a new integrated chip called RTX Spark. Nvidia stock soared more than 4% on Monday, while Intel and AMD fell. Co-designed with Taiwan’s MediaTek, RTX Spark is considered a system-on-chip (SoC). This essentially means that a set of computing functions will be integrated onto a single silicon rather than stitching together separate chips. For RTX Spark, Nvidia has designed the central processing unit (CPU), graphics processing unit (GPU), and neural processing unit (NPU) required for powerful on-device AI computing all integrated into a single SoC package. As a result, for the first time, Nvidia will fully dictate the performance, power requirements, and AI capabilities of these PCs. The GPU is based on the Blackwell family architecture. The CPU is based on the power-efficient Arm instruction set, another win for the Arm club name. Granted, this isn’t a fully integrated computer like what Apple offers. Apple designs every aspect of its devices, from the physical appearance to the operating system software and even the silicon. However, it’s much more vertically integrated than what Nvidia has traditionally offered in the consumer PC space. Nvidia has historically only made graphics cards for PCs, making them extremely popular among PC gamers. It was Nvidia’s original dominant market before expanding into data centers. Although no official release date has been announced, we expect these Nvidia-based PCs to start hitting stores in the fall, in time for the holiday shopping season. Early laptop and desktop PC models are manufactured by ASUS, Dell, HP, Lenovo, club namesake Microsoft, and MSI. NVDA 1Y Mountain Nvidia’s stock price performance over the past 12 months. When purchasing a PC, there are often three options for graphics cards or GPUs: dedicated, integrated, or SoC. SoC configurations are the newest configurations and have become increasingly popular in recent years. Clearly, Nvidia has indicated that it expects it to become even more popular. Here’s why: Integration: The graphics card is integrated with the CPU on the same silicon base, called the board, and shares power and memory (RAM). Slow memory, either due to lack of physical RAM or memory usage by the CPU, reduces performance (latency). RAM is also optimized for the CPU rather than the GPU. Dedicated: The graphics card is separate from the CPU and has its own dedicated memory (video RAM or VRAM) and power supply. The memory is designed specifically for graphics data, and because it is dedicated to graphics data, it has lower latency. However, the problem with dedicated memory is that the CPU and GPU cannot communicate efficiently. The GPU has all the VRAM it needs, but it doesn’t really matter until the information from the CPU’s RAM is copied into the GPU’s VRAM. This copy/paste process becomes the bottleneck. System-on-a-chip: All components are on the same “package” and share a unified memory that can be accessed simultaneously by the CPU and GPU. This is similar to the unified option, but with much higher bandwidth, the CPU can treat the unified memory component like RAM, the GPU can treat it like VRAM, and both can utilize the memory component at the same time without one slowing down the other. Additionally, since the memory is shared, the copy-and-paste bottleneck associated with dedicated options is also eliminated. The takeaway here is that SoC designs are the best choice for AI-oriented PCs. However, products for Windows PCs have traditionally been available only from Intel, AMD, and to a lesser extent Qualcomm. If you wanted a Windows-based PC with an Nvidia GPU, you had to get it as a dedicated option for an AMD or Intel CPU. Additionally, AMD and Intel are based on the x86 architecture, which is essentially the language that software and CPUs use to communicate. If I wanted an Arm-based PC, I was looking at either a Mac (which would mean abandoning Windows and committing to a completely different ecosystem) or a Qualcomm Snapdragon-based PC. The problem with the Snapdragon option is that it’s built on an Arm-based architecture in a world where Windows was designed for x86, so there are some compatibility issues. But Nvidia claims to have fully addressed the compatibility issue, making this the first Arm-based Windows PC that users can buy without worrying about compatibility issues with programs they rely on. According to Huang, Nvidia’s Arm-based PCs can not only run 100% of Windows applications, but they can also run everything Nvidia has ever done. Running Windows is key to adoption, but being able to run the entire Nvidia stack locally? That’s exciting. Think about it. If you’re an AI engineer who needs to develop and test language models or new agents at scale, but want to work on the Windows operating system, you can now work locally. This allows you to run and test your application locally before deploying it to the cloud, saving you significant cloud computing rental costs while building your application. This leads to another major theme. We are now firmly entering the era of “edge computing.” Cloud computing is the sending of data over the Internet to large data centers for processing. Edge computing means performing that computing at the “edge,” where the users are local. By introducing a computer capable of running everything Nvidia has ever created into the PC form factor, Nvidia is essentially putting a supercomputer in your home that can act as the brains for all your other smart devices. While the most intensive tasks may still need to be sent to the cloud for datacenter-level computation, most tasks can definitely be executed locally, reducing latency and cost while improving reliability, execution time, and security. Nvidia’s decision to aggressively compete for market share in a consumer PC market dominated by Intel, AMD, and Apple reinforces our belief that this is a stock to own for the long term. Nvidia’s data center business will continue to be its biggest opportunity, given that annual AI infrastructure spending is expected to reach trillions of dollars. But the PC market is not a niche, with industry firm IDC predicting sales of $274 billion in 2026. As NVIDIA maintains its momentum in the AI data center market and moves into the stronghold of Intel and AMD, we’re having a hard time seeing how the price-to-earnings ratios of these three stocks will converge over time. Intel’s forward P/E over the next 12 months is 91x, AMD’s 52x, and Nvidia’s 21x. Of course, these three chipmakers are not carbon copies of each other. Intel, for example, has a fledgling third-party manufacturing business that has investors excited. However, we do not believe current valuations are reasonable as the companies are now competing more directly than ever before. This is even truer when you consider that the other two companies haven’t taken up the level of software moats or vertically integrated hardware stacks that Nvidia has. I’m not explicitly saying that Intel and AMD are overvalued. Try calling someone else. Our point is that NVIDIA has been, and continues to be, abnormally undervalued when compared to these two companies, as well as the S&P 500 as a whole, which is considered a proxy for “average price.” Nvidia and the S&P 500 currently have roughly the same multiple. With its sights set on new markets, Nvidia is clearly no ordinary company or ordinary stock. (Jim Cramer’s charitable trusts are long NVDA, ARM, AAPL, and MSFT. See here for a complete list of stocks.) As a subscriber to Jim Cramer’s CNBC Investment Club, you will receive trade alerts before Jim makes a trade. After Jim sends a trade alert, he waits 45 minutes before buying or selling stocks in his charitable trust’s portfolio. If Jim talks about a stock on CNBC TV, he will issue a trade alert and then wait 72 hours before executing the trade. The above investment club information is subject to our Terms of Use and Privacy Policy, along with our disclaimer. No fiduciary duties or obligations exist or arise from your receipt of information provided in connection with the Investment Club. No specific results or benefits are guaranteed.
