CES 2025 Silicon: Qualcomm Disrupts the PC Market. Again.
There was a lot of silicon and computing news at CES 2025:
Intel showed off new Core Ultra chips in a dizzying array of wattage, performance and efficiency core, and NPU configurations. Perhaps more importantly, Intel showed off working systems using processors from its 18A process due out much later this year/into 2026. These “Panther Lake” chips should be more competitive and more profitable for Intel to make than the Core Ultra chips it is partly sourcing from TSMC.
AMD announced new confusingly named Ryzen processors aimed at bringing 50 TOPS of NPU performance slightly down-market from its AI 300. It also moved upmarket, announcing Strix Halo processors with huge numbers of cores to compete with Apple’s M4 Pro and M4 Max processors. I am looking forward to testing out HP’s ZBook Ultra G1a with AMD’s new Ryzen AI Max PRO processor, which HP dubbed, “the world’s most powerful 14-inch mobile workstation.”
NVIDIA announced its 50 series GPUs along with Project DIGITS, a miniature $3,000 Linux desktop AI computer powered by a tiny version of its datacenter silicon. The use case for the PC GPUs is primarily gaming with additional frame insertion when playing games at higher resolutions and faster refresh rates. NVIDIA will have founder’s edition cards available in the coming weeks, and within the next three months we should start seeing laptops with the mobile versions from Lenovo, HP, Dell, ASUS, and others.
NVIDIA’s most intriguing announcement* at its CES keynote was Project DIGITS. DIGITS is a response to AI researchers who want to build and test larger models on their desks without having to work entirely in the cloud. DIGITS has a GB10 Grace Blackwell “Superchip” that delivers a petaflop of AI performance. NVIDIA’s AI software stack is preinstalled along with 128GB of memory allowing prototyping and AI model inference of up to 200B parameters locally. The most intriguing detail of Project DIGITS – beyond simply how small the unit is – isn’t the GPU/NPU but the CPU. The CPU for Project DIGITS is an Nvidia design, but MediaTek played a public, though unspecified role, in bringing it to market. I talked to both NVIDIA and MediaTek at CES and couldn’t get either to divulge specifics. Is MediaTek providing NVIDIA with its TSMC allocation to bring DIGITS to market without impacting its Blackwell production for hyperscalers? Is this the first step in NVIDIA and MediaTek jointly entering the Windows PC silicon market? We’ll have to see. In any case, the partnership between the two companies is running deeper than just automotive.
MediaTek also quietly had another win at the show behind the scenes. Every large TV vendor at CES announced new mini LED HDTVs with increasing numbers of dimming zones, often with additional control over the RGB in each LED. Managing all of these elements, at higher refresh rates for gaming, along with more AI in the user interface has required more powerful silicon than vendors have used in the past. Each vendor highlighted their branded AI processors that improve picture quality, but aside from Sony, which credits MediaTek publicly, and Samsung, which makes its own processors, nearly all of these processors are actually just MediaTek’s chips in disguise. While MediaTek would surely like some marketing attribution, the fact is that its customers are finally moving up the value chain and buying more expensive cards.
Despite all this, Qualcomm may emerge as the ultimate “winner” of CES 2025 Silicon Wars with its new Snapdragon X chips for Windows laptops. Today’s Qualcomm is diversified across IP, phones, automotive, and IoT, but that wasn’t always the case. One of the first phone-adjacent markets that Qualcomm tried to enter was computing, starting with the Snapdragon 835 for Windows laptops and convertibles in 2017, though not with much success. To help break through, Qualcomm acquired NUVIA in 2021 for its engineering talent. In 2024, Qualcomm started shipping the Snapdragon X SoC with an extremely powerful and efficient NUVIA-derived Oryon CPU architecture, Hexagon NPU with 45 TOPS, and a more serious commitment from Microsoft to make the software work this time.
Qualcomm launched Snapdragon in stages with the highest-performing tier first. This made excellent overall marketing sense: by starting at the top, Qualcomm ensured that consumers and the tech press could see what the platform could do and restore a premium brand image to Qualcomm-powered PCs. However, that has meant that the ensuing mid-tier Snapdragon X Plus and entry level Snapdragon X didn’t get the same push as they hit the market three and six months later.
That’s a shame because, as much as PC OEMs (and Microsoft and Qualcomm) would like to cut into Apple MacBook sales, relatively few laptops are purchased at $1,000+ price points. The Snapdragon X Plus was launched at IFA which brought Qualcomm into the $700 - $1,000 system range. While there were some Snapdragon X Plus design wins at IFA, we also saw the first desktops based on the chipset from Lenovo at CES. With the base Snapdragon X, Qualcomm’s lineup for this generation is complete, and brings Snapdragon X all the way down to $600 systems.
It may be the baby of the family, but the Snapdragon X is still a 4nm chip with all the benefits of the Oryon CPU and Hexagon NPU architecture in terms of performance, on-device AI, and extensive battery life. You’ll get a little less performance in the CPU and GPU – without any drop in NPU capabilities – for a lower price. That dip in performance shouldn’t matter for mainstream laptops for home, work, and school. In fact, there are an increasing number of use cases where the faster NPU will actually make up for the slower CPU and GPU. Qualcomm had a private demo area at CES where it showed off how on-device AI can be used to edit videos, mix audio, and apply color filters to live photography (I volunteered for the headshot). Those specific scenarios may not be relevant to everyone buying a $600 laptop, but they are no longer entirely niche, either. With social media, everyone is a creator now. Even for those who will never crack open an editing app, the NPU will handle background blur and noise cancellation during video calls without slowing anything else down or crashing your battery. This should endear Qualcomm to budget-conscious IT departments as well as consumers. Once Microsoft un-recalls Recall, you’ll also be able to run Recall in the background as well because even a $600 laptop with a Snapdragon X qualifies as a Microsoft Copilot+ PC.
Even with a Snapdragon X, a $600 laptop is still going to need to cut corners to hit that price point; the chassis will be plastic, not aluminum or magnesium, and the displays in particular are going to be less than ideal. However, the thermal envelope and power efficiency of the Snapdragon should allow designers to make thinner and lighter designs than today’s inexpensive laptops.
This is bad news for Intel, which profitably serves this market segment today using chips manufactured in-house on its older process nodes. Intel cannot profitably compete with Qualcomm at these price points using its own efficient AI PC chips like Lunar Lake because those are partly manufactured by TSMC, not in Intel’s own fabs. Until Intel has its new fabs fully optimized (hopefully by 2026 or 2027), Qualcomm will likely be the only way to get extremely long battery life and 45 TOPS of performance at these price points unless AMD wants to take a serious hit to its own margins.
To discuss the implications of this report on your business, product, or investment strategies, contact Techsponential at avi@techsponential.com.
*NVIDIA CEO Jensen Huang made a lot of announcements and pronouncements and proclamations at his CES keynote. This report is focused on consumer and end user silicon, but NVIDIA’s Cosmos AI framework for the real world may be the most far reaching. If the world is subjugated by sentient AI robots, we might have Cosmos to blame. Of course, if a robot doctor saves your life or fills vacant eldercare positions, we might have Cosmos to thank.