On Thursday, Intel (INTC 4.53%) and Nvidia (NVDA 0.27%) announced a shocking deal that will pair the two companies' technologies in PC and data center central processing units (CPUs). This news looks problematic for Advanced Micro Devices (AMD -1.05%). AMD has made an extraordinary comeback over the past decade in both markets, with a unit market share across PCs and servers of around 24% in the second quarter, up from 11% in 2016. Those gains could come under pressure once the Intel-Nvidia partnership bears fruit.
In the PC market, Intel and Nvidia will develop systems-on-a-chip (SOCs) that integrate Intel's CPU cores with an Nvidia RTX graphics processing unit (GPU) chiplet. In the data center market, custom Intel CPUs with Nvidia's NVLink technology will be used directly by Nvidia, and the GPU giant will also offer the chips to third-party customers. NVLink is Nvidia's proprietary high-speed connection technology that's a more performant alternative to the standard PCI Express interface.
In the PC market, in particular, AMD will have a fight on its hands as it looks to preserve its market share.
Image source: Getty Images.
AI PCs have been slow to take off
The PC CPUs that come out of this deal are targeting "a wide range of PCs that demand integration of world-class CPUs and GPUs," according to Intel's press release. Gaming laptops and gaming handhelds would be the natural form factors for the new CPUs, given the gaming performance that Nvidia's GPU technology will enable. But Nvidia's GPUs will also provide hefty artificial intelligence (AI) computing horsepower.
As a category, AI PCs aren't particularly impressive right now. PCs running Intel, AMD, and Qualcomm CPUs with enough AI computing capacity to run some AI workloads locally get lumped into this category, but they're just not capable enough for AI to be a strong selling point. Microsoft's Copilot+ initiative, which puts that label on certain AI PCs, was a disappointment when it debuted last year. The AI features included brought little to the table, and anything useful still required calling out to a cloud service.
For AI PCs to truly take off, they need to be powerful enough to run capable AI models locally that can perform useful tasks for the user. Once a PC can handle running a competent AI writing assistant or a useful AI coding assistant, the appeal of the AI PC expands dramatically.
A potential AI powerhouse
The CPUs that will emerge from the Intel-Nvidia deal appear to be squarely aimed at the AI PC opportunity. Nvidia's GPU technology has become the industry standard for AI computing, and the GPU chiplet included in these new CPUs could provide enough AI computing capacity to handle more complex AI workloads.
Some of AMD's Ryzen PC CPUs include both a neural processing unit and a GPU to accelerate AI workloads. The high-end Ryzen AI Max+ 395, for example, offers overall AI performance of 126 TOPs, or trillions of operations per second. That's nothing compared to dedicated GPUs, and it's certainly not enough to run truly capable AI models. Nvidia's midrange RTX 5060 laptop GPU, for example, offers 440 TOPS, while its high-end RTX 5090 laptop GPU clocks in at 1,824 TOPS.
We don't know exactly what the Intel-Nvidia CPUs will look like, but they could push the envelope when it comes to AI computing capacity, thanks to Nvidia's GPUs. By putting an Nvidia GPU on the same chip as the CPU, rather than pairing the CPU with a discrete GPU, latency can be reduced, power efficiency can be improved, and costs can be brought down.
AMD makes competitive products, but the PC CPU market is a whole new ball game when Intel CPUs integrated with Nvidia GPUs start shipping. AMD's presence in gaming laptops and gaming handhelds will likely come under pressure, and the larger battle for AI PC market will be much tougher with Intel and Nvidia teaming up.
