The first mobile AI PCs became available from Acer, Asus, Dell, HP, Lenovo, Samsung, and other PC OEMs in June, powered by Microsoft’s Copilot+ and Qualcomm’s Snapdragon X Elite SoCs. Another wave of mobile PCs powered by AMD’s AI 300 and Intel’s Core Ultra Series 2 SoCs are rolling out now, with Copilot+ support from Microsoft scheduled for November. And sometime early next year, Tirias Research expects Microsoft support for both desktop and mobile PCs using discrete graphics processing units (GPUs). But it doesn’t end there.
What is a Copilot+ PC?
According to Microsoft’s original definition, a Copilot+ PC is configured with at least 16GB of memory, at least a 256GB solid-state drive, and a mobile SoC (System-on-Chip) with at least 40 teraoperations per second (TOPS) of AI performance from an integrated neural processing unit (NPU). While the memory and storage requirements are simply a system configuration option, the SoC requirement is the key requirement because not only must the SoC have the required performance for on-device AI processing, but Microsoft must also provide software support for the specific NPU. Tirias research views these as “Productivity AI PCs,” because of the lack of support for a discrete GPU as an AI processing engine. However, this will change as support for discrete GPUs is added.
Adding Microsoft support for discrete GPUs will not only allow new PCs with discrete GPUs to be classified as Copilot+ PCs, but it will also enable hundreds of millions of mobile and desktop PCs that already have discrete GPUs and are already running various AI workloads like video upscaling, to be classified as Copilot+ PCs. It also creates a new category of AI PCs that Tirias Research calls “Enthusiast AI PCs.” Just as they enable better graphics and processing performance, leveraging discrete GPUs will push the limits of AI workloads by orders of magnitude more and open the door to even more powerful traditional and generative AI applications.
Going beyond Microsoft
On the software side, third-party developers are emerging with AI-enabled applications. As with smartphones, many of the PC OEMs, which are considered first-party developers, are offering and developing AI applications that can run on devices with the power of the Copilot+ PCs. However, again similar to the smartphone ecosystem, independent software developers (ISVs) are modifying current applications and developing new ones to leverage the power of the Copilot+ PCs. If history is any indication, the software community will exceed the AI resources available and drive a need for even more performance in both productivity and enthusiast-class AI PCs.
A glimpse at the future
While some view these new classes of intelligent devices with concern or skepticism, on-device AI is here today in PCs. Although on-device applications are limited today, they will grow rapidly, enabling new usage models for PCs, such as becoming digital agents or what some are starting to call “Agentic AI” – having the ability to operate without human intervention but with knowledge of the user and the environment, process new information, and interact with a user in a variety of ways. A change such as this in the PC usage model is long overdue, and it will revolutionize the PC and lead to a wave of other intelligent devices.
Note that not all information or intelligence is likely to reside on the device, so cloud connectivity is still essential. While this can be achieved over Wi-Fi, Tirias Research believes that cellular connectivity will also be a critical requirement for AI-enabledmobile devices, including PCs. While only a handful of higher-end Copilot+ PCs offer 5G connectivity today, it is another area where AI PCs are likely to evolve.
