Microsoft's AI PC roadmap shows where Ryzen AI 500 and Intel Nova Lake fit into Windows

Microsoft's AI PC roadmap presents a clear position for Ryzen AI 500 and Intel Nova Lake within the Windows ecosystem. The company has outlined four AI tiers for Windows PCs based on memory capacity, ranging from 8GB to beyond 128GB. The Ryzen AI 500 and Core Ultra Series 4 both occupy the mainstream Copilot+ class, while Ryzen AI 400 and Core Ultra Series 3 and 4 are positioned in the lower tiers.

Microsoft's slide, titled "Windows is the home for hybrid intelligence," categorizes PCs according to their memory scale. The chip names appear as footnotes, but the information they provide is crucial. The slide divides the Copilot+ tier into three bands: quick local questions at 16GB, delegated tasks at 32GB, and long-running jobs at 64GB. Core Ultra Series 3 and 4, as well as Ryzen AI 400 and 500, are situated in the first two bands.

The timing of the chip release is suggested by a small "CY26-27" label next to the chip list. This aligns with leaks indicating that Core Ultra Series 4 will be Intel's Nova Lake, and Ryzen AI 500 will be AMD's Zen 6-based Medusa Point, both expected to launch in 2027. Upgrading the memory capacity changes the chip lineup significantly.

At 64GB, mainstream parts are replaced by Ryzen AI Max+ 400, Nvidia's RTX Spark, and an unnamed Core Ultra Next from Intel. These components are also found in the 128GB Builder PC tier, designed for local model building and serving.

Beyond the 128GB tier lies the Supercomputer tier, intended for high-performance machines running frontier models, demonstrating Microsoft's ambitious vision for Windows. An 8GB laptop, however, remains in the basic "Windows PC" tier, limited to local context and fundamental actions, regardless of the chip it uses. This includes Microsoft's 8GB Surface models from June.

Microsoft and Qualcomm announced in September that they had ceased referring to new devices as Copilot+ PCs. However, the label persists in the slide, underscoring the slide's alignment with Microsoft's hybrid intelligence strategy announced last month with the RTX Spark-powered Surface Laptop Ultra. The strategy implies that Copilot will eventually process local files, execute actions, and utilize on-device models.

The slide emphasizes that a powerful Neural Processing Unit (NPU) alone will not unlock Copilot+ features without at least 16GB of memory support. Moreover, the increasing cost of memory this year highlights the significance of extra memory capacity in AI applications.