Microsoft is reportedly organizing Windows PCs with AI capabilities into four tiers, distinguished mainly by memory. The map, reported by Wccftech from a leaked internal roadmap, runs from budget laptops with 8 GB of RAM up to machines with 128 GB or more built to run large models directly on the device.
At the bottom is Tier 0, with 8 GB of memory. According to the roadmap, these PCs handle the most basic AI tasks on the device and use chips such as Intel Core Series 3 or 4, AMD Ryzen 100 and 200 and Snapdragon X and X Plus. Most of these machines are already on sale at around $500. The limit is clear: 8 GB is enough for browser tabs, email, Office files and video calls, but becomes a bottleneck as the load grows.
Image: freepik.com — https://www.freepik.com/premium-photo/artificial-intelligence-hardware-concept-3d-rendering-generative-ai_41999509.htm
Tier 1 covers the mainstream segment, from 16 to 64 GB. Microsoft defines it as Copilot+ compliant, meaning capable of local answers, delegated tasks and long-running jobs with less reliance on the cloud. Processors cited include Intel Core Ultra Series 3 "Panther Lake", AMD Ryzen AI 400 and upcoming next-generation chips. The premium end includes Nvidia RTX Spark and AMD Ryzen AI Max+ 400.
The two top tiers target professionals and developers. Tier 2 offers at least 128 GB of memory and the ability to build and serve models locally. According to the roadmap, AMD Ryzen AI Max+ 400 and Nvidia RTX Spark are available today, while Intel Core Ultra Next and Snapdragon Next are future offerings. Tier 3 is the workstation class, which the press calls personal supercomputers: machines built with server-grade chips such as Intel Crescent Island GPUs, which according to the roadmap scale from 160 GB up to 512 GB of memory per GPU.
The structure is striking because memory, rather than raw compute, becomes the classification criterion. That detail should be read with caution. Microsoft has not published the table in this form; the roadmap is a document reported by Wccftech and other outlets. What the company does confirm publicly is its hybrid intelligence strategy, in which part of the AI workload runs on the PC and part in the cloud. In the Copilot+ context, the NPU, the chip dedicated to AI, is the component expected to support local processing, as outlined in the announcement on Windows 11 hybrid intelligence features.
Image: dreamstime.com — https://www.dreamstime.com/glowing-brain-circuit-microchip-computer-motherboard-big-data-processing-ai-trading-machine-learning-ai-technology-image290717268
The table also explains why makers are pushing high-memory configurations. AMD is aiming at PCs with at least 64 GB for heavier workloads, as shown by the Gorgon Halo line with 192 GB of RAM, while Intel is preparing 480 GB GPUs for data center inference, as detailed in the D-series report. Tier 3 seems to gather machines the market is already building.
What remains open is when the tiers would start and the timing of chips not yet presented. For now the roadmap sketches the overall framework, not prices or release dates for individual machines.
No comments yet. Be the first!