Microsoft Copilot is coming to local PCs, but needs beefy chips

Every time you ask Copilot to rewrite a sentence on Windows today, your prompt takes a round trip to Microsoft's cloud servers. That is about to change, but your current laptop probably won't survive the upgrade.

Intel revealed to Tom's Hardware that Microsoft's Copilot will soon run key elements locally on personal computers. To make that happen, Microsoft is setting a strict requirement for next-gen AI PCs: an integrated NPU capable of at least 40 TOPS (trillion operations per second).

Every current chip just got disqualified

That 40 TOPS requirement instantly renders every consumer processor currently on store shelves obsolete for next-gen local AI.

To put that number in perspective, Apple's M3 chips only top out at 18 TOPS. AMD's Ryzen 8040 laptop processors hit 16 TOPS, while its older 7040 reaches 10 TOPS. Intel's current Meteor Lake lineup also sits at just 10 TOPS.

Surprisingly, Qualcomm is positioned to cross the finish line first. Its upcoming Snapdragon X Elite promises 45 TOPS of AI performance, making it the only announced processor ready for Microsoft's local standard.

Intel is asking users to wait for its Lunar Lake chips, slated for 2025, which promise to triple current NPU speeds. In the meantime, Intel introduced 300 AI features optimized for its OpenVino toolkit and launched an AI PC dev kit built around an Asus NUC Pro with Meteor Lake silicon.

Previous leaks suggested next-gen AI PCs would require both 40 TOPS and a baseline of 16GB of RAM. Intel's comments now make that baseline official.

Windows finally gives the NPU real work

Until now, Windows has barely touched the NPUs inside existing PCs. Outside of webcam parlor tricks like background blur and eye tracking on Surface devices, dedicated AI silicon has mostly sat idle.

Ars Technica noted that both macOS and ChromeOS have spent years utilizing NPUs for real productivity tasks, including system-wide OCR, live audio transcription, and offline language translation.

Running Copilot locally fixes the noticeable lag that plagues quick desktop commands, while also offering major privacy benefits by keeping user queries off remote servers.

Why it matters

For everyday users, this is basically an eviction notice for your current PC. If you want full-fledged, zero-latency AI integrated into your desktop workflow, you simply cannot buy a chip today that qualifies. It is worth holding off on any major Windows PC purchase until the next wave of silicon arrives.

For the broader industry, Qualcomm has a massive opportunity to shake up the PC hierarchy. By beating both Intel and AMD to the 40 TOPS threshold with the Snapdragon X Elite, Qualcomm has its best shot in years at taking real market share from the x86 duopoly.

We think Microsoft's decision to set a 40 TOPS floor is smart, but risky. It provides a real technical definition for the nebulous "AI PC" buzzword and stops OEMs from slapping stickers on underpowered hardware. The downside is clear: with compatible silicon largely absent until late this year and 2025, user frustration with cloud latency could easily stall Windows AI momentum before it really starts.