Microsoft and Nvidia Are Building AI Agent PCs, Because Copilot Wasn't Embarrassing Enough
Microsoft and Nvidia are reportedly joining forces on a new class of Windows PCs that use Nvidia silicon as the primary processor. According to Axios, the first devices will be shown off at Computex in Taiwan and Microsoft's Build conference in San Francisco next week. Both Microsoft's Surface line and Dell are expected to have hardware ready to show.
More interesting than the chips is what Microsoft apparently wants to run on them. The company is developing software to let AI agents execute tasks locally on Windows, and it has been quietly building around a framework called OpenClaw since the start of the year. A dedicated team was assembled under developer Omar Shahine, who recently announced he's bringing OpenClaw and personal agents into Microsoft 365. OpenClaw's founder Peter Steinberger, now at OpenAI, has a session scheduled at Build, which suggests Microsoft is planning to put the framework front and centre.
This is Microsoft's second crack at the AI PC concept. The first, branded Copilot+ PC, was essentially a marketing exercise that slapped AI branding on mid-range laptops and pushed Copilot into the default experience whether users wanted it or not. It did not set the world on fire. The new approach at least has a more concrete proposition: agents that do actual work locally, rather than a chatbot sidebar nobody asked for.
That said, OpenClaw has known security and reliability concerns that haven't gone away just because the processing happens on-device. Local execution solves the data privacy problem to some degree, but it doesn't automatically make the agents trustworthy or robust. Whether Microsoft has addressed any of that before shipping will be worth watching closely at Build.