Key Points
- Microsoft introduced an AI coding model that runs directly on personal computers, not cloud servers.
- The company also showed security technology to block AI agents from accessing data without permission.
- The push would shift workloads off Microsoft's costly Azure data centres onto customer-owned hardware.
The latest:
An AI coding model capable of running directly on personal computers was unveiled by Microsoft on Wednesday, alongside new security technology built to stop AI agents from reaching data without permission. At an event in San Francisco, the company also showed a high-performance Surface laptop called Ultra, powered by Nvidia’s RTX Spark chips. The announcements anchor a wider effort to make Windows a platform for AI agents.
Details:
- The model: The coding model runs on the device itself rather than in a remote data centre, according to Microsoft’s announcement in San Francisco. Microsoft did not detail the model’s size, pricing or release timing at the event.
- The security layer: Alongside the model, Microsoft presented a security technology designed to prevent AI agents from accessing data without permission — a direct response to the central risk of letting autonomous software act on a user’s own machine and files.
- The hardware: The company showcased a high-performance Surface laptop named Ultra, built around Nvidia’s RTX Spark chips. The device is positioned as the class of machine capable of carrying AI workloads locally rather than sending them to the cloud.
- The Windows strategy: Microsoft aims to turn Windows into a platform for AI agents that can handle tasks ranging from writing computer code to completing complex business projects, running on desktop and laptop computers rather than remote servers.
- The economics: The strategy reflects a bet that some tasks currently handled by Microsoft’s costly Azure cloud computing data centres can be shifted to high-performance Windows devices in businesses and homes, where customers bear the hardware costs rather than Microsoft.
- The market position: Microsoft holds a strong position in the Windows device market, giving it an installed base to target with on-device AI features — a channel its cloud-first rivals in generative AI do not control to the same degree.
- Nvidia’s role: The Surface Ultra’s reliance on Nvidia’s RTX Spark chips extends the chipmaker’s presence from data centre AI hardware into consumer and business laptops, tying Microsoft’s on-device ambitions to Nvidia silicon.
- What was not set: Microsoft did not announce availability dates, prices, or which Windows versions will carry the agent and security features. No figures were given for the share of Azure workloads the company expects to move onto local devices.
Background:
Generative AI has largely run in cloud data centres, where compute and energy costs sit with the provider. Running models locally shifts that cost to the device owner and reduces reliance on constant server capacity expansion.
Between the lines:
The pairing of an on-device coding model with a permission layer for AI agents points to the obstacle Microsoft must clear: agents that write code and complete business projects need access to local files, and enterprise buyers will not grant it without controls. The Azure cost argument explains the incentive to move those workloads onto hardware customers already pay for.
What’s next
Watch for pricing and availability for the Surface Ultra, the Windows releases carrying agent support, and any Microsoft guidance on Azure capital spending as workloads shift toward local devices.
