Microsoft AI Updates: October 13, 2026

1. Microsoft Details Maia 300 Inference Chip, Plans Multi-Gigawatt Azure Maia Deployments From 2027

Microsoft. Scott Guthrie announced Maia 300, Microsoft’s third-generation inference accelerator, with compute chiplets on TSMC’s N2P process, eight HBM4 stacks delivering 18.0 TB/s of peak bandwidth, a 950 W TDP, and support for low-precision formats including enhanced MXFP4. A rack holds 96 accelerators in a 167 kW envelope with 27.6 TB of HBM, and Microsoft projects 40% more tokens per megawatt than leading GPU-based systems and up to 7x the rack-scale throughput of Maia 200, figures that are not yet measured results. Maia 300 begins production deployments in 2027 and is slated to become the primary inference system for Microsoft products, while the current Maia 200 already serves Copilot, OpenAI models, MAI models, and open-source model families, with GitHub Copilot and Microsoft Foundry moving onto it by the end of 2026. The post mentions no external customer access or SDK, so the chip matters to developers mainly through the cost and latency of Copilot and Foundry model serving. Source

2. GitHub Copilot Adds Admin Controls to Attribute Agent Telemetry to Users

GitHub Copilot. Enterprise administrators can now use managed telemetry settings to tie Copilot agent activity to individual users without capturing prompts or responses. The telemetry.capture.identity setting adds three OpenTelemetry attributes (Copilot account name, local OS account, and machine name), and five content classes (prompts, responses, tool arguments, tool output, and policy detail) can be enabled independently, all off by default. Managed values override local settings, SDK configuration, and environment variables, and the controls apply to Copilot CLI, the GitHub Copilot app, and VS Code sessions using Agent Host. GitHub warns that content capture can include source code, file contents, and commands, so teams should enable only what an audit requires. Source