This week in Microsoft AI

Microsoft signaled sustained Copilot demand while tightening practical controls and data tooling.

A reported Microsoft earnings datapoint put paid Copilot enterprise users above 20 million, alongside another large year of AI infrastructure spend. In product-level updates, Microsoft documented how to reduce Copilot surfaces in Edge and highlighted new AI functions embedded directly in Fabric T‑SQL.

Reported: Copilot passes 20M paid enterprise users

A Windows Central article cited Microsoft’s Q3 FY26 earnings discussion as stating that Copilot has more than 20 million paying enterprise customers and that Microsoft plans up to $146B of AI infrastructure investment in 2026.

Use the adoption figure as a benchmark when sizing internal demand, licensing, and support capacity for Copilot rollouts in large tenants.
Treat the infrastructure spend as a signal to plan multi-year Copilot dependencies (latency, availability, service limits) into business-critical workflows rather than pilots.
Ask Microsoft and partners for EU-region service assurances and capacity guidance if you expect heavier Copilot usage in regulated Czech environments.

Source — Windows Central CopilotAzure

Edge documents how to reduce Copilot surfaces

Microsoft Q&A described how users can reduce or disable AI-related content and Copilot modules on the Microsoft Edge new tab page using layout settings and the edge://settings/ai toggle for the Copilot new tab page.

Document the settings as part of your Copilot governance pack for business units that must minimize AI prompts on shared or kiosk-style devices.
Use the toggle and layout controls to stage Copilot exposure by group, which helps change management during an M365 standardization program.
Include the exact setting path in your helpdesk knowledge base to cut tickets after Edge updates that change default new-tab content.

Source — Microsoft Q&A EdgeGovernance

Fabric adds AI functions directly into T‑SQL

Microsoft’s Fabric “What’s new” page stated that Fabric Data Warehouse can run AI tasks directly in T‑SQL, including text categorization, sentiment analysis, and structured data extraction from text.

Enable data teams to prototype classification and sentiment features inside existing SQL workflows without introducing a separate ML service layer.
Reassess architecture for text-heavy analytics (support tickets, emails, claims notes) because in-SQL AI can reduce pipeline steps and operational overhead.
Update data governance reviews to cover how AI-in-SQL features handle prompts, outputs, and logging in the warehouse environment.

Source — Microsoft Learn (Fabric) Microsoft FabricData