Fabric IQ is on by default in Copilot. Own the semantic layer before clients trust every answer.
Consultiply’s independent MSP read of Microsoft’s FabCon Barcelona announcements: Fabric IQ generally available in Copilot Chat and Cowork, grounded on Power BI semantics, with zero extra AI token cost (vendor claim).
Published September 30, 2026 · ~8 min read
What shipped
On September 28, 2026, at FabCon and SQLCon in Barcelona, Microsoft announced that Fabric IQ is generally available in Microsoft Copilot Chat and Cowork. Per Microsoft’s official blog and Azure Blog, Fabric IQ brings governed business context into Copilot by grounding answers on Power BI semantic models, metrics, relationships, and business definitions organizations already trust.
For Fabric and Power BI customers, Microsoft says this capability is on by default in Copilot Chat and Cowork, and that it preserves existing access controls and governance. Microsoft also says customers can tap Fabric insights in Copilot with no additional AI token costs (vendor claim). Integration with the new Copilot Code experience is listed as coming through the Frontier program.
Microsoft describes Fabric IQ as a shared intelligence layer that combines unified data from OneLake, trusted metrics from Power BI semantic models, and operational context from ontologies and real-time intelligence. The New Stack reported Microsoft Fabric CTO Amir Netz comparing agents to waking up with no memory of prior context: each session needs to be told where it is and who it works for. Fabric IQ is Microsoft’s answer for the “state of the business” piece of that context stack, alongside other Microsoft IQ surfaces for work content, documents, and the public web.
Related FabCon items that sit next to this story, still mostly preview or roadmap unless noted: IQ Sharing in preview for sharing governed data and context across tenants; ontology tooling that builds from semantic models and documents; Fabric data engineering agents in preview; observability in Fabric in preview; and agentic app creation in Power BI Desktop coming in preview in the coming weeks (vendor claims).
Verified scale (vendor claims)
Figures below appear in Microsoft’s Azure Blog, CIO’s FabCon coverage, and/or The New Stack’s reporting. None were measured by Consultiply.
Fabric customers
more than 40,000 worldwide
Active Fabric users
2.6 million (CIO)
Power BI semantic models already created
22 million (The New Stack, citing Netz)
OneLake storage under management growth
300% year over year (The New Stack, citing Netz)
FabCon Barcelona
nearly 5,000 attendees from 71 countries (Azure Blog / CIO)
What this means for MSP operators
Agents forget everything when they wake up. If Copilot Chat and Cowork now answer from Power BI semantics by default, a wrong semantic model becomes a wrong answer at Microsoft 365 scale.
Own the semantic catalog.
Inventory which Power BI / Fabric models clients already treat as system of record. If the model is wrong, every grounded Copilot answer is wrong.
Confirm who can ask.
Default on means business users may start querying metrics in Chat and Cowork without a fresh project kickoff. Map roles, workspaces, and sensitivity labels before demos become production answers.
Price model hygiene as a service line.
Microsoft says there is no extra AI token surcharge for this grounding. Your margin is model ownership, change control, eval cadence, and incident response when an agent cites the wrong definition of revenue.
Gap-check ontologies and real-time intelligence.
Semantic models cover the past definitions. Ontologies add rules and relationships. Real-time intelligence covers live signals. Clients who only refresh nightly batch will feed agents stale context.
Gate downstream agent surfaces.
The same Fabric IQ layer can reach Foundry, Copilot Studio, GitHub Copilot, and MCP-based agents. Name who approves those paths.
Offer moves this week
Run a five-point Fabric IQ readiness check.
Inventory semantic models; confirm Copilot askers; gap-check ontologies and RTI; gate Foundry/Studio/GitHub paths; name owners and eval cadence for semantic changes.
Pick one client pilot.
Choose a book where Power BI already defines the board metrics. Turn the semantic model review into a paid governance engagement before they treat Chat as finance.
Update the SOW language.
Who owns semantic changes, who can pause Copilot grounding paths, and how wrong-number incidents get triaged.
Watch Ignite (Nov 17-20).
Microsoft flagged a broader Frontier ecosystem story there. Treat Barcelona as the data-foundation chapter, not the end of the packaging.
Numbers table
Figures below appear in Microsoft’s Azure Blog, CIO’s FabCon coverage, and/or The New Stack’s reporting. None were measured by Consultiply.
| Claim | What was measured | Status |
|---|---|---|
| Fabric IQ in Copilot Chat and Cowork | GA Sep 28, 2026; on by default for Fabric and Power BI customers. From Microsoft Blog / Azure Blog. | Vendor claim |
| Extra AI token cost | None. Vendor claim (Azure Blog, CIO, The New Stack). | Vendor claim |
| Power BI semantic models | 22 million. The New Stack citing Netz. | Vendor claim |
| Fabric customers | 40,000+. Azure Blog / CIO. | Vendor claim |
| Active Fabric users | 2.6 million. CIO. | Vendor claim |
| FabCon Barcelona | ~5,000 attendees / 71 countries. Azure Blog / CIO. | Vendor claim |
This guide is Consultiply’s independent analysis of Microsoft’s own FabCon Barcelona announcements and related coverage from The New Stack and CIO, translated for MSP operators. It is not affiliated with or endorsed by Microsoft.
Reviewed, not run by us.
This guide is Consultiply’s independent analysis of Microsoft’s own FabCon Barcelona announcements and related coverage from The New Stack and CIO, translated for MSP operators. It is not affiliated with or endorsed by Microsoft. Official Microsoft Blog and Azure Blog FabCon announcements (Sep 28, 2026) →
