From Model Power to Data Context in Enterprise AI
Microsoft’s latest enterprise AI strategy centers on the idea that powerful models matter less than giving AI agents rich, shared data context about how an organization works, what it values, and which actions are allowed across its systems. This shift moves attention from training ever-larger models to building reliable data infrastructure, semantics, and governance so AI agents behave more like informed employees than outsiders. At Build, Microsoft framed this as the key difference between consumer AI and enterprise AI: the latter must work inside complex business processes, policies, and data estates. The company is positioning Microsoft Fabric as the substrate where that context lives, spanning operational databases, analytical stores, semantics, and real-time signals. This reflects a broader market trend: as foundational models become widely available, competitive advantage is migrating upstream into enterprise AI strategy built around data context AI and organizational memory.
Azure HorizonDB: A Database Built for AI-Scale Context
Azure HorizonDB is Microsoft’s new managed, PostgreSQL-compatible database, designed to store and serve the context that AI agents need at enterprise scale. In public preview, it promises elastic storage up to 128 TB, compute scaling to 3,072 vCores, and sub-millisecond multi-zone commit latency, targeting high-demand transactional workloads that sit at the heart of business operations. HorizonDB combines transactional data, vector search, and integrated AI model management in one platform, connecting directly to Microsoft Foundry and Fabric. That design is meant to avoid brittle architectures where developers stitch together separate systems for search, AI, and line-of-business data. Mohsin Shafqat, Director of Software Engineering at NASDAQ, said HorizonDB “brings transactional data, vector search, and AI capabilities into a single platform, which simplifies the architecture without forcing a complete rethink.” In practice, it turns operational databases into live context hubs for AI agents organizational data.
GPU-Accelerated Fabric Data Warehouse and Real-Time AI Workloads
Microsoft’s GPU-accelerated Fabric Data Warehouse pushes the same context-first idea into analytics, where AI agents need fast access to large, shared datasets. Entering early access preview in July 2026, the warehouse integrates NVIDIA accelerated computing directly into the query engine with no query rewrites, a key requirement for existing BI and analytics teams. Microsoft claims up to 7x faster performance than three unnamed cloud data warehouse competitors at 64-user concurrency in internal benchmarks, and UNC Health is reporting up to 5x query speed improvements. These gains matter for AI agents that reason over enterprise-wide data while serving many users at once. Ian Buck of Nvidia notes that such low-latency, multi-user workloads align well with GPU acceleration. For Microsoft, this is about more than faster analytics; it is about ensuring that data context AI has responsive, scalable access to the shared data fabric that underpins enterprise decisions.
Fabric IQ and Ontologies: Turning Data into Organizational Memory
Fabric IQ, now generally available, is Microsoft’s semantic and ontology layer aimed at turning scattered enterprise data into coherent organizational memory for AI agents. Built on top of Power BI semantic models that serve roughly half a million organizations, Fabric IQ adds business entities, relationships, rules, real-time signals from Fabric Real-Time Intelligence, and explicit definitions of which actions agents may take. This context flows into operations agents, now generally available, which continuously monitor live data and act on predefined business logic. Microsoft is threading Fabric IQ through its broader ecosystem: as a knowledge source in Microsoft Foundry, as a first-party MCP tool for Microsoft Agent 365, and as grounding for Microsoft 365 Copilot, including Cowork and Copilot Chat. Planning in Fabric will write forecasts back into the platform, so, as Amir Netz describes it, Fabric’s ontology can now cover the past, present, and expected future of the business.
A Maturing Enterprise AI Strategy Built on Shared Context
Taken together, Azure HorizonDB, GPU-accelerated Fabric Data Warehouse, and Fabric IQ show Microsoft reframing enterprise AI strategy around data context, not model supremacy. Fabric is pitched as both the data foundation and deployment target for AI agents, unifying operational and analytical workloads in OneLake while layering semantics and governance on top. This stance contrasts with rivals that focus mainly on analytics; Microsoft argues that AI agents that build and run applications need a single platform that combines transactions, analytics, semantics, and planning. New additions such as Database Hub for central database management, graph in Fabric to model relationships, and Rayfin for agent-built backends further extend this context-centric approach. As foundational models commoditize, the advantage shifts to organizations that treat AI agents organizational data as a managed asset, with clear semantics and performance guarantees. Microsoft’s Build announcements are a direct bid to be the platform where that asset lives.