From Model Power to Organizational Data Context
Enterprise AI data context refers to the shared, structured understanding of an organization’s data, relationships, rules, and goals that allows AI agents to act like informed insiders instead of generic tools. At Microsoft Build, the message was clear: the hard part of enterprise AI is no longer picking or tuning a large language model. It is giving that model consistent organizational data context so it does not start from zero with every query or workflow. Amir Netz, CTO of Microsoft Fabric, framed the difference as AI that behaves like an employee who knows how the machinery operates, rather than a stranger. This shift moves enterprise AI strategy away from model horsepower toward data organization, common semantics, and governance that can serve many AI agents at once.
Azure HorizonDB and the New AI Agent Data Infrastructure
Azure HorizonDB sits at the core of Microsoft’s vision for AI agent data infrastructure, bringing transactional data, vector search, and AI features into one managed PostgreSQL-compatible platform. Designed for AI-scale workloads, it offers elastic storage up to 128 TB and compute scaling to 3,072 vCores, with sub-millisecond multi-zone commit latency for high-demand transactions. HorizonDB connects directly to Microsoft Foundry and Fabric, so developers do not need to stitch together separate systems for operations, search, and inference. As Mohsin Shafqat of NASDAQ notes, “Instead of stitching together multiple components, it brings transactional data, vector search, and AI capabilities into a single platform, which simplifies the architecture without forcing a complete rethink.” This integration turns the operational database into a live context hub that AI agents can query, reason over, and update in near real time.
GPU-Accelerated Fabric Data Warehouse and Performance as Context Enabler
Microsoft’s GPU-accelerated Fabric Data Warehouse extends this enterprise AI data context story into analytics performance. By integrating NVIDIA accelerated computing directly into the warehouse, Microsoft reports up to 7x faster performance than three unnamed competitors at 64-user concurrency in internal benchmarks. In data warehousing, even a 10 percent yearly gain is notable; according to Amir Netz, GPU acceleration is now yielding speedups “anywhere from 5x to 100x.” For AI agents that must reason over large volumes of enterprise data while serving many users at once, this matters as much as model choice. UNC Health’s early results show up to 5x faster queries, freeing teams to focus on insights instead of tuning. High-throughput, low-latency analytics turns historical and near-real-time data into accessible organizational data context that agents can use for troubleshooting, forecasting, and decision support.
Fabric IQ: Turning Disparate Data into Shared Organizational Memory
Fabric IQ, now generally available, is Microsoft’s semantic and ontology layer that converts scattered datasets into a single organizational data context for AI agents. Built on Power BI semantic models, already used by roughly half a million organizations, Fabric IQ adds business entities, relationships, rules, permitted actions, and real-time signals from Fabric Real-Time Intelligence. Operations agents can use this layer to monitor live data and act on predefined logic, while ontologies (due for GA soon) encode how the business is structured. Fabric IQ is integrated across Microsoft’s agent ecosystem: as a knowledge source in Microsoft Foundry, as a first-party MCP tool for Microsoft Agent 365, and inside Microsoft 365 Copilot and GitHub Copilot CLI. Netz describes the result as an ontology that covers past data in OneLake, present signals, and future plans so that agents see a continuous, closed-loop view of the business.
A Maturing Enterprise AI Strategy Focused on Data Infrastructure
The Build announcements signal a maturation of enterprise AI strategy: success now hinges on enterprise AI data context, not chasing the latest model. Microsoft is pitching Fabric as both data foundation and deployment target, spanning operational and analytical workloads with a shared context layer on top. This contrasts with platforms framed primarily as analytical warehouses, and it aligns with how AI agents actually work: they need operational state, historical trends, and future plans in one governed fabric. New pieces like the Database Hub for managing HorizonDB, Azure Database for PostgreSQL, and Azure Cosmos DB, as well as graph and planning in Fabric, all point to the same goal. Enterprise AI is shifting from model selection to building reliable, governed data infrastructure and semantics that any agent can plug into, turning AI from isolated experiments into an integrated part of organizational memory.
