From Smarter Models to Smarter Context: A New Enterprise AI Bottleneck
Enterprise AI data context refers to the structured business meaning wrapped around raw information—entities, relationships, rules, history, and real-time signals—that allows AI agents to reason like informed insiders instead of generic chatbots, and it is quickly becoming more decisive for AI agent performance than incremental advances in model architectures or parameter counts. At Microsoft Build, Microsoft argued that the hard part of enterprise AI is no longer the model itself but the lack of shared organizational memory across systems. Amir Netz, CTO of Microsoft Fabric, framed the shift clearly: enterprises want AI that behaves like an employee who knows “how the machinery operates, what the goals are,” not an outsider starting from zero. The Build announcements centered on solving this gap through a unified data and context layer that spans databases, analytics, and agents.
Azure HorizonDB: A Database Built for AI-Scale Context
Azure HorizonDB is Microsoft’s new PostgreSQL-compatible database designed to keep enterprise AI agents close to the data context they need. Now in public preview, it supports elastic storage up to 128 TB and compute scaling to 3,072 vCores, with sub-millisecond multi-zone commit latency for demanding workloads. For AI agent performance, the key is that transactional data, vector search, and AI model management live in one place, instead of being spread across separate systems. According to NASDAQ’s engineering director Mohsin Shafqat, HorizonDB “brings transactional data, vector search, and AI capabilities into a single platform, which simplifies the architecture without forcing a complete rethink.” Direct links to Microsoft Foundry and Fabric mean agents can treat HorizonDB as both system of record and contextual memory, reducing latency between operations, retrieval, and reasoning over organizational data.
GPU-Accelerated Fabric Data Warehouse: Speed for Context-Hungry Agents
To feed AI agents with large volumes of shared context at low latency, Microsoft is adding GPU acceleration to Fabric Data Warehouse, entering early access preview in July 2026. Microsoft has integrated NVIDIA accelerated computing directly into the warehouse engine, so existing queries gain speed without rewrites. Internal benchmarks show up to 7x faster performance than three unnamed competitors at 64-user concurrency, and preview customers like UNC Health report up to 5x faster queries. This matters because organizational data strategy increasingly depends on agents that must scan, join, and analyze live data for many users at once. Ian Buck, Vice President of Hyperscale and HPC at Nvidia, notes that AI agents reasoning over enterprise data are a natural fit for GPUs. Faster warehousing effectively widens the contextual “window” agents can operate on in real time.
Fabric IQ and Ontologies: Embedding Shared Meaning Into Enterprise Workflows
If HorizonDB stores facts and Fabric Data Warehouse accelerates queries, Fabric IQ supplies the meaning that turns data into organizational memory. Now generally available, Fabric IQ sits on top of Power BI’s widely adopted semantic models and extends them with business entities, relationships, rules, live signals from Fabric Real-Time Intelligence, and explicit action permissions for agents. Operations agents that monitor live data and execute predefined business logic also reached general availability, while ontologies in Fabric IQ are expected to become generally available soon. Microsoft is weaving Fabric IQ across its agent ecosystem: as a knowledge source in Microsoft Foundry, a first-party tool in Microsoft Agent 365, and a grounding layer for Microsoft 365 Copilot experiences. Graph and planning capabilities in Fabric further enrich this context, letting agents connect past data, present signals, and forward-looking plans in one loop.
Organizational Data Strategy: Context as the New Enterprise AI Foundation
The Build announcements add up to a clear bet: enterprise AI data context, not raw model power, will decide which platforms win. Microsoft positions Fabric as both data foundation and deployment target for enterprise AI agents, spanning operational and analytical workloads in a single environment. A new Database Hub in Fabric will centralize management of HorizonDB, Azure Database for PostgreSQL, and Azure Cosmos DB, mirroring data into OneLake to reduce silos. Cosmos DB itself gains AI-focused features such as semantic reranking and an agent memory toolkit, underscoring that every data tier is expected to feed context into agents. With tools like the Rayfin SDK for building backends directly on Fabric, Microsoft’s organizational data strategy is to make context the default: agents should inherit a consistent ontology and semantic layer wherever they run, turning AI from a clever outsider into a dependable digital coworker.
