Organizations have long struggled with moving data out of relational stores to run AI models, incurring latency, security risks, and overhead.
SQL Server 2025 brings AI inference directly into the database engine, enabling vector searches, embedding generation, and language model completions where data resides.
This tight integration removes ETL pipelines that copy sensitive data to external services, reducing attack surfaces and simplifying compliance.
Enterprises can invoke models with familiar T-SQL syntax while benefiting from built-in security, backup, and high-availability features.
The result is a streamlined architecture that accelerates time-to-insight and lowers total cost of ownership for AI-driven applications.
Early adopters report up to 60% reduction in data preparation time, freeing data scientists for feature engineering and model tuning.
Keeping data inside the database ensures consistent lineage, versioning, and access controls, vital for regulated industries.
The collaboration between Dell Technologies and Microsoft provides a cohesive stack that runs on premises, at the edge, or in Azure Local without re-architecting.
Dell’s firmware management and power-efficiency telemetry are exposed through Azure Arc, giving a single pane of glass for monitoring.
SQL Server 2025 hosts embedding and chat completion models inside the engine, using GPU-accelerated kernels when available.
Vector similarity searches leverage native HNSW indexes for sub-second nearest-neighbor lookups without moving data.
Transactions involving AI-derived insights can be rolled back or committed atomically, preserving data integrity.