Per-Device License
A single per-device license covers an unlimited number of users or connections accessing that device โ no CAL math, no per-core multipliers.
Your IoT/Embedded Channel Partner & more
SQL Server IoT
Predictable per-device database licensing for embedded appliances, edge gateways, and industrial systems โ including full AI and vector search capabilities in SQL Server 2025.
SQL Server IoT is a dedicated licensing model for fixed-function devices and appliances. Unlike commercial SQL Server editions โ which are licensed per core or per user โ SQL Server IoT is licensed per device, making it highly cost-effective for OEM manufacturers deploying hundreds or thousands of units. The functionality is identical to the standard SQL Server editions; only the licensing model differs.
Why SQL Server IoT?
No Subscription Required
SQL Server IoT is a perpetual license. Pay once per device, own the license for the life of the product โ no annual renewal.
10-Year Availability Guarantee
All SQL Server IoT editions are available for 10 years after release through the embedded channel, matching your hardware and OS lifecycle.
Full SQL Server Feature Set
SQL Server IoT is functionally identical to commercial SQL Server โ no feature limitations. Every SQL Server capability is available, including AI, replication, and HA features.
Edge & On-Premise Ready
Runs entirely on-premises, with no cloud dependency. Ideal for air-gapped, offline, or bandwidth-constrained industrial environments.
OEM Activation Support
OEM activation keys are available for volume manufacturing, enabling pre-activation during image creation without per-unit internet connectivity.
SQL Server IoT 2025 โ AI-Native Database for the Edge
New
-
Native Vector Search & Vector Data Type
SQL Server 2025 introduces a native `vector` data type and built-in vector distance functions (cosine similarity, Euclidean distance, dot product). Store, index, and query high-dimensional vector embeddings directly in the relational database โ no separate vector store required.
This enables semantic similarity search over structured operational data without additional infrastructure. Embeddings generated by AI models (local or cloud) can be persisted and searched at scale directly within T-SQL.
-
Retrieval-Augmented Generation (RAG) in T-SQL
SQL Server 2025 is the first relational database with built-in RAG pipeline support. The database can act as the knowledge source for Large Language Models โ retrieve semantically relevant data from your operational database and inject it into AI model context windows at query time.
RAG allows AI models to answer questions about your specific product data, maintenance records, sensor history, or customer data โ without retraining the model. The LLM stays current because it reads live from your database.
-
AI Model Integration via REST (sp_invoke_external_rest_endpoint)
Call external AI inference endpoints (Azure OpenAI, local Ollama, ONNX Runtime, or any REST API) directly from T-SQL stored procedures and functions. Enrich query results with AI-generated text, classify sensor readings, summarize log entries, or translate device status messages โ all within a SQL transaction.
For embedded and edge scenarios, this means local AI model inference can be triggered from within the database engine without leaving the device.
-
Semantic Chunking & Document AI Pipelines
Ingest, chunk, and vectorize unstructured documents (manuals, maintenance logs, inspection reports) directly in SQL Server 2025 using built-in chunking functions. Combined with vector search and RAG, this creates a complete document Q&A pipeline on-device โ enabling AI assistants that understand your machine documentation without internet access.
-
ONNX Model Inference in-Database
Execute ONNX format AI models directly inside SQL Server 2025 using `PREDICT()`. Deploy trained anomaly detection, predictive maintenance, quality inspection, or classification models as SQL functions. Inference happens within the SQL engine โ no Python runtime, no external process, no round-trip latency.
-
JSON Native Type & LLM Response Handling
SQL Server 2025 introduces a native `json` data type with validation constraints and a rich set of JSON path functions. This simplifies storing and querying structured AI model responses, IoT telemetry payloads, and REST API results. Combined with AI integration, parse and shred LLM JSON output directly in T-SQL.
-
DiskANN Vector Index for High-Performance Similarity Search
SQL Server 2025 includes the DiskANN approximate nearest neighbor index โ a disk-aware graph-based index optimized for billion-scale vector search on commodity hardware. For embedded edge systems, this means accurate and fast semantic search over large embedding collections without GPU acceleration.
-
Available Editions (IoT Licensed)
- Microsoft SQL Server 2025 IoT Enterprise
- Microsoft SQL Server 2025 IoT Standard
Key Use Cases for SQL Server IoT + AI on the Edge
The combination of SQL Server IoT's per-device licensing and SQL Server 2025's AI capabilities enables new classes of intelligent embedded applications:
Predictive Maintenance
Run ONNX anomaly detection models on live sensor data within the database. Detect failure patterns before they cause downtime โ entirely on-device, no cloud latency.
AI-Assisted Service Terminals
Embed a RAG pipeline over maintenance manuals and service history on a technician terminal. Ask natural language questions, get answers grounded in your actual device data.
Quality Inspection & Vision AI
Store inspection results, defect classifications, and camera metadata in SQL Server. Run ONNX classification models in-database to detect quality issues and log structured outcomes.
Edge Analytics & Telemetry
Aggregate, filter, and analyze IoT sensor streams on the edge server. Use vector search to find similar historical patterns across time-series data.
Document Q&A on Kiosks
Vectorize product documentation, safety datasheets, or compliance records in SQL Server. Power touchscreen kiosk Q&A with a RAG pipeline that requires no internet connectivity.
Medical & Diagnostic Systems
Store patient measurement records, device calibration logs, and imaging metadata in a HIPAA-compatible on-premise SQL Server instance. AI-classify structured diagnostic data in-database.
SQL Server IoT 2022
-
SQL Server IoT 2022
Stable, long-term supported edition with Azure Synapse Link, built-in query intelligence (Query Store improvements), enhanced HA with contained Availability Groups, and improved T-SQL language features. Extended support until January 2033.
-
Available Editions
- Microsoft SQL Server 2022 IoT Enterprise
- Microsoft SQL Server 2022 IoT Standard
SQL Server IoT 2019 & 2017
-
SQL Server IoT 2019
Intelligent query processing, Big Data Clusters integration, UTF-8 support, and enhanced Java/Spark extensibility. Extended support until January 2030.
-
SQL Server IoT 2017
Cross-platform support (Linux, Docker), Graph database extensions, adaptive query processing. Extended support until October 2027.
-
Available Editions
- Microsoft SQL Server 2019 IoT Enterprise
- Microsoft SQL Server 2019 IoT Standard
- Microsoft SQL Server 2017 IoT Enterprise
- Microsoft SQL Server 2017 IoT Standard
SQL Server IoT Edition Comparison
| Feature | IoT Standard | IoT Enterprise |
|---|---|---|
| Licensing model | Per device | Per device |
| Max CPU sockets used | 4 | Unlimited |
| Max memory used | 128 GB | OS maximum |
| Replication | Subscriber only | Full publish & subscribe |
| Always On AG | Basic (2-node) | Full (8 secondaries) |
| Columnstore indexes | Yes | Yes |
| In-Memory OLTP | Yes | Yes |
| Vector search (2025) | Yes | Yes |
| ONNX inference (2025) | Yes | Yes |
| RAG pipeline (2025) | Yes | Yes |
| Data compression | Yes | Yes |
| Policy-based management | Limited | Full |
| Resource Governor | No | Yes |
Licensing & Ordering
SQL Server IoT licenses are available exclusively through authorized Microsoft embedded distributors. Elbacom supplies SQL Server IoT licenses individually or in volume for OEM manufacturing runs. Contact us for current pricing, edition guidance, and compatibility with your target hardware platform.