Azure SQL
Azure SQL is Microsoft's family of SQL Server engine services on Azure: Azure SQL Database for cloud applications (with Hyperscale and serverless options), Azure SQL Managed Instance for near-complete SQL Server compatibility, and SQL Server on Azure Virtual Machines for full OS and engine control.1
Also known as Azure SQL Database, Azure SQL Managed Instance, SQL Server on Azure Virtual Machines
- Intended for
- Database administrators, Application developers, SaaS and ISV teams, IT teams migrating SQL Server
- Runs on
- Azure SQL Database (PaaS), Azure SQL Managed Instance (PaaS in your virtual network), SQL Server on Azure VMs (IaaS, Windows or Linux)
At a glance
- What is it?
- Azure SQL is not a single service but a family of products that run the SQL Server database engine in Azure. Azure SQL Database is a platform-as-a-service relational database based on the latest stable Enterprise Edition engine, offered as single databases (with Hyperscale and serverless options) or elastic pools. Azure SQL Managed Instance is PaaS with almost 100% SQL Server feature parity and instance-scoped features, which Microsoft recommends for most migrations to the cloud. SQL Server on Azure VMs is infrastructure as a service with full control of the engine and operating system. SQL database in Microsoft Fabric is a related but distinct product.1
- What does it do?
- Azure SQL runs transactional workloads with automated patching, backups and high availability in the PaaS options, and online changes to CPU and storage. It stores vectors natively: the vector data type and vector functions are generally available in Azure SQL Database and Managed Instance, and DiskANN-based vector indexes are generally available in Azure SQL Database and in Managed Instance on the Always-up-to-date update policy. Azure SQL Database and Managed Instance can be mirrored into Microsoft Fabric OneLake for analytics.23
- Who needs it?
- Teams moving SQL Server applications to Azure, developers building new relational applications or multitenant SaaS products with many databases, and organizations that want retrieval-augmented generation over relational data without a separate vector store. DBAs who need OS access, specific SQL Server versions or full engine control choose SQL Server on Azure VMs.1
- What does it need?4
- An Azure subscription
- A logical server for Azure SQL Database, or a virtual network subnet for Managed Instance
- A firewall rule or private endpoint so clients can connect
- For migrations: an assessment of SQL Server features in use, such as CLR, SQL Agent jobs and OS dependencies
- For vector search: an embedding model, for example from Azure OpenAI
- What it is not
- Azure SQL is not a resource you create; you create a SQL database, a managed instance or a SQL Server VM. SQL database in Microsoft Fabric is distinct from Azure SQL Database even though it uses the same engine. Azure SQL is not a globally distributed NoSQL store; for that pattern Microsoft points to Azure Cosmos DB. The PaaS options do not support every SQL Server feature (CLR, for example, is not supported in SQL Database), and compatibility with older SQL Server versions is handled only through database compatibility levels.15
Availability and licensing. Generally available. Microsoft's Azure SQL overview says SQL Database prices start from $5 per month for the Basic tier (checked 26 Oct 2026). The free tier gives each database 100,000 vCore seconds, 32 GB of data and 32 GB of backup storage per month for the life of the subscription, up to 10 databases per subscription, without an SLA. SQL Server 2025 became generally available at Ignite 2025. Vector indexes are still preview in SQL Server 2025 and in Managed Instance on the SQL Server 2025 update policy.167
The problem it solves
On-premises SQL Server estates carry hardware, patching, backup and high-availability work, plus recurring end-of-support upgrades. Moving them to the cloud raises the question of how much compatibility to keep and how much administration to hand over.
Azure SQL offers three points on that scale: SQL Database for new cloud applications, Managed Instance for migrations with minimal changes, and SQL Server on Azure VMs for full control. The Azure SQL hub in the Azure portal includes a decision tree to choose between them.1
How it works
- Azure SQL Database: databases on a logical server, as a single database (with Hyperscale and serverless options) or in an elastic pool that shares resources; databases up to 128 TB.
- Azure SQL Managed Instance: an instance with system and user databases inside your Azure virtual network with a private IP, up to 16 TB; instance pools share resources between instances.
- SQL Server on Azure VMs: Windows or Linux VMs with SQL Server 2016, 2017, 2019, 2022 or 2025 in any edition and up to 256 TB of storage; you manage the OS and engine, with optional automation through the SQL Server IaaS Agent extension.
PaaS options are billed hourly by service tier and compute size under the vCore or DTU purchasing models; VMs by size and cores, with per-minute licensing for platform images. Azure Hybrid Benefit applies existing SQL Server licenses to all three.1
Diagram as a list
Copilot & business apps
- Application or AI agentConnects with standard SQL Server drivers and tools
AI models & platform
- Vector type and DiskANN vector indexesStores embeddings and runs similarity search in T-SQL
Data & analytics
- Azure SQL Database (single database or elastic pool)PaaS relational database, including Hyperscale and serverlessConnects to Logical server
- Azure SQL Managed InstancePaaS instance with near-complete SQL Server compatibilityConnects to Azure Virtual Network
- Fabric mirroring to OneLakeReplicates operational data for analytics
Cloud & hybrid infrastructure
- Logical serverManagement and connection boundary for SQL databases
- SQL Server on Azure VMsIaaS with full OS and engine controlConnects to Azure Virtual Network
- Azure Virtual NetworkPrivate network for Managed Instance and VMs
Capabilities
Three deployment options on one engine1
SQL Database (PaaS for modern cloud apps), SQL Managed Instance (PaaS with almost 100% SQL Server feature parity) and SQL Server on Azure VMs (IaaS with OS-level access).
Why it matters: Teams pick the trade-off between compatibility and administration per workload while keeping the same engine and tools.
Limits: SQL Database lacks some SQL Server features; SQL Server on VMs leaves high availability and patching largely to you.
Hyperscale and serverless1
Hyperscale is a separate architecture with independently scalable storage and compute tiers; serverless compute is available for single databases.
Why it matters: Supports very large databases and workloads with variable activity.
Limits: Azure SQL Database databases go up to 128 TB; Hyperscale behaves differently from other tiers, so test before moving.
Elastic pools for SaaS1
Elastic pools share a set of resources across many databases on one logical server, with tenant isolation per database.
Why it matters: Lowers cost for multitenant SaaS apps whose databases peak at different times.
Limits: Pool sizing needs monitoring when many tenants peak together.
Native vector search2
The vector data type and vector functions such as VECTOR_DISTANCE, plus DiskANN-based vector indexes searched with VECTOR_SEARCH.
Why it matters: Retrieval-augmented generation can run next to relational data without a separate vector database.
Limits: Vector indexes are preview in SQL Server 2025 and in Managed Instance on the SQL Server 2025 update policy; Microsoft recommends exact search for up to about 50,000 vectors.
Mirroring into Microsoft Fabric3
Azure SQL Database and Azure SQL Managed Instance can be continuously mirrored into OneLake as Delta tables.
Why it matters: Analytics and Power BI reports can use near real-time operational data without ETL.
Limits: Mirroring needs a running Fabric capacity; queries on mirrored data consume capacity.
Free tier for Azure SQL Database4
Up to 10 General Purpose serverless databases per subscription, each with 100,000 vCore seconds, 32 GB of data and 32 GB of backup storage free per month for the life of the subscription.
Why it matters: Development and proofs of concept can start at no cost.
Limits: No SLA; at the monthly limit the database either pauses until next month or bills overage, and you cannot switch back to auto-pause.
Licensing flexibility1
Azure Hybrid Benefit applies existing SQL Server licenses to all three products; SQL Server VMs from platform images bill licensing per minute.
Why it matters: Organizations with SQL Server licenses can reduce the cost of moving to Azure.
Limits: Savings depend on your existing license agreements; check them in the pricing calculator.
Practical use cases
A SQL Server estate with SQL Agent jobs and instance-level features must leave an ageing data center.
- Approach
- Move the databases to Azure SQL Managed Instance, which Microsoft recommends for most migrations.
- Role of Azure SQL
- Managed Instance hosts the databases with near-complete SQL Server compatibility and PaaS operations.
- Data, infrastructure and skills
- A virtual network subnet, connectivity to on-premises (ExpressRoute or VPN) and a feature assessment.
- Type of benefit
- Less infrastructure administration
- Caveats
- A few SQL Server features are not available; apps with OS dependencies may need SQL Server on VMs instead.
- First step
- Use the Azure SQL hub decision tree to confirm Managed Instance fits, then assess one database.
A SaaS vendor runs one database per customer and pays for idle capacity.
- Approach
- Group tenant databases in elastic pools on Azure SQL Database.
- Role of Azure SQL
- Elastic pools share compute across tenants with variable usage.
- Data, infrastructure and skills
- Usage data per tenant and an application that routes to per-tenant databases.
- Type of benefit
- Lower cost per tenant
- Caveats
- Noisy tenants can affect a shared pool; move them to their own database when needed.
- First step
- Move a group of low-usage tenant databases into one pool and watch resource use.
A product catalog in SQL needs semantic search for a support assistant.
- Approach
- Store embeddings in a vector column, create a vector index and query with VECTOR_SEARCH from the assistant.
- Role of Azure SQL
- Azure SQL holds both the product data and its embeddings.
- Data, infrastructure and skills
- An embedding model such as Azure OpenAI and Azure SQL Database or an up-to-date Managed Instance.
- Type of benefit
- Simpler AI architecture
- Caveats
- Recall of approximate search should be tested; vector indexes are preview in some products.
- First step
- Add a vector column to one table and test exact search with VECTOR_DISTANCE on a sample.
Who uses it
C.H. Robinson
In productionC.H. Robinson: AI agents for freight quotes, orders and tracking
from hours to an average of 32 seconds per quote (measured)
Logistics and transportation, United States (global operations)
Make-A-Wish America (national office and chapters)
In productionMake-A-Wish America: unified data, Copilot and an IT help bot
roughly two-thirds of staff (measured)
Nonprofit, United States
Air India
In productionAir India: AI.g customer service agent on Azure OpenAI
about 40,000 (2026), up from an average of 30,000 (2024) (measured)
Travel and hospitality, India
Works with
Optional integration
- Microsoft FabricAzure SQL Database and Managed Instance can be mirrored into OneLake.
- Microsoft Foundry ModelsEmbeddings from Azure OpenAI models can be stored and searched with the vector type.
- Microsoft FoundryFoundry IQ knowledge bases can use Azure SQL as a knowledge source.
Alternative approaches
- Azure Cosmos DBMicrosoft suggests Azure SQL for highly relational apps.
Optional integration for
- Microsoft FabricAzure SQL Database and Managed Instance can be mirrored into OneLake.
Relationship labels follow Microsoft's documentation. "Alternative approaches" does not mean one is better: each profile says when it fits.
Compared with
Getting started
Pick the product
Open the Azure SQL hub (aka.ms/azuresqlhub) and use the Get recommendation decision tree or the side-by-side comparison.
Check: You know whether the workload goes to SQL Database, Managed Instance or a SQL Server VM.
Create a free database
In the Azure SQL hub select Start free, or run az sql db create with --use-free-limit true, choosing AutoPause at the monthly limit.
Check: The database Overview shows the free monthly vCore amount remaining.
Connect
Add a firewall rule for your client IP and connect with the portal query editor, SQL Server Management Studio or the MSSQL extension for Visual Studio Code.
Check: A test query returns results.
Try vector search
Add a vector column, load embeddings and query with VECTOR_DISTANCE; for larger sets create a vector index.
Check: Similar rows come back ranked by distance.
Connect analytics
Create a mirrored database in Microsoft Fabric from the Azure SQL Database.
Check: Tables appear in OneLake and the SQL analytics endpoint.
Official resources
Deciding whether you need it?
Three free tools apply Microsoft's own published guidance to your situation, and say when Azure SQL or any new technology is not needed.
Sources
Each statement above links to the source it comes from. Labels say who reported it.
- What is Azure SQL? (opens in a new tab)
- Vector search and vector index in the SQL Database Engine (opens in a new tab)
- Mirroring in Microsoft Fabric (opens in a new tab)
- Try Azure SQL Database for free (opens in a new tab)
- Azure Cosmos DB overview (Unified AI Database) (opens in a new tab)
- Microsoft Ignite 2025 Book of News (opens in a new tab)
- Azure SQL product page (opens in a new tab)
Thank you. Your correction was sent.
The editors check it against the sources. If you left an email address, they may reply about it.