- Azure SQL Database is a database-scoped PaaS service: Microsoft runs the latest stable SQL Server engine, patching, backups and high availability, and you manage databases, not servers.
- Two purchasing models: DTU (bundled Basic, Standard and Premium sizes) and vCore (General Purpose, Business Critical and Hyperscale, with provisioned or serverless compute). Microsoft labels vCore the recommended model.
- Instance-level SQL Server features such as SQL Server Agent, cross-database queries, CLR and Database Mail are not available; if you need them, look at Azure SQL Managed Instance instead.
- A free offer gives each subscription up to 10 General Purpose serverless databases with a monthly compute and storage allowance, which is enough for learning and prototypes.
What is Azure SQL Database?
Azure SQL Database is a platform as a service (PaaS) relational database built on the SQL Server Database Engine. Microsoft's overview states that it always runs on the latest stable version of the engine and a patched operating system, so there is no version to choose and no upgrade project to plan. Microsoft handles patching, automated backups, high availability and monitoring; you create databases, size them and connect with the usual SQL Server tools and drivers.
People search for it under many names (Azure SQL DB, Microsoft Azure SQL, SQL Server Azure database, Azure MSSQL database), but they usually mean this product. It is one of three members of what Microsoft calls the Azure SQL family: Azure SQL Database, Azure SQL Managed Instance and SQL Server on Azure Virtual Machines. The three share the engine but differ in how much of SQL Server you get and how much you manage yourself.
A database in Azure SQL Database lives on a logical server, which is a management and connection endpoint rather than a machine you can log in to. IP firewall rules can be set for the whole logical server (stored in its master database) or for an individual database.
Deployment options: single database, elastic pool, Managed Instance or VM
Azure SQL Database itself has two deployment models. A single database is a fully managed, isolated database with its own resources; Microsoft compares it to a contained database in SQL Server. An elastic pool is a collection of single databases that share a set amount of CPU and memory at a set price, which suits many small databases whose usage peaks at different times, such as one database per customer in a SaaS application. Databases can move into and out of a pool.
The other two Azure SQL options are separate products. Azure SQL Managed Instance gives you a whole managed SQL Server instance inside your virtual network with near-complete compatibility. SQL Server on Azure VMs is infrastructure as a service: you get the full product and operating system access, and you also do the patching and backups. Our comparison of Azure SQL Database vs Managed Instance goes deeper on that choice.
| Option | What you manage | SQL Server compatibility | Typical fit |
|---|---|---|---|
| Single database | Database settings, size, security | Database-scoped features only | New cloud applications, microservices |
| Elastic pool | Pool size and which databases are in it | Same as single database | Many databases with uneven, unpredictable use |
| SQL Managed Instance | Instance settings, virtual network | Near 100% with SQL Server Enterprise Edition | Lift-and-shift migrations that need instance features |
| SQL Server on Azure VM | Operating system, SQL Server, patching, backups | 100%, with OS access | Workloads that need full control or unsupported features |
DTU vs vCore purchasing models
You pay for Azure SQL Database through one of two purchasing models. The DTU model sells preconfigured bundles: a database transaction unit is a blended measure of CPU, memory, reads and writes, and each compute size includes a fixed amount of storage at a fixed price. It has three service tiers, Basic, Standard and Premium. Microsoft describes it as suited to customers who want simple, preconfigured resource options.
The vCore model, which Microsoft's comparison marks as recommended, lets you choose compute (vCores and memory) and storage separately, pick a hardware configuration, and use discounts that DTU does not offer: Azure Hybrid Benefit for existing SQL Server licences and reserved capacity. It has three service tiers, General Purpose, Business Critical and Hyperscale. You can convert a database between the two models online.
For list prices, worked monthly estimates and how serverless billing is calculated, see Azure SQL pricing explained.
Azure SQL Database service tiers: General Purpose, Business Critical, Hyperscale
In the vCore model the service tier sets the storage architecture, the high availability design and some features. General Purpose separates a stateless compute node from data files held in Azure premium remote storage; Microsoft gives typical storage latency of 5 to 10 ms and describes it as budget-oriented. Business Critical runs a cluster of replicas on local SSD, with a free readable secondary and In-Memory OLTP support; the documentation states the extra replicas make it about 2.7 times the price of General Purpose. Hyperscale decouples compute from a distributed storage layer that grows to 128 TB, and lets you choose 0 to 4 high availability replicas plus named read replicas.
Microsoft's tier table now calls Hyperscale the recommended and default tier for new and modernising OLTP and HTAP workloads, while General Purpose remains the budget option and the tier used by the free offer. In the DTU model, Basic and the smallest Standard sizes run on standard (HDD-based) storage and provide less than one vCore, which Microsoft says suits development, testing and infrequently accessed workloads.
| General Purpose | Business Critical | Hyperscale | |
|---|---|---|---|
| Compute size | 2 to 128 vCores | 2 to 128 vCores | 2 to 192 vCores |
| Storage size | 1 GB to 4 TB | 1 GB to 4 TB | 10 GB to 128 TB |
| Storage type | Premium remote storage | Local SSD | Decoupled storage with local SSD cache |
| Availability | One replica, no read scale-out | Three replicas, one read scale-out replica | Up to 4 read scale-out replicas |
| In-memory tables | No | Yes | No |
| Serverless compute | Yes | No | Yes |
Serverless compute and auto-pause
Within the vCore model you choose provisioned compute (a fixed size billed per hour) or serverless compute, which scales between a minimum and maximum number of vCores that you set and is billed per second for what it uses. Serverless is available in General Purpose and Hyperscale, on standard-series (Gen5) hardware only. In General Purpose a serverless database can pause automatically after an inactivity delay of at least 15 minutes (the default is 60), and while paused only storage is billed; auto-pause is a preview feature in Hyperscale.
Microsoft positions serverless for single databases with intermittent, unpredictable usage that can tolerate a warm-up delay after idle periods, and provisioned compute for steady workloads or for several databases that fit an elastic pool. Because the first connection after a pause has to wait for the database to resume, Microsoft's guidance is to build retry logic into applications that connect to serverless databases.
Azure SQL Database vs SQL Server: what is different
Because each database is isolated and the instance is managed for you, several SQL Server features are absent or replaced. Microsoft's feature comparison lists, among others:
- No SQL Server Agent: use elastic jobs to schedule T-SQL across databases.
- No cross-database or three-part name queries and no linked servers: elastic queries are the documented alternative.
- No CLR, Database Mail or Service Broker.
- No user-initiated
BACKUP: backups are automatic, and you cannot restore a database to an on-premises SQL Server; use a BACPAC export instead.
Connections go through a firewall. Clients outside Azure need outbound TCP port 1433 and a server-level or database-level IP firewall rule; a blocked address produces error 40615, and wrong credentials produce the familiar error 18456. You can connect with SSMS, the MSSQL extension for Visual Studio Code, sqlcmd or the portal's query editor; our SQL Server GUI tools guide covers the desktop options. For a fuller list of differences, see Azure SQL Database vs SQL Server.
When to use Azure SQL Database, and when not to
In our assessment Azure SQL Database suits:
- New applications written for SQL Server or T-SQL that need a database and no instance-level features.
- SaaS products with a database per tenant, where elastic pools spread capacity across many small databases.
- Development, test and low-traffic databases that sit idle for long periods, using serverless auto-pause or the free offer.
- Large or fast-growing OLTP databases that would outgrow a single machine's storage, using Hyperscale.
It is the wrong choice when an application depends on SQL Server Agent jobs, cross-database queries, CLR assemblies or linked servers (choose Managed Instance), when you need operating system access or a feature the PaaS services do not support (choose SQL Server on a VM), or when your team prefers an open source engine (see Azure Database for PostgreSQL and Azure Database for MySQL). If you are weighing clouds rather than Azure products, compare AWS RDS vs Azure SQL Database and Azure SQL vs Google Cloud SQL. For all Azure database services side by side, start at the Azure database services comparison.
Frequently asked questions
Is Azure SQL Database the same as SQL Server?
It runs the same SQL Server Database Engine, always on the latest stable version, so T-SQL, drivers and tools carry over. It is not a full SQL Server instance: instance-level features such as SQL Server Agent, cross-database queries, CLR and Database Mail are not available, and backups are managed for you.
What is the difference between Azure SQL Database and Azure SQL Managed Instance?
Azure SQL Database is scoped to individual databases (or a pool of them). Managed Instance is a whole SQL Server instance in your virtual network with near-complete compatibility, including SQL Server Agent, cross-database queries, CLR and native backup restore. Managed Instance suits migrations; SQL Database suits new cloud applications. See Azure SQL Managed Instance.
Should I choose DTU or vCore?
Microsoft recommends vCore: it offers more tiers and sizes, separate compute and storage, serverless compute, Azure Hybrid Benefit and reserved capacity. DTU remains a simple option with a fixed monthly price for small databases. You can convert between them online.
Is there a free Azure SQL Database?
Yes. The free offer lets each Azure subscription create up to 10 General Purpose databases, each with 100,000 vCore seconds of serverless compute, 32 GB of data and 32 GB of backup storage per month, for the lifetime of the subscription. Details are in Azure SQL Database free offer.
How large can an Azure SQL database be?
General Purpose and Business Critical databases can be configured from 1 GB up to 4 TB. Hyperscale databases grow automatically up to 128 TB.
Sources
- Microsoft Learn: What is Azure SQL Database?
- Microsoft Learn: What is Azure SQL? (IaaS vs PaaS)
- Microsoft Learn: Compare vCore and DTU purchasing models
- Microsoft Learn: vCore purchasing model
- Microsoft Learn: DTU-based purchasing model
- Microsoft Learn: Serverless compute tier
- Microsoft Learn: Features comparison, SQL Database and SQL Managed Instance
- Microsoft Learn: IP firewall rules
- Microsoft Learn: Azure SQL Database free offer
Checked 8 October 2026.
How we research cloud database guides: our editorial method. CodeWithSQL earns nothing from the providers mentioned.