IT, Data & Development

Best Database Software in 2026

Thirteen database systems compared on data model, licensing, deployment, AI features and published pricing, from Oracle and SQL Server to PostgreSQL.

Database software (a database management system, or DBMS) stores, organises and retrieves structured data and enforces rules such as transactions, access control and recovery. The category spans relational databases queried with SQL, document, key-value and in-memory stores, embedded engines and managed cloud services.

This list covers thirteen database products in November 2026: commercial relational systems (Oracle AI Database, Microsoft SQL Server, IBM Db2), open-source relational engines (PostgreSQL, MySQL, MariaDB, SQLite), the document database MongoDB, the in-memory store Redis, the distributed SQL database CockroachDB, and cloud services built on open engines (Amazon Aurora, Supabase, Neon). Prices come from vendors' own pages where published.

Contents
  1. What is database software
  2. History and market
  3. Key features to compare
  4. Pricing models
  5. How to choose
  6. Methodology
  7. The list
  8. Comparison table
  9. Alternatives by use case
  10. Recent developments
  11. Frequently asked questions
  12. See also

What is database software

A DBMS manages persistent data on behalf of applications. Relational systems organise data in tables with schemas and support SQL, ACID transactions and constraints; PostgreSQL, for example, implements all SQL-standard isolation levels, write-ahead logging and point-in-time recovery. Non-relational systems trade some of these properties for flexibility or speed: MongoDB stores JSON-like documents, and Redis keeps data structures in memory. Embedded databases such as SQLite run inside the application process without a server.

Adjacent categories are distinct. A data warehouse is optimised for analytical queries across large historical datasets rather than transactional updates. Vector databases specialise in similarity search over embeddings, a capability now added to general databases such as SQL Server 2025, Oracle AI Database and Supabase's pgvector. ETL tools and data integration tools move data between databases, and backup and cloud storage products protect it.

History and market

Relational databases became the dominant model in the 1980s. PostgreSQL's ancestor, the POSTGRES project at the University of California, Berkeley, began in 1986. Commercial engines from Oracle, IBM and Microsoft dominated enterprise computing, while MySQL and later MariaDB, started in 2009 by original MySQL developers, became standard for web applications. The 2000s and 2010s brought NoSQL systems such as MongoDB, founded in 2007, and Redis, alongside distributed SQL databases such as CockroachDB.

The 2020s are defined by managed cloud services and licensing shifts. Cloud providers built engines compatible with open-source databases, such as Amazon Aurora for MySQL and PostgreSQL. Redis moved from BSD to source-available licences in March 2024 and added AGPLv3 with Redis 8. Serverless Postgres services multiplied, and Neon now operates as part of Databricks. Vendors have added vector search and AI functions, with Oracle rebranding its flagship as Oracle AI Database 26ai and Microsoft adding a vector type in SQL Server 2025. No market-size estimate from a named research firm was verified for this list.

Key features to compare

Data model

Relational engines enforce schemas and joins; document and key-value stores favour flexible or very fast access. Several products now combine models: Oracle offers JSON Relational Duality and property graphs, and PostgreSQL supports JSONB.

Transactions and consistency

ACID transactions are standard in relational engines; CockroachDB extends them across distributed nodes with globally consistent transactions.

Scalability and availability

Options range from single-file SQLite to Aurora's multi-region availability of up to 99.999% and Redis Pro's active-active replication.

AI and vector support

Vector types and indexes now appear in SQL Server 2025, Oracle AI Database, Redis and Supabase via pgvector.

Licensing

Licences range from public domain (SQLite) and permissive (PostgreSQL) to GPL (MySQL, MariaDB), source-available or AGPL (Redis) and commercial (Oracle, SQL Server, Db2).

Deployment and operations

Self-managed software gives control but requires staff for backups, patching and failover; managed services such as Atlas, Aurora, Supabase and Neon shift that work to the provider.

Ecosystem and tools

Extensions, drivers and skills availability matter for long-term cost; PostgreSQL's extensions, including PostGIS, are a notable advantage.

Pricing models

Commercial databases are licensed per core, processor or user. SQL Server 2022 Enterprise lists at US$15,123 per two-core pack and Standard at US$3,945, with subscriptions and pay-as-you-go per-core rates through Azure Arc. IBM Db2 Standard and Advanced are sold as perpetual, monthly or subscription licences. Open-source engines (PostgreSQL, MySQL Community, MariaDB Community, SQLite) are free, with paid support or enterprise editions from vendors.

Managed services bill on compute and storage. Published entry points include MongoDB Atlas Dedicated from US$56.94 per month, Redis Cloud Essentials from US$5 per month, Supabase Pro from US$25 per month, Neon at US$0.106 per CU-hour, CockroachDB Standard at US$0.092 per vCPU-hour, Aurora serverless at US$0.156 per ACU-hour (I/O-Optimized example) and Db2 on IBM Cloud from US$630 per month. Free tiers exist for MongoDB Atlas, Redis Cloud, Supabase, Neon and Db2 on IBM Cloud.

How to choose

Start-ups and new applications often choose PostgreSQL, either self-managed or through Supabase or Neon, for its free licence and ecosystem.

Microsoft-centric enterprises typically standardise on SQL Server, while organisations running Oracle applications or very large transactional systems stay with Oracle AI Database.

Applications with evolving, document-shaped data fit MongoDB Atlas, and caching or real-time features fit Redis.

Global applications needing consistency across regions should evaluate CockroachDB and Aurora DSQL.

Embedded and mobile software uses SQLite, and IBM mainframe shops continue with Db2. Web workloads built on the LAMP stack commonly use MySQL or MariaDB.

Methodology

Products were included if they are generally available database management systems or managed database services in November 2026 with documentation on their own sites. Ordering is editorial, reflecting breadth of documented capabilities, maturity and installed base as described by the projects themselves, deployment flexibility and licensing transparency; commercial relational engines and the main open-source engines rank first, followed by specialised and cloud-native services. Oracle's price list and product pages returned access errors, so Oracle licence prices are omitted. Google Cloud and Azure SQL database pricing pages did not render rates and are not ranked. No benchmarks, user ratings or sponsored placements were used.

The list

1

Oracle AI Database

Best for: Large enterprises running mission-critical transactional systems

Oracle Database is Oracle Corporation's flagship commercial relational database. The current release is branded Oracle AI Database 26ai. Its documentation highlights JSON Relational Duality (which exposes relational data as JSON documents), SQL property graphs, JavaScript stored procedures, Sagas for microservices, in-database machine learning, and AI Vector Search for storing and querying embeddings alongside business data. The New Features guide for Release 26ai was last revised on 20 October 2026.

The documented features show a multi-model engine handling relational, JSON document, graph, machine-learning and vector workloads in one database. A free edition, Oracle AI Database 26ai Free, is offered for development and learning. Oracle's public price list and product pages returned access errors during this review, so licence prices are not quoted here. The feature breadth is the widest in this list; the main drawback for buyers is that licence costs cannot be checked against a public price list and must be negotiated with Oracle.

Key features

  • JSON Relational Duality
  • AI Vector Search
  • SQL property graphs
  • JavaScript stored procedures
  • In-database machine learning
  • Free edition for development

Pricing

Licence prices not verified (Oracle price list inaccessible during review). A no-cost Oracle AI Database 26ai Free edition is available.

Vendor
Oracle Corporation
Current release
26ai
Free edition
Oracle AI Database 26ai Free

Strengths

  • Most feature-complete relational engine
  • Multi-model including vectors and graphs
  • Proven at very large scale
  • Broad deployment options

Limitations

  • Licence prices not verifiable online
  • Commercial licence required for production
  • Proprietary features increase switching costs
  • Complex product to administer
2

Microsoft SQL Server

Best for: Organisations standardised on Microsoft technology

Microsoft SQL Server is the company's relational database for Windows and Linux, closely integrated with Azure, Power BI and Microsoft Fabric. SQL Server 2025 (version 17.x) adds a native vector data type and vector functions, a vector index in preview, management of external AI models for generating embeddings, a native JSON type, regular-expression functions, change event streaming to Azure Event Hubs or Fabric, an MCP server for AI agents and mirroring to Microsoft Fabric. Standard edition limits rose to the lesser of four sockets or 32 cores and 256 GB of buffer-pool memory, Express edition now allows 50 GB databases, Web edition was discontinued, and new free Standard Developer and Enterprise Developer editions were introduced.

Microsoft's published SQL Server 2022 prices give a guide to licensing costs: Enterprise at US$15,123 per two-core pack, Standard at US$3,945 per two-core pack or US$989 per server plus US$230 per client access licence, with Developer and Express free. Subscriptions cost US$5,434 per year (Enterprise) or US$1,418 (Standard) per two-core pack, and pay-as-you-go through Azure Arc costs US$274 or US$73 per core per month. Several components were dropped in 2025, including Master Data Services, Data Quality Services and Synapse Link.

Key features

  • Vector data type and AI functions
  • Native JSON type
  • Regular expression functions
  • Change event streaming
  • Mirroring to Microsoft Fabric
  • Free Express (50 GB) and Developer editions

Pricing

SQL Server 2022 list prices: Enterprise US$15,123 per 2-core pack; Standard US$3,945 per 2-core pack, or US$989 per server + US$230 per CAL; Developer and Express free. Subscriptions US$5,434 (Enterprise) / US$1,418 (Standard) per 2-core pack per year; Azure Arc pay-as-you-go US$274 / US$73 per core per month.

Vendor
Microsoft
Current release
SQL Server 2025 (17.x)
Enterprise list price
US$15,123 per 2-core pack (2022)

Strengths

  • Strong tooling and Microsoft integration
  • Published list prices
  • Free Express and Developer editions
  • Built-in AI and vector features in 2025

Limitations

  • Enterprise licensing expensive
  • Several components discontinued in 2025
  • Some new features still preview
  • Best fit in Microsoft-centric estates
3

PostgreSQL

Best for: Teams wanting a free, extensible, standards-compliant database

PostgreSQL began in 1986 as the POSTGRES project at the University of California, Berkeley, and is developed by a global community under the PostgreSQL Licence, a liberal open-source licence similar to BSD or MIT. The current major release is PostgreSQL 18, with point release 18.7 published on 12 November 2026 alongside updates to versions 14 to 17, and PostgreSQL 19 at release-candidate stage.

Its feature set is broad: JSON and JSONB, arrays, ranges, UUIDs and custom types; B-tree, GiST, GIN and BRIN indexes; multiversion concurrency control, partitioning, parallel query and JIT compilation; all SQL-standard isolation levels including Serializable; write-ahead logging with asynchronous, synchronous and logical replication and point-in-time recovery; and authentication via LDAP, SCRAM-SHA-256, certificates or OAuth 2.0 with row- and column-level security. Extensions such as PostGIS and foreign data wrappers allow it to act as a platform rather than a single product. Most cloud providers and many start-ups, including Supabase, Neon and Amazon Aurora, build services on PostgreSQL compatibility. The main limitation is that operations, such as high availability and connection pooling, require third-party tools or a managed service.

Key features

  • JSON/JSONB and rich data types
  • GIN, GiST, BRIN and B-tree indexes
  • Logical and synchronous replication
  • Row- and column-level security
  • Extensions including PostGIS
  • Parallel query and JIT

Pricing

Free and open source under the PostgreSQL Licence.

First released
1986 (POSTGRES project, UC Berkeley)
Licence
PostgreSQL Licence
Current version
18.7 (12 November 2026)

Strengths

  • Free with permissive licence
  • Highly extensible
  • Strong standards compliance
  • Huge ecosystem of managed services

Limitations

  • High availability needs external tooling
  • Tuning required at scale
  • No single vendor support by default
  • Write scaling requires sharding tools
4

MySQL

Best for: Web applications and LAMP-stack deployments

MySQL is a multithreaded, multi-user SQL database server owned by Oracle Corporation, intended for heavy-load production systems and for embedding in mass-deployed software. It is dual-licensed: free under the GNU General Public License, or under a commercial licence purchased from Oracle. The reference manual covers the 9.x innovation series, long-term support releases 8.4 and 8.0, and newer release tracks numbered 26.x.

Oracle sells several commercial offerings: MySQL HeatWave, a cloud service combining transactions with lakehouse-scale analytics, generative AI and machine learning; MySQL Enterprise Edition for business-critical applications; and an embedded edition for ISVs, OEMs and resellers. Standard Edition and NDB Cluster CGE are also listed; prices are obtained from Oracle sales. MySQL's strengths are its simplicity, enormous installed base in web applications and broad hosting support. Its limitations include a feature set narrower than PostgreSQL in areas such as advanced indexing and extensibility, and the fact that some enterprise features, such as advanced backup and auditing, are reserved for commercial editions.

Key features

  • GPL Community Server
  • MySQL HeatWave analytics and ML (cloud)
  • Enterprise Edition for production support
  • Embedded licensing for ISVs
  • NDB Cluster CGE

Pricing

Community Server free under GPL; Enterprise, Standard, Cluster CGE and embedded licences priced by Oracle sales; HeatWave billed as a cloud service.

Owner
Oracle Corporation
Licence
GPL or commercial

Strengths

  • Free and very widely supported
  • Simple to operate
  • Large talent pool
  • HeatWave adds analytics

Limitations

  • Enterprise features require paid licence
  • Prices not published
  • Fewer advanced features than PostgreSQL
  • Release track naming has changed
5

MongoDB

Best for: Applications with flexible, JSON-like data models

MongoDB was founded in 2007 (originally as 10gen) by Dwight Merriman, Eliot Horowitz and Kevin Ryan, and reports more than 67,000 customers. It is the best-known document database: data is stored as flexible JSON-like documents rather than fixed tables, which suits evolving application schemas.

MongoDB Atlas, the company's managed cloud service, has a free M0 tier with 512 MB of storage, a Flex tier at US$0.011 per hour capped at US$30 per month with 5 GB of storage, and Dedicated clusters from US$0.08 per hour (US$56.94 per month for an M10) scaling from 10 GB to 4 TB of storage and from 2 to 96 vCPUs. MongoDB Enterprise Advanced is a commercially licensed self-managed edition priced on request, and Community Edition is available at no cost for local development and self-hosting. The trade-offs are that joins and multi-document transactions are less natural than in relational systems, and Dedicated cluster costs rise steeply with larger instances.

Key features

  • Flexible document model
  • Atlas free M0 tier
  • Flex tier capped at US$30/month
  • Dedicated clusters up to 96 vCPUs
  • Self-managed Enterprise Advanced
  • Free Community Edition

Pricing

Atlas Free (M0): US$0, 512 MB. Flex: US$0.011/hour, up to US$30/month, 5 GB. Dedicated: from US$0.08/hour (US$56.94/month), 10 GB-4 TB. Enterprise Advanced: quote. Community: free.

Founded
2007
Founders
Dwight Merriman, Eliot Horowitz, Kevin Ryan
Customers
67,000+ (company figure)
Pricing
Free tier; Dedicated from US$56.94/month

Strengths

  • Flexible schema
  • Managed free tier
  • Published Atlas prices
  • Free Community Edition

Limitations

  • Joins and transactions less natural than SQL
  • Free tier only 512 MB
  • Costs rise quickly at scale
  • Enterprise Advanced quote-only
6

Amazon Aurora

Best for: AWS customers wanting a managed relational database at scale

Amazon Aurora is a relational database service built by AWS that is compatible with MySQL and PostgreSQL, allowing existing databases to be migrated with standard tools. AWS claims up to six times the throughput of standard PostgreSQL and MySQL and offers availability of up to 99.99% in a single region and up to 99.999% across regions. Aurora Serverless scales capacity automatically, and Aurora DSQL is a PostgreSQL-compatible distributed SQL option for globally distributed applications.

Pricing is usage-based. AWS's worked examples for the I/O-Optimized configuration in US East (N. Virginia) use US$0.156 per Aurora capacity unit (ACU) hour for serverless capacity and US$0.225 per GB-month for storage. Provisioned instances, reserved pricing and a standard configuration with I/O charges are also available. Aurora removes most operational work around replication, backups and failover, but it ties workloads to AWS, its performance claims are vendor figures, and compatibility with every community PostgreSQL extension or MySQL feature should be checked before migration.

Key features

  • MySQL and PostgreSQL compatibility
  • Aurora Serverless auto-scaling
  • Aurora DSQL distributed SQL
  • Up to 99.999% multi-region availability
  • I/O-Optimized configuration

Pricing

Usage-based. Examples (I/O-Optimized, US East): US$0.156 per ACU-hour serverless; US$0.225 per GB-month storage. Provisioned and reserved options available.

Provider
Amazon Web Services
Compatibility
MySQL, PostgreSQL
Availability
Up to 99.999% multi-region (AWS figure)

Strengths

  • Managed replication and failover
  • Serverless option
  • Compatible with common engines
  • Multi-region availability

Limitations

  • AWS lock-in
  • Performance claims are vendor figures
  • Costs can exceed self-managed engines
  • Compatibility details need checking
7

Redis

Best for: Low-latency caching and real-time workloads

Redis is an in-memory data structure store used as a cache, message broker, session store and, increasingly, a vector database. Its licensing has changed repeatedly: versions up to 7.2 were BSD-licensed; from March 2024 Redis 7.4 onward moved to dual RSALv2 and SSPLv1 source-available licences; and with Redis 8 the company added the OSI-approved AGPLv3 as a third option. Modules such as RediSearch, RedisJSON, RedisTimeSeries and RedisBloom are now part of Redis Open Source under the same tri-licence. The pricing page references Redis 8.8.

Redis Cloud has a free 30 MB database on shared infrastructure; Essentials from US$0.007 per hour (US$5 per month) covering 250 MB to 100 GB with SSO, role-based access and encryption and up to 99.99% uptime; and Pro from US$0.014 per hour with a US$200 monthly minimum (the first US$200 free), unlimited memory, multiple databases, active-active multi-region replication and up to 99.999% uptime, on AWS, Azure and Google Cloud. Redis is very fast for key-value workloads, but memory-resident data is costly at large volumes, and buyers need to check which of the three licences applies to their use.

Key features

  • In-memory key-value and data structures
  • Search, JSON, time series and probabilistic modules
  • Vector search
  • Active-active multi-region (Pro)
  • Redis Flex RAM-plus-SSD tiering

Pricing

Free: 30 MB. Essentials: from US$0.007/hour (US$5/month), 250 MB-100 GB. Pro: from US$0.014/hour, US$200/month minimum, first US$200 free. Open source under RSALv2, SSPLv1 or AGPLv3.

Licence
RSALv2 / SSPLv1 / AGPLv3 (Redis 8+)
Pricing
Free; Essentials from US$5/month

Strengths

  • Very low latency
  • Low-cost managed entry tier
  • Multi-model modules included
  • Multi-cloud managed service

Limitations

  • Licence history complex
  • Memory costs at scale
  • Pro has US$200 monthly minimum
  • Not a primary system of record for most workloads
8

IBM Db2

Best for: Enterprises with IBM mainframe or Db2 heritage

Db2 is IBM's relational database. The current Db2 12.1 line is sold in three editions: a free Community edition limited to four CPU cores and 16 GB of memory with community support; Standard, limited to 16 cores and 128 GB with enterprise support; and Advanced, with no resource limits, for mission-critical workloads. Standard and Advanced are licensed standalone (perpetual or monthly) or as cartridges (perpetual or subscription).

IBM also offers Db2 as a managed service on IBM Cloud. A perpetually free tier provides a serverless configuration with one schema, 200 MB of storage and five connections, and the Performance plan starts at US$630 per month, billed hourly at US$0.22 or US$0.29 per vCPU plus storage and IOPS charges, scaling from 4 vCPUs and 50 GB to 128 vCPUs and 40 TB. IBM describes these prices as indicative. Db2 is robust and well supported, but its ecosystem and talent pool are smaller than those of PostgreSQL, SQL Server or Oracle outside IBM-centric organisations.

Key features

  • Community edition (4 cores, 16 GB) free
  • Standard (16 cores, 128 GB)
  • Advanced with no resource limits
  • Managed Db2 on IBM Cloud
  • Perpetual, monthly or subscription licensing

Pricing

Community: free (4 cores, 16 GB). Standard, Advanced: licence prices via IBM. Db2 on IBM Cloud: free tier (200 MB); Performance from US$630/month, US$0.22-0.29 per vCPU-hour.

Vendor
IBM
Current version
Db2 12.1
Managed price
From US$630/month (IBM Cloud)

Strengths

  • Free Community edition
  • Unlimited resources on Advanced
  • Flexible licensing models
  • Managed cloud option

Limitations

  • Smaller ecosystem
  • On-premises prices not published
  • Community edition resource-capped
  • Less common in new cloud-native projects
9

MariaDB

Best for: Organisations wanting a community-governed MySQL alternative

MariaDB was started in 2009 by original MySQL developers, with Michael "Monty" Widenius regarded as its founder, and was designed as an enhanced drop-in replacement for MySQL. The MariaDB Foundation, a Delaware non-profit funded by sponsors, states that MariaDB Server will remain free and open source under GPLv2 independent of any commercial entity. The commercial company, MariaDB plc (MariaDB Corporation), is the primary code contributor and sells subscriptions and enterprise features; Widenius serves as its chief technology officer and as a Foundation board member.

MariaDB's commercial Enterprise Platform is sold in three self-managed subscription tiers, Platform, Platform Plus and Platform Ultimate, each priced on request, and a separate MariaDB Cloud service has its own pricing. Compatibility with MySQL has diverged over time, however, so migrations in either direction need testing, and enterprise pricing is not published.

Key features

  • MySQL-compatible SQL
  • GPLv2 Community Server
  • Foundation-governed open source
  • Enterprise Platform subscriptions
  • MariaDB Cloud service

Pricing

Community Server free under GPLv2; Enterprise Platform (Platform, Platform Plus, Platform Ultimate) quote-based; MariaDB Cloud priced separately.

First released
2009
Licence
GPLv2
Pricing
Free; enterprise by quote

Strengths

  • Free and GPL-licensed
  • Familiar to MySQL users
  • Non-profit foundation stewardship
  • Commercial support available

Limitations

  • MySQL compatibility has diverged
  • Enterprise prices unpublished
  • Naming of corporate entity inconsistent
  • Smaller managed-service footprint than MySQL
10

SQLite

Best for: Embedded, mobile, desktop and edge applications

SQLite is an in-process SQL database library: there is no separate server, and a complete database with tables, indexes, triggers and views lives in a single ordinary disk file. Its source code is in the public domain, so it can be used for any purpose without a licence. The file format is cross-platform between 32-bit and 64-bit and big- and little-endian systems, and the project describes it as stable and enduring; the US Library of Congress recommends SQLite files as a storage format.

The project describes SQLite as the most widely deployed database in the world, embedded in billions of devices and applications. Limits are large for an embedded engine: a maximum database size of 281 terabytes and a maximum row size of 1 gigabyte. SQLite is ideal for mobile apps, browsers, desktop software, IoT devices and test environments, and is increasingly used at the edge. It is not designed for high write concurrency from many clients over a network, so it rarely replaces a client-server database for multi-user transactional systems.

Key features

  • Serverless, zero-configuration
  • Single-file database
  • Public-domain source
  • Cross-platform stable file format
  • Up to 281 TB per database

Pricing

Free; public domain.

Licence
Public domain
Max database size
281 TB

Strengths

  • No licence or server required
  • Extremely portable
  • Stable file format
  • Tiny footprint

Limitations

  • Limited write concurrency
  • No built-in network access
  • No user management
  • Not suited to multi-user OLTP
11

CockroachDB

Best for: Globally distributed applications needing strong consistency

CockroachDB, developed by Cockroach Labs, is a distributed SQL database that combines relational consistency and ACID transactions with horizontal scaling: nodes can be added to scale reads and writes, with automatic rebalancing and hotspot detection. It offers a PostgreSQL-compatible SQL interface with migration tools, globally consistent transactions, global writes and data locality controls for placing data near users or within jurisdictions.

The managed service, Cockroach Continuum, is sold in two editions. Standard is priced at US$0.092 per vCPU-hour, about US$203 per month for 2 vCPUs and 100 GB on AWS us-east-1, although the features section quotes a rate from US$0.057; Mission Critical costs US$0.162 per vCPU-hour, about US$1,528 per month for 12 vCPUs and 100 GB. Both come with a 30-day trial and US$400 of credit. A self-hosted edition with enterprise security features is sold through sales. CockroachDB suits applications that must survive regional outages, but its distributed architecture adds latency to some transactions compared with single-node PostgreSQL, and not every PostgreSQL feature or extension is supported.

Key features

  • Distributed ACID transactions
  • PostgreSQL-compatible SQL
  • Automatic rebalancing
  • Data locality controls
  • Global writes

Pricing

Standard: US$0.092/vCPU-hour (about US$203/month for 2 vCPU, 100 GB); Mission Critical: US$0.162/vCPU-hour (about US$1,528/month for 12 vCPU). 30-day trial with US$400 credit. Self-hosted via sales.

Developer
Cockroach Labs
Pricing
From US$0.092/vCPU-hour (Standard)

Strengths

  • Horizontal scaling with strong consistency
  • Multi-region resilience
  • PostgreSQL wire familiarity
  • Published cloud rates

Limitations

  • Higher latency for some transactions
  • Incomplete PostgreSQL feature parity
  • Conflicting rates on pricing page
  • Self-hosted licence via sales
12

Supabase

Best for: Start-ups and app developers wanting a managed Postgres backend

Supabase describes itself as the Postgres development platform: every project is a dedicated PostgreSQL database, optionally combined with authentication (20+ social sign-in providers, email, phone and magic links, with authorisation via row-level security), S3-compatible storage with a CDN, TypeScript edge functions, realtime WebSocket sync, auto-generated APIs and pgvector-based vector search. Its core is open source.

The Free plan includes a 500 MB database per project, a shared CPU with 500 MB of RAM, up to 50,000 monthly active users, 5 GB of egress and 1 GB of file storage, but limits accounts to two active projects, pauses projects after a week of inactivity and excludes automatic backups. Pro starts at US$25 per month and Team at US$599 per month, each including 8 GB of disk per project and then US$0.125 per GB; Enterprise is custom. Supabase gives small teams a full backend quickly while keeping standard PostgreSQL underneath, which eases later migration. Usage-based compute and egress charges need monitoring as applications grow, and the free tier is unsuited to production.

Key features

  • Dedicated PostgreSQL per project
  • Auth with row-level security
  • S3-compatible storage
  • Edge functions
  • Realtime sync
  • pgvector vector search

Pricing

Free: US$0, 500 MB database, 2 active projects. Pro: from US$25/month; Team: from US$599/month (8 GB disk per project, then US$0.125/GB). Enterprise: custom.

Engine
PostgreSQL
Pricing
Free; Pro from US$25/month

Strengths

  • Full backend around standard Postgres
  • Generous free tier for prototypes
  • Open-source core
  • Low Pro entry price

Limitations

  • Free projects pause after inactivity
  • No backups on free plan
  • Usage charges grow with traffic
  • Large jump from Pro to Team
13

Neon

Best for: Developers and AI agent platforms needing on-demand Postgres

Neon is a serverless PostgreSQL service that separates storage from compute so databases can scale to zero and be branched like code. Its pricing page describes it as built around Lakebase Postgres, for developers, start-ups and agent platforms, "from Databricks", reflecting its acquisition by Databricks.

Pricing is usage-based without monthly minimums on paid plans. The Free plan is permanent, requires no card and includes 100 compute-unit (CU) hours per project and 1 GB of storage per project, up to 20 GB across the account. Launch costs US$0.106 per CU-hour and Scale US$0.222 per CU-hour, with storage at US$0.35 per GB-month and no storage cap. Branching and scale-to-zero suit preview environments and AI agent platforms, which Neon names as target users. Its limitations are cold-start latency when compute resumes from zero, a product direction now tied to Databricks' Lakebase strategy, and storage prices higher than raw block storage.

Key features

  • Serverless Postgres with scale-to-zero
  • Database branching
  • Usage-based compute (CU-hours)
  • Free plan with 100 CU-hours per project
  • Unlimited storage on paid plans

Pricing

Free: 100 CU-hours and 1 GB per project (20 GB total). Launch: US$0.106/CU-hour. Scale: US$0.222/CU-hour. Storage US$0.35/GB-month. No monthly minimum.

Owner
Databricks
Engine
PostgreSQL
Pricing
Free; from US$0.106/CU-hour

Strengths

  • Scale-to-zero saves cost
  • Branching for dev and preview environments
  • No monthly minimum
  • Permanent free plan

Limitations

  • Cold starts after idle
  • Storage priced above raw disk
  • Roadmap tied to Databricks
  • Usage billing harder to predict

Comparison table

DatabaseModelLicenceFree optionDeploymentEntry paid priceOwner
Oracle AI DatabaseMulti-model relationalCommercialYes (Free edition)On-prem / cloudNot verifiedOracle
SQL Server 2025RelationalCommercialExpress, DeveloperOn-prem / AzureUS$3,945 per 2 cores (Std 2022)Microsoft
PostgreSQLObject-relationalPostgreSQL LicenceFreeSelf-managed / many cloudsFreeCommunity
MySQLRelationalGPL / commercialCommunitySelf-managed / HeatWaveQuote (Enterprise)Oracle
MongoDBDocumentCommercial / communityAtlas M0Atlas / self-managedUS$56.94/mo (M10)MongoDB
Amazon AuroraRelational (MySQL/PG)Proprietary serviceNoAWSUS$0.156/ACU-hr exampleAWS
RedisIn-memory key-valueRSALv2/SSPLv1/AGPLv330 MB cloudSelf-managed / Redis CloudUS$5/moRedis
IBM Db2RelationalCommercialCommunity (4 cores)On-prem / IBM CloudUS$630/mo (cloud)IBM
MariaDBRelationalGPLv2 / commercialCommunitySelf-managed / cloudQuoteMariaDB plc
SQLiteEmbedded relationalPublic domainFreeIn-processFreeSQLite project
CockroachDBDistributed SQLCommercialTrial creditCloud / self-hostedUS$0.092/vCPU-hrCockroach Labs
SupabasePostgres platformOpen-source core500 MBCloudUS$25/moSupabase
NeonServerless PostgresProprietary service100 CU-hrsCloudUS$0.106/CU-hrDatabricks

Alternatives by use case

  • Mission-critical enterprise OLTP: Oracle AI Database, SQL Server Enterprise, IBM Db2 Advanced.
  • Free general-purpose relational: PostgreSQL, MySQL, MariaDB.
  • Managed relational on AWS: Amazon Aurora.
  • Document data: MongoDB Atlas.
  • Caching and real-time: Redis.
  • Global consistency: CockroachDB, Aurora DSQL.
  • App backend in a box: Supabase.
  • Ephemeral and branched databases: Neon.
  • Embedded and mobile: SQLite.

Recent developments

  • 2025: SQL Server 2025 (17.x) introduces vector data types, native JSON, regular expressions and Fabric mirroring, and discontinues Web edition, Master Data Services and Data Quality Services.
  • 2026: Oracle's flagship database is released as Oracle AI Database 26ai; the New Features guide is updated on 20 October 2026.
  • 12 November 2026: PostgreSQL 18.7 and updates to versions 14-17 are released, with PostgreSQL 19 at release-candidate stage.
  • 2025-2026: Redis 8 adds the AGPLv3 open-source licence option alongside RSALv2 and SSPLv1.
  • 2026: Neon is marketed as Lakebase Postgres from Databricks.
  • 2026: Cockroach Labs prices its managed service as Cockroach Continuum in Standard and Mission Critical editions.

Frequently asked questions

What is the most popular database software?

Popularity depends on the measure used. Oracle Database, MySQL, Microsoft SQL Server and PostgreSQL are the established client-server relational engines, MongoDB reports more than 67,000 customers for its document database, and Redis is widely used as an in-memory store. SQLite's developers describe it as the most widely deployed database overall because it is embedded in billions of devices and applications.

What is the best free database?

PostgreSQL is free under a permissive licence and has a broad feature set and extension ecosystem. MySQL Community and MariaDB are free under the GPL, and SQLite is public domain for embedded use. Commercial vendors also offer free editions: SQL Server Express and Developer, Oracle AI Database Free and Db2 Community.

What is the difference between SQL and NoSQL databases?

SQL databases store data in tables with fixed schemas, support joins and enforce ACID transactions. NoSQL databases use other models, such as documents in MongoDB or key-value structures in Redis, and favour flexible schemas or very low latency. Many systems now blur the line, for example PostgreSQL's JSONB and Oracle's JSON Relational Duality.

How much does SQL Server cost?

Microsoft's SQL Server 2022 list prices are US$15,123 per two-core pack for Enterprise and US$3,945 per two-core pack for Standard, or US$989 per server plus US$230 per client access licence. Subscriptions and Azure Arc pay-as-you-go pricing are also available, and the Express and Developer editions are free.

Is PostgreSQL better than MySQL?

Each suits different needs. PostgreSQL offers richer data types, more index types, all SQL isolation levels and extensibility through extensions such as PostGIS, under a permissive licence. MySQL is simpler, very widely hosted and dual-licensed under the GPL or a commercial Oracle licence. Many new projects choose PostgreSQL; many existing web applications run on MySQL.

Which databases support vector search?

SQL Server 2025 adds a vector data type, vector functions and a preview vector index; Oracle AI Database 26ai includes AI Vector Search; Redis includes vector search; and PostgreSQL supports it through the pgvector extension, which Supabase includes in every project.

What is a serverless database?

A serverless database scales compute automatically and bills for capacity actually used, sometimes scaling to zero when idle. Examples include Aurora Serverless, priced per Aurora capacity unit hour, and Neon, priced per compute-unit hour with a permanent free plan. The trade-off is possible cold-start latency when an idle database resumes.

See also

Should your company be on this list?

Submit your logo, description, image and link for review. Approved listings appear in the Featured section above.

Add your company

Is your company missing from a list?

Send your logo, description and link. Editors review every submission, and approved listings are clearly labelled.

Add your company