Relational Databases Software Market

When Amazon Web Services made Aurora DSQL generally available on 27 May 2025, it sold a relational database that no customer would ever install, patch or size, and that single launch captures where the money in this market is going. Douglas Insights values relational databases software at USD 88,400.0 million in 2025, rising to USD 252,337.8 million by 2035 at 11.06% a year. Our build counts roughly 2.6 million organisations paying for relational database engines, licences, support or managed database services, at an average annual spend of USD 34,000 each. Paying organisations grow 3.6% a year, while spend per organisation rises 7.2% a year as workloads move to consumption priced cloud services and data volumes climb. The analysis sits within our enterprise software coverage and follows the Douglas Insights methodology.

Who controls relational database software spending?

Relational database software spending is controlled by a small group of companies: Douglas Insights estimates the top three, Microsoft, Oracle and Amazon Web Services, take about 60% of the USD 88,400.0 million spent in 2025, and the balance of power has shifted from the vendors who wrote the classic engines to the cloud providers who run them. For thirty years, relational database spending meant licences and support contracts from Oracle, Microsoft, IBM and SAP, installed on servers inside the customer’s data centre. That business still exists and still earns very high margins, but the growth now sits with Amazon Web Services, Microsoft Azure and Google Cloud, which rent managed relational databases by the hour, by the gigabyte or by the request. The three hyperscalers compete partly on their own engines, such as Aurora, Azure SQL, Spanner and AlloyDB, and partly on hosting open source engines, above all PostgreSQL and MySQL, which carry no licence fee but generate large service revenue.

Aurora DSQL illustrates the new competitive weapon. AWS previewed it at re:Invent in December 2024 and launched it in eight regions in May 2025, promising a distributed SQL database with 99.999% multi-region availability and PostgreSQL compatibility, with no instances to provision. Customers such as ADP and Robinhood were named at launch. Features of this kind were once sold by specialist distributed SQL vendors; bundling them into a cloud platform shows how the hyperscalers absorb innovation from the edges of the market. The traditional vendors have responded by moving their own engines into every cloud: Oracle now runs its database inside Microsoft Azure, Google Cloud and AWS data centres, so customers can keep Oracle while buying through their cloud contract.

Is relational database software being replaced by NoSQL?

Relational database software is not being replaced, because it solves the problem most businesses actually have, recording transactions correctly every time, which is why relational engines still take about three quarters of all database software spending in our 2025 estimate. A relational database stores data in tables with defined relationships and guarantees that a payment, an order or an inventory change is either fully recorded or not recorded at all, even if a server fails mid-transaction. During the 2010s, NoSQL databases promised to replace this model for web scale applications, and document, key-value and graph databases did take important workloads. Yet relational engines kept their central role in finance, retail, healthcare, government and enterprise applications, because rewriting systems of record is risky and because the relational engines themselves absorbed the ideas that made NoSQL attractive, such as JSON documents, horizontal scaling and flexible schemas. PostgreSQL in particular has become the default choice for new applications, used by startups and banks alike. The result is a mature technology whose spending still grows faster than most enterprise software categories, because data volumes grow and because cloud consumption pricing captures that growth directly.

How does buying change when databases move to the cloud?

Relational database software buying stops being a licence negotiated every few years and becomes a meter that runs every hour, which changes who buys, how much is spent and how quickly spending grows. Under the licence model, an organisation paid for processor cores or named users, plus roughly a fifth of the licence price each year in support, and its costs moved only when it bought more hardware. Under the managed service model, it pays for compute capacity, storage, input and output operations and backups as they are consumed. Developers can create a database in minutes without procurement, so databases multiply. Spending follows data growth and application traffic automatically. For vendors, this produces faster revenue growth than licence renewals ever did; for buyers, it produces convenience and, frequently, surprise bills. Our spend per organisation leg of 7.2% a year reflects this: it is not price inflation in the usual sense but the meter capturing more usage, more replicas, more regions and more premium features per customer. As database fleets multiply, many teams add monitoring tools of the kind covered in our Enterprise Data Observability Platforms Market report to keep track of them.

What is driving relational database software growth?

Relational database software growth is driven first by data growth inside transactional systems, which Douglas Insights estimates adds 5 to 6 points a year to spend. Digital payments, e-commerce orders, connected devices and application logs add rows faster than old data is archived, and managed cloud databases bill directly for the storage and input and output this creates. We model stored relational data at leading customers growing around 20% a year, which translates into roughly 5 to 6 points of spend growth once storage price declines are netted off.

The second driver is cloud migration. We estimate that about 55% of relational database workloads by spend still run on self-managed infrastructure in 2025, falling to around 30% by 2035. Each migrated workload typically moves from a fixed licence and support cost to a managed service bill, and although migration can reduce total cost of ownership, it usually increases the revenue captured by database providers because infrastructure and operations are bundled into the price. Migration also moves in large blocks: a single bank shifting its core ledger from an on-premise Oracle cluster to a managed service can move several million dollars a year from licences and hardware into database service revenue, and programmes of that size typically run for two to four years. That is why the managed services share of relational database software climbs by around 2 points a year rather than in one step.

The third driver is resilience requirements. Regulators and customers expect critical services to survive regional outages. Aurora DSQL’s promise of 99.999% multi-region availability, equal to about five minutes of downtime a year, is a response to that demand, and multi-region deployments multiply database capacity and cost for the same application, often by two to three times.

The fourth driver is artificial intelligence. Relational engines now add vector search, so that applications can combine customer records with embeddings used by language models, and PostgreSQL’s pgvector extension has become widely used for this purpose. We count AI related features, retrieval workloads and associated storage as adding around 1 point a year to spend per organisation through the forecast. The same AI build out is reshaping data centre hardware, as our Data Centre Liquid Cooling Market report shows.

What could slow relational database software down?

Relational database software growth could be slowed first by cost control. After the rapid cloud growth of 2020 to 2022, many organisations began optimising cloud spend in 2023, rightsizing instances, committing to reserved capacity and shutting idle databases. We assume optimisation removes about 1.5 points a year from spend per organisation growth that would otherwise occur, and a harder optimisation cycle is the main downside risk.

The second restraint is open source substitution. Every workload moved from a commercial licence to PostgreSQL removes licence revenue that can be worth tens of thousands of dollars per server core. We model open source engines rising from around 40% of new relational deployments to more than 55% by 2035, a shift that grows the market in managed services while shrinking commercial licence revenue.

The third restraint is migration difficulty. Moving a large Oracle or Db2 system with thousands of stored procedures to another engine can take one to three years and cost millions of dollars, which slows both cloud migration and open source substitution. That friction protects incumbents’ revenue but also delays growth in the fastest segments.

The fourth restraint is sovereignty and regulation. Rules on where data may be stored, particularly in Europe, China and India, limit the use of global cloud services for some workloads and favour local providers or self-managed deployments. We estimate that around 15% of regulated workloads face such constraints.

Which deployment segments carry the most value?

Within relational database software, cloud managed relational database services lead with 52% of 2025 value, USD 45,968.0 million. This covers Amazon RDS and Aurora, Azure SQL Database and managed instances, Google Cloud SQL, AlloyDB and Spanner, Oracle’s cloud database services and hosted open source engines. It is where new applications land by default and where migrated workloads end up, and we expect its share to exceed 70% by 2035.

Commercial on-premise licences and support hold 34%, USD 30,056.0 million. This is the installed base of Oracle Database, Microsoft SQL Server, IBM Db2, SAP HANA and similar engines running in customer data centres. It grows slowly, in low single digits, and support revenue is extremely profitable because customers running critical systems rarely stop paying.

Open source subscriptions and support account for 8%, USD 7,072.0 million. Organisations running PostgreSQL, MySQL and MariaDB themselves still pay for enterprise distributions, support, tooling and consulting from companies such as EDB and Percona, and from Oracle for MySQL Enterprise.

Distributed SQL and serverless databases contribute 6%, USD 5,304.0 million, and grow fastest. This includes Google Spanner, CockroachDB, YugabyteDB, TiDB and newer serverless services such as Aurora DSQL, which spread data across many servers and regions while keeping SQL and transactional guarantees.

Which other lines do buyers draw?

Beyond deployment, buyers segment relational database software in three further ways, and each changes the forecast. By engine origin, proprietary engines still earn most of the money, but open source based engines, including hyperscaler services built on PostgreSQL and MySQL, take the majority of new deployments. By workload, online transaction processing remains the core of spending, while hybrid transactional and analytical processing, which runs reports directly on live data, and vector search for AI are the growing edges. By buyer size, large enterprises account for roughly two thirds of spend, while mid-sized organisations, and small organisations and startups, are the fastest growing buyers because managed services removed the need for specialist database administrators; a startup can run a production database for under USD 50 a month.

Where is database spending concentrated?

North America leads relational database software spending with 46% of 2025 value, USD 40,664.0 million, growing 10.08% a year. The United States is home to the three largest cloud providers and to the heaviest enterprise users of commercial databases, and its growth is slightly slower than the global rate only because migration is further advanced.

Europe holds 24%, USD 21,216.0 million, growing 10.8%. Data protection rules and sovereignty concerns slowed early cloud adoption, but sovereign cloud offerings and Oracle’s multicloud arrangements are now accelerating migration.

Asia Pacific also holds 24%, USD 21,216.0 million, and grows fastest at 12.8%, driven by digital payments and e-commerce in India and Southeast Asia, and by domestic database vendors and cloud providers in China, such as Alibaba Cloud and PingCAP, which serve a market largely closed to Western cloud services. Latin America contributes USD 2,652.0 million at 11.6%, led by Brazil’s banking and fintech sector, the Middle East USD 1,944.8 million at 12.4%, where Gulf states are building national cloud regions, and Africa USD 707.2 million at 11.0%.

Who are the main vendors and what are they betting on?

Relational database software is concentrated: Douglas Insights estimates the top three vendors hold about 60% of 2025 spending, based on vendor database revenue disclosures and cloud database service estimates. Oracle remains the largest commercial vendor of relational database licences and is betting on multicloud: running Oracle Database and Exadata inside the three rival clouds so customers never have to leave Oracle to use Azure, Google Cloud or AWS. Microsoft sells SQL Server licences and Azure SQL services and benefits from customers already committed to its software. Amazon Web Services leads managed databases through RDS and Aurora, and uses launches such as Aurora DSQL to compete with specialist vendors. Google Cloud competes on Spanner and AlloyDB and on analytics integration. IBM and SAP serve large installed bases through Db2 and HANA. Alibaba Cloud, Huawei and PingCAP lead in China, while EDB, Percona, Cockroach Labs and Yugabyte serve open source and distributed SQL users.

How are relational databases priced?

Relational database software spend averages USD 34,000 per paying organisation in 2025, but pricing models differ by segment. Commercial licences are typically priced per processor core, with enterprise editions costing tens of thousands of dollars per core and support adding about a fifth of that each year. Managed services charge for compute by the hour, storage by the gigabyte-month, and input and output, backups and data transfer separately; a small managed database can cost under USD 50 a month while large production clusters cost hundreds of thousands a year. Serverless services such as Aurora DSQL charge by distributed processing unit and storage, with a free tier covering the first 100,000 processing units and 1 gigabyte of storage each month. Reserved capacity commitments reduce prices by a third or more. The mix shift toward managed and multi-region deployments raises spend per organisation.

What do the scenarios imply for relational database software by 2035?

The Douglas Insights base case for relational database software, with 3.6% growth in paying organisations and 7.2% growth in spend per organisation, reaches USD 252,337.8 million in 2035. In a cost-discipline case, where organisations optimise aggressively and open source substitution erodes commercial licences faster, the legs fall to 2.0% and 5.4%, leaving the market near USD 182,330 million. In a data-surge case, where AI retrieval workloads and multi-region resilience requirements expand quickly, the legs rise to 4.8% and 8.6%, carrying the market past USD 322,350 million. The Aurora DSQL launch sits at the centre of this range: if serverless, distributed SQL becomes the default for new applications, spend per organisation follows usage rather than licences and the upside case becomes more likely. Each 1 point change in organisation growth moves the 2035 figure by roughly USD 24,360 million.

Which rules shape database choices?

Relational database software choices are shaped by three sets of rules. Data protection and localisation laws, including Europe’s General Data Protection Regulation and data residency requirements in India, China and the Gulf, determine where data can be stored and favour local cloud regions. Financial sector resilience rules, such as the European Union’s Digital Operational Resilience Act, Regulation (EU) 2022/2554, which applied from 17 January 2025, require banks and insurers to manage risks from technology providers, including critical cloud databases, and to plan for outages and exits. Licensing terms are the third: changes to how commercial vendors license software in virtualised and cloud environments can shift costs by millions of dollars and prompt audits and migrations.

Is PostgreSQL the new default?

PostgreSQL is increasingly the default engine in relational database software. PostgreSQL has become the engine most developers choose for new applications, the basis for managed services from every major cloud provider, and the foundation of several distributed SQL products, including Aurora DSQL’s compatibility layer. Its permissive licence means no single company controls it, and its extension system lets vendors add capabilities such as vector search without forking the engine. This does not mean commercial engines disappear: systems built on Oracle and SQL Server will run for decades. But the centre of gravity for new spending has moved to PostgreSQL compatible services, where revenue comes from running the database rather than licensing it. We treat PostgreSQL compatibility as the most important product decision vendors make over the forecast.

Douglas Exclusive: the workload migration ledger

The Douglas Insights workload migration ledger follows relational database software workloads by industry and region from self-managed licences to managed services and from commercial to open source engines, recording migration timelines, cost changes and the resulting shifts in vendor revenue. It is the basis for the spend per organisation leg in this forecast.

What methodology sits behind the USD 88.4 billion estimate?

The relational database software estimate rests on a bottom-up count of about 2.6 million paying organisations. We counted paying organisations by size, industry and region, estimated relational database spend per organisation from vendor disclosures, cloud provider reporting and pricing, and split spend across deployment models and engines. Non-relational databases, data warehouses sold as standalone analytics platforms, database tooling sold separately and IT services are excluded. The full build, sources and confidence grades are set out in the fact sheet above.