Ever opened a database vendor’s pricing page, scrolled for ten minutes, and still had no idea what the bill would be? That confusion is common. Database pricing is one of the murkiest corners of software buying, because the sticker price rarely reflects what a team ends up paying. Between licensing models, compute hours, storage tiers, and support contracts, the real number hides in footnotes.
This comparison breaks down how the major database management tools price their products, where the hidden costs live, and how to build a realistic budget before you sign anything. The goal is simple: fewer surprises on the invoice.
Why database pricing is harder than it looks
A database is not a single purchase. It is a license, an operating environment, a backup strategy, and a person who knows how to run it. Each of those carries a cost.
The biggest trap is comparing list prices across models that do not measure the same thing. A per-core license, a per-hour cloud instance, and a free open-source download can look like apples and oranges, because they are. A free license still requires servers, storage, monitoring, and a database administrator’s salary. A cheap cloud instance gets expensive fast when your query volume climbs.
Total cost of ownership is the only fair comparison. That means adding licensing, infrastructure, staff time, migration effort, and the cost of getting locked into a platform you cannot easily leave.
Open-source databases: free license, real costs
MySQL Community Edition is free under the GNU GPL v2. PostgreSQL is free under its own permissive, BSD-style PostgreSQL License. Neither asks for a licensing fee, which makes them the default starting point for cost-conscious teams.
But “free” describes the license, not the bill. What you pay for instead:
- Managed hosting. Running PostgreSQL on a cloud provider’s managed service turns a free license into a monthly consumption bill.
- Commercial support. Vendors sell support subscriptions for open-source databases, typically priced by server count or support tier.
- Commercial editions. Oracle sells paid MySQL editions and support alongside the free Community version, priced by quote rather than a stable public list.
- Operational labor. Someone has to patch, back up, tune, and monitor the thing.
The license model itself has been stable for years. Oracle acquired MySQL in 2010 and has kept the open-core-plus-support structure ever since.
Commercial relational databases: the licensing maze

Once you move to SQL Server or Oracle Database, pricing shifts from “free” to “read the contract carefully.”
Microsoft SQL Server
SQL Server has been licensed by core since SQL Server 2012, replacing the older per-socket model. Standard Edition caps at 24 cores per instance and 4 sockets as of the 2019 and 2022 documentation. That cap matters: a big server can force you into Enterprise Edition pricing whether you wanted it or not.
Core-based licensing also means virtualized environments need careful counting. Many teams discover mid-negotiation that their VMware cluster technically requires far more licenses than the physical host they assumed.
Oracle Database
Oracle licenses by processor core or by Named User Plus, and applies a core-factor multiplier to certain processors. That multiplier is where budgets go sideways. A processor that counts as one core in your head may count as several on Oracle’s price list.
Oracle publishes a Technology Price List, but the exact multipliers and list prices should always come from the current version of that document, not from a blog post or a reseller’s summary.
Managed cloud databases: consumption is the price
AWS RDS, Azure SQL Database, and Google Cloud SQL all bill by consumption. The components are broadly similar:
| Cost component | What drives it |
|---|---|
| Instance size | vCPUs and memory of the database instance |
| Storage | Provisioned or consumed gigabytes |
| I/O | Read and write operations, especially on high-throughput workloads |
| Backup retention | How long you keep automated backups |
| Data transfer | Moving data out of the region or provider |
Azure SQL adds serverless and vCore purchasing models on top, which lets you pause compute during idle periods. That can cut costs dramatically for development and low-traffic workloads, and just as dramatically confuse a forecast built on always-on assumptions.
The practical rule: cloud list prices change often. Rather than quoting a figure that will be stale in a month, cite the provider’s own pricing documentation with a “checked on” date.
NoSQL and data stores: license changes that move the bill
Licensing shifts in the NoSQL world have real financial consequences, and they are easy to miss.
MongoDB moved its Community Server to the Server Side Public License (SSPL) in October 2018. Several Linux distributions, including Debian, dropped it from default repositories as a result. MongoDB Atlas is priced by cluster tier and consumption.
Redis changed its license in March 2024, moving from BSD to a dual RSALv2 and SSPLv1 model. The Linux Foundation responded with the Valkey fork the same month. Elasticsearch adopted the Elastic License 2.0 in 2021, prompting the OpenSearch fork.
Why does this matter for pricing? Because a license change can force a migration, and migrations are expensive. A database that was free last year may now require either a paid license or a re-platforming project. When comparing data stores, weight vendor lock-in risk as a line item.
A practical cost comparison by model
Here is how the main pricing models stack up when you look past the headline number.
| Model | Upfront license | Ongoing cost | Best fit |
|---|---|---|---|
| Open source, self-managed | None | Infrastructure and staff | Teams with strong ops skills |
| Open source, managed | None | Monthly consumption | Teams wanting low admin load |
| Commercial, on-premises | Core or user based | Support and renewal | Regulated, high-utilization workloads |
| Cloud managed | None | Pay-as-you-go | Variable or fast-growing demand |
| NoSQL managed | None | Cluster tier and usage | Document or key-value workloads |
Cloud pay-as-you-go can exceed on-premises licensing at steady high utilization, but the crossover point is workload-specific. Any single “cloud is cheaper after X” claim should be treated as uncertain without a cited study for your exact pattern.
Hidden costs that blow up database budgets
The line items that wreck forecasts are rarely the license itself. Watch for these:
- Non-production environments. Dev, staging, and QA often need the same licenses as production. Many contracts let you discount or defer these, but only if you ask.
- High availability and disaster recovery. A standby replica can double your license count or compute spend.
- Backup storage growth. Retention policies quietly accumulate terabytes.
- Egress and data transfer. Moving data between regions or out of a cloud provider is billed separately and adds up fast.
- Migration and re-platforming. Moving between database engines is a project with a real cost, not a switch you flip.
- Free tiers that expire. Cloud free tiers are time- or usage-limited and should never be modeled as ongoing zero-cost options.
A useful discipline: build the budget from four buckets, not one. Licensing, infrastructure, people, and migration. Teams that only budget the first bucket are the ones surprised in month three.
Practical tips that actually work
Model three scenarios, not one. Build a low, expected, and high growth forecast for compute and storage. Consumption pricing punishes optimistic single-line projections.
Ask for the renewal clause in writing. Discounts in year one often evaporate at renewal. Get the multi-year trajectory.
Count cores before you negotiate. For SQL Server and Oracle, an accurate core count (including virtualization rules and core factors) is your strongest negotiating tool. Getting it wrong in either direction costs money.
Test the exit. Before committing to a managed platform, estimate what it would cost to leave. If the answer is “a year of engineering,” price that risk in.
Watch the license changelog. For open-source data stores, follow license announcements. The Redis and MongoDB changes show how quickly a “free” component can become a budget line.
Broader cloud and infrastructure decisions interact with database spend more than most teams expect. If you are also evaluating the surrounding stack, it helps to see how cloud cost management tools that actually cut your bill handle the same consumption-pricing problem, and how multi cloud management tools that actually reduce complexity keep costs visible across providers. Teams running databases alongside containerized workloads often pair this with a look at function as service tools practical comparison 2025, since serverless compute changes how you size the database behind it.
Security tooling is another adjacent cost that database buyers forget to budget. If your database sits behind privileged access controls, the privileged access management pricing buyers actually pay breakdown shows how those subscriptions are structured, and cloud security posture management pricing buyers actually covers the monitoring layer that often gets bundled into compliance requirements.
Common mistakes to avoid
The most frequent error is comparing a free open-source license to a commercial one without adding operational cost. The second is assuming cloud means no licensing thinking at all; cloud bills are licensing decisions in disguise, just measured in hours instead of cores.
A third mistake: ignoring the cost of the people. A database administrator is not free, and a managed service that removes DBA toil can pay for itself even at a higher monthly rate. Run the math on staff time before rejecting a managed option on price alone.
Finally, teams routinely forget that support contracts renew. A support subscription that looked reasonable in year one is a recurring line item, and it usually escalates.
Conclusion
Database pricing resists simple comparison because the industry sells five different things under one label: a license, a hosted service, a support contract, a consumption meter, and an operational burden. The teams that budget well do three things. They compare total cost of ownership rather than list price. They model consumption scenarios instead of a single forecast. And they treat license changes and vendor lock-in as financial risks, not legal trivia.
Start with your workload profile. Steady, high-utilization, and regulated points toward commercial licensing with careful core counting. Variable, fast-growing, or small-team points toward managed open source. Either way, build the budget from licensing, infrastructure, people, and migration, and re-verify every price against the vendor’s own documentation before you sign. The number on the pricing page is a starting point, never the answer.
FAQ
Is PostgreSQL really free to use commercially?
Yes. PostgreSQL is distributed under the PostgreSQL License, a permissive BSD-style license that allows commercial use without a fee. The cost comes from hosting, support subscriptions if you buy them, and the staff time to operate it.
Why is SQL Server priced by core instead of by server?
Microsoft moved to core-based licensing with SQL Server 2012, replacing per-socket licensing. It aligns cost more closely with compute power consumed. Standard Edition caps at 24 cores per instance and 4 sockets as of the 2019 and 2022 documentation, which can push large servers toward Enterprise pricing.
What is an Oracle core-factor multiplier?
Oracle licenses by processor core or Named User Plus and applies a multiplier to certain processors, so one physical chip may count as multiple cores for licensing purposes. Always confirm the current multiplier in Oracle’s own price list rather than a secondary source.
Can a free cloud database tier replace a paid one long term?
Usually not. Free tiers are time- or usage-limited, and their terms change without notice. They are useful for prototyping and learning, but they should never be modeled as an ongoing zero-cost production option.
How did the Redis and MongoDB license changes affect costs?
MongoDB moved Community Server to the SSPL in October 2018, and Redis shifted from BSD to a dual RSALv2 and SSPLv1 model in March 2024. Both changes pushed some users toward forks or paid offerings, adding migration and licensing costs that a simple “free versus paid” comparison misses.
How do I estimate total cost of ownership for a database?
Add four buckets: licensing or subscription fees, infrastructure and storage, staff time to operate and secure it, and the cost of migrating in or out. Model low, expected, and high growth scenarios for consumption-based pricing, and get renewal terms in writing before committing.

