MariaDB adoption is growing, and the way people get it is changing – Adoption Index 2026-09
The MariaDB Adoption Index is up 126% in a year. More people are downloading MariaDB, more people are contributing to it, and more projects are building on it than at any point since we started measuring.
At the same time, the numbers show something else: the way people acquire and discover MariaDB is changing. Some of our older metrics are falling – not because MariaDB is, but because the channels they measure are.
The index combines 24 metrics – downloads, contributors, community, social – into one monthly number, baseline January 2024. Background in the original post.
1. Adoption is up across the board
| Aug 2025 | Aug 2026 | YoY | |
|---|---|---|---|
| mariadb.org downloads | 99,496 | 526,572 | +429% |
| Docker official image pulls | 18.7M | 25.3M | +35% |
| GitHub READMEs mentioning MariaDB | 174,123 | 314,637 | +81% |
| Unique external PR contributors, 12 months | 64 | 115 | +80% |
| Returning external contributors | 5 | 16 | +220% |
| GitHub stars | 6,345 | 8,161 | +29% |
| Reddit subscribers | 2,951 | 3,978 | +35% |
| LinkedIn followers, Foundation | 4,866 | 6,579 | +35% |
Downloads from mariadb.org have grown fivefold. Docker pulls are at 25 million a month. The number of GitHub projects that mention MariaDB in their README is up 81% – 314,637 projects now say so. External contributors to MariaDB Server have nearly doubled, and the number who come back and contribute again has tripled.
Downloads are now the largest part of the index, so it is worth saying: without them, the index is still up 53%. Every growth metric points the same way.
Why downloads are up fivefold, our current theory is that it may have something to do with MySQL 8.0 reaching EOL and Magento choosing MariaDB as default. There may be other trends in motion too. We are running our yearly survey at the moment – please participate so we can share a broader picture of what is going on.
2. How people get MariaDB is changing: packages to containers
| Aug 2025 | Aug 2026 | YoY | |
|---|---|---|---|
| Docker official image pulls | 18.7M | 25.3M | +35% |
| mariadb.org downloads | 99,496 | 526,572 | +429% |
| Debian popcon (machines reporting the package) | 457,244 | 367,002 | −20% |
| Homebrew installs, 30 days | 5,107 | 3,940 | −23% |
Debian’s popcon and brew install mariadb are both down about a fifth. Docker and direct downloads are up far more than that. This is not MariaDB losing users – it is developers changing how they work. Docker usage jumped 17 points to 71% in the 2025 Stack Overflow Developer Survey, the largest single-year gain of any technology, and Docker’s own 2025 State of Application Development report found the share of developers working primarily in a local environment fell from 64% to 36% in a year. The developer who used to brew install a local MariaDB now runs docker run mariadb – or does not run it locally at all.
Last year’s State of MariaDB survey already had 34% of respondents on Docker against 40% on distribution packages. That balance has likely tipped since. Popcon was a good adoption proxy when Debian packages were how MariaDB reached machines; it is becoming a proxy for one shrinking channel.
3. How people discover MariaDB is changing: Q&A to AI
| Dec 2023 | Aug 2026 | |
|---|---|---|
| Stack Exchange new MariaDB questions, per month | 84 | 7 |
| Google Trends | 42 | 32 |
| Wikipedia views, all languages | 28,765 | 20,046 |
New MariaDB questions on Stack Exchange have gone from 84 a month to 7. Search and Wikipedia are down too. None of this is specific to MariaDB – Stack Overflow’s overall question volume has collapsed since 2023, and Wikipedia’s traffic has declined. The questions have not stopped; they have moved to AI assistants, where nobody can count them yet.
The metrics that measure people doing something with MariaDB – contributing, starring, building on it, joining the community – are all up. The current available metrics measure people asking about it in public are down. That is a change in where the conversation happens, not in whether it happens.
We are still restricted in knowing exactly what users are asking AI about MariaDB, though we have made some internal tests to see how well LLMs know MariaDB. We are watching this space if somebody starts sharing stats on what users are asking AI.
The numbers
What we want to measure next
The index measures attention and acquisition well. Adoption moves through four layers – running, available, chosen, visible – and we want to see more of the ones the index only sees partly:
- Running. Deployment, not downloads. WordPress.org publishes the database split across its installs – MariaDB passed MySQL there last year. It measures machines actually running MariaDB, and it will not fall when people stop asking questions on Stackexchange and anywhere else.
- Available. Distribution points: managed-service listings, hoster defaults, and applications that name MariaDB as their default. A count that only moves when a provider adds or drops MariaDB.
- Say the name. We already count READMEs mentioning MariaDB, up 81% in a year. What we do not count is how many projects running MariaDB call it MySQL. We originally left out comparisons to focus on MariaDB’s metrics, but “say the name” is a MariaDB problem, and it may deserve some kind of metric.
As the channels change, so will the index. Adoption is growing; we want to keep measuring it where it happens.
