Before the numbers, Bruno explained the nature of the data. Quarkus build analytics are opt-in only. The first time a developer uses dev mode, they are prompted to contribute anonymous build-time data. If they agree, each subsequent build or dev-mode execution sends a small JSON payload.

What is collected:

What is explicitly not collected: source code, IP addresses, or any other personal data.

The analytics file sent is stored locally at target/quarkus-build-analytics.json, so developers can inspect exactly what is being shared. To check or change the setting, the local config lives at ~/.redhat/ on your home directory — a disabled: false entry means analytics are on.

The team noted a deliberate gap: only extensions from known group IDs (the core Quarkus and a handful of community namespaces) are tracked. Internal or corporate extensions are never reported, to avoid inadvertently revealing what proprietary components a company is using. Community extensions wanting to be included can contact the team.

Because the data is opt-in, the absolute numbers are never shown publicly — the team looks at relative growth and proportions. Maven download stats (opt-out rather than opt-in) show higher absolute numbers, but the two data sources tend to tell consistent stories about trends.

The analytics page at quarkus.io/usage has more detail on what is collected and why.

The headline numbers for May 2026 vs May 2025:

Max observed that AI-assisted development likely explains part of the shift: developers using agents to write and run code may run fewer manual start-stop-start dev cycles, but still trigger dependency downloads and CI builds. The build analytics capture the latter but undercount the former.

The most eye-catching slide: Windows 11 is now the most-used OS among Quarkus developers, overtaking macOS which held the top spot last year. Linux slipped slightly.

Bruno attributed this to two overlapping factors: a large migration from Windows 10 to Windows 11, and what looks like growing enterprise adoption — organizations where developers work on locked-down corporate Windows laptops. Max noted that in most developer tooling data over the years, Windows has consistently been around 30-40%, macOS around 30%, and Linux the remainder, so seeing Windows at the top is unusual and significant for prioritization decisions.

The data reinforces that Windows support is not optional for Quarkus.

Time zone is used as a geographic proxy. The top three cities by build activity were São Paulo (Brazil), Germany (multiple cities), and Kolkata (India). Bruno noted that China is growing fast — roughly doubling year-over-year — though time zone data for China is complicated because the country spans multiple natural time zones but uses a single official one, meaning a large number of users are concentrated in a single time zone.

The broader trend is clear: Quarkus is no longer primarily a European project. Latin America, India, and China are all growing, and activity in Bangkok, Mexico City, and Bogotá is notable — Bogotá now has more build activity in the last 30 days than London.

At the other end, the least-active time zones include small Pacific islands, Malta, San Marino, and — to some amusement on the stream — Perth, Australia.

Java 25, released in September 2025, already accounts for 22% of all Quarkus builds — a rate faster than any previous Java version adoption. Java 21 and Java 25 together now represent around 76% of all builds.

This is directly relevant to Quarkus 4, which will require Java 21 or later. The data suggests the community is already well-positioned for that transition. Java 17 adoption has roughly halved since last year and continues to fall.

Max attributed the acceleration to a combination of factors: the painful breaking changes that accompanied older version jumps are largely behind the ecosystem, and Java 25 offers practical, visible benefits (notably performance) that make upgrading feel worthwhile rather than just necessary. The Quarkus team did a significant amount of work across two or three releases to handle Java 25’s stricter warnings around native memory access, with extensions now able to declare the JVM flags they need so Quarkus can set them automatically at launch.

Maven remains dominant at 85% of builds, with Maven 3.9 being the most-used version. Maven 4.x is beginning to appear. Gradle sits at roughly 15%, consistent with what the code.quarkus.io project-creation data shows — the two data sources agree, which Bruno took as a sign the analytics are representative.

Maven’s share actually grew slightly (+2.5 percentage points) year-over-year. The team’s default wrapper version being Maven 3.9 likely contributes to that number’s persistence.

Native compilation via GraalVM dropped from 2.5% to 2% of builds. Bruno interpreted this as developers building less natively on local machines, not a sign that native is less important in production.

Java remains dominant at nearly 97% of builds. Kotlin grew almost 12% year-over-year in absolute terms — still a small share, but meaningful enough that the team treats Kotlin investment as informed by real data rather than guesswork.

Based on the data, the modal Quarkus developer:

Bruno and Max were both clear that the analytics are not vanity metrics. Kotlin adoption data informs how much the team invests in Kotlin support. Gradle’s share informs tooling investment. Extension growth rates influence where limited engineering resources get focused. The Java version distribution directly shaped the decision to target Java 21+ for Quarkus 4.

The team encouraged anyone not yet opted in — especially Kotlin or Gradle users — to enable the analytics, since the current numbers likely undercount those populations. More representative data leads to better prioritization decisions.

The episode also mentioned that Quarkus 3.37 has been released, and development on Quarkus 4 continues alongside the 3.x series (expected through 3.40). Quarkus Insights will be on a summer break for two weeks; the next episode will return to domain-driven design and hexagonal architecture.

The build analytics give the Quarkus team — and the wider community — a grounded, data-driven view of how Quarkus is being used in practice. The picture that emerges from May 2026 is of a maturing, production-focused ecosystem: developers are building more, upgrading Java versions faster than ever, reaching for production essentials like schema migration and secrets management, and coming from a more geographically diverse set of time zones than ever before.

If you want to help make the data more representative, you can enable build analytics by setting disabled: false in ~/.redhat/ — or just answer "yes" next time dev mode prompts you. Every build counts.