DevOps Statistics 2026
By Axis Intelligence Research
Co-author: James Porter | Last updated: September 30, 2026 | License: CC BY 4.0
Automation now proposes fixes much faster than teams ship them. Axis Intelligence Research calculates a Patch Merge Rate (PMR™) of 35.29% for 2025: of 40.15 million Dependabot pull requests opened on GitHub, only 14.17 million were merged. That leaves roughly 26 million automated fixes unmerged.
Quick Answer
Nine in ten technology professionals now use AI at work, and nine in ten organizations run at least one internal platform (DORA 2025). Kubernetes runs in production at 82% of container users (CNCF). The weak link sits downstream. Axis Intelligence Research finds that only 35.29% of automated dependency fixes were merged in 2025. CI compute per merged pull request rose 5.04% to 22.17 CPU minutes.
Key Findings
- Axis Intelligence Research calculates a 2025 Patch Merge Rate (PMR™) of 35.29%. That is 14.17 million merged out of 40.15 million Dependabot pull requests opened, per GitHub Octoverse 2025.
- The PMR™ fell from 38.56% in 2024 to 35.29% in 2025, according to Axis Intelligence Research. Automated patch volume held flat while merged fixes declined.
- Axis Intelligence Research finds public GitHub projects spent 22.17 CI minutes per merged pull request (CMM™) in 2025, up 5.04% from 21.11 in 2024.
- DORA’s 2025 report found 90% of technology professionals use AI at work, yet only 24% trust AI output “a lot” or “a great deal.” Axis Intelligence Research puts that at an adoption-to-trust multiple of 3.75.
- CNCF’s 2025 survey found 58% of cloud native “innovators” use GitOps extensively, against 23% of “adopters.” Axis Intelligence Research calculates a GitOps Maturity Multiple of 2.52.
How Many Companies Use DevOps and Platform Engineering in 2026?
The trunk of the DevOps tree is no longer “should we automate delivery.” It is “whose platform does the delivery run on.” Google Cloud’s DORA program surveyed nearly 5,000 technology professionals for its 2025 State of AI-assisted Software Development report. DORA found that 90% of organizations have adopted at least one internal platform. It also found a direct link between platform quality and an organization’s ability to get value from AI.
Cloud native techniques have crossed from experiment into default. The CNCF Annual Cloud Native Survey, published January 20, 2026, found that 98% of organizations use cloud native techniques. It also found that 59% say “much” or “nearly all” of their development and deployment is cloud native.
DevOps adoption indicators, 2025
| Indicator | Value | Population | Source |
|---|---|---|---|
| Organizations with at least one internal platform | 90% | Organizations | DORA 2025 (Google Cloud) |
| Organizations using cloud native techniques | 98% | Organizations | CNCF Annual Survey 2025 |
| “Much” or “nearly all” dev and deployment cloud native | 59% | Organizations | CNCF Annual Survey 2025 |
| Team archetypes identified by cluster analysis | 7 | Teams | DORA 2025 (Google Cloud) |
James Porter’s read: A 90% platform figure means the internal developer platform has become plumbing. Plumbing gets judged on whether water comes out, not on whether it exists. DORA identified seven team archetypes that ran from “Harmonious high-achievers” to teams stuck in a “Legacy bottleneck.” That split is the real story: the same platform headline covers teams shipping cleanly and teams burning out behind a ticket queue.
What Percentage of Companies Run Kubernetes in Production?
CNCF reports that 82% of container users run Kubernetes in production, up from 66% in 2023. The same survey found 66% of organizations hosting generative AI models use Kubernetes for some or all of their inference workloads.
Model delivery still lags code delivery. Only 7% of organizations deploy AI models daily, 47% deploy occasionally, and 44% run no AI/ML workloads on Kubernetes at all.
Kubernetes and cloud native statistics
| Metric | 2023 | 2025 | Source |
|---|---|---|---|
| Kubernetes in production (container users) | 66% | 82% | CNCF |
| GenAI inference hosted on Kubernetes | — | 66% | CNCF |
| Organizations deploying AI models daily | — | 7% | CNCF |
| Organizations deploying AI models occasionally | — | 47% | CNCF |
| No AI/ML workloads on Kubernetes | — | 44% | CNCF |
GitOps adoption: innovators vs adopters
CNCF found that 58% of “cloud native innovators” use GitOps principles extensively, compared with 23% of “adopters.” Axis Intelligence Research calculates a GitOps Maturity Multiple of 2.52 (58 ÷ 23). At that multiple, declarative, pull-based deployment separates the mature organizations more sharply than Kubernetes adoption itself does.
James Porter’s read: Kubernetes at 82% is a scheduler decision. GitOps at 58% versus 23% is a discipline decision. Clusters are easy to stand up. A repo that is the only source of truth for what runs in prod, with drift reconciled automatically, is the harder part. That is where the 2.52× gap lives.
What Are the Biggest DevOps Challenges in 2026?
For the first time in CNCF’s survey, the top blocker is organizational rather than technical. CNCF found that 47% of respondents cite “cultural changes with the development team” as their top challenge. That puts culture ahead of lack of training (36%), security (36%) and complexity (34%).
| Challenge | Share of respondents | Source |
|---|---|---|
| Cultural changes with the development team | 47% | CNCF 2025 |
| Lack of training | 36% | CNCF 2025 |
| Security | 36% | CNCF 2025 |
| Complexity | 34% | CNCF 2025 |
Observability is moving the same way. OpenTelemetry is CNCF’s second-highest-velocity project, with more than 24,000 contributors, and nearly 20% of respondents now use profiling in their observability stack.
How Many Automated Security Fixes Actually Get Merged? The Patch Merge Rate (PMR™)
This is the number no DevOps survey publishes, and the reason this page exists.
PMR™ — Patch Merge Rate — is an Axis Intelligence Research metric. It is the share of automated dependency-update pull requests (Dependabot) that are actually merged in a given year.
Formula: PMR™ = Dependabot PRs merged ÷ Dependabot PRs opened × 100
Inputs: GitHub Octoverse 2025, Dependabot activity series 2023–2025.
| Year | Dependabot PRs opened | Dependabot PRs merged | PMR™ | Source |
|---|---|---|---|---|
| 2023 | 57.74M | 12.15M | 21.04% | GitHub Octoverse 2025; Axis calculation |
| 2024 | 39.91M | 15.39M | 38.56% | GitHub Octoverse 2025; Axis calculation |
| 2025 | 40.15M | 14.17M | 35.29% | GitHub Octoverse 2025; Axis calculation |
Worked example for 2025: 14.17 ÷ 40.15 = 0.3529, so the PMR™ is 35.29%. The unmerged remainder is 40.15 − 14.17 = 25.98 million automated fix proposals.
GitHub’s Octoverse 2025 gives the context. Dependabot is enabled across more than 2.668 million repositories, and average fix time for critical vulnerabilities fell from 37 to 26 days. So fixes that do land are landing faster. What the PMR™ shows is that roughly two in three automated patches never land at all.
James Porter’s read: The 2023 reading of 21.04% came off the post-Log4Shell alert flood, so read it as a spike rather than a floor. The movement that matters is 2024 to 2025. Open volume barely moved (39.91M to 40.15M), and merges fell by 1.22 million. The bot is not the bottleneck. The merge button is. Check your branch protection rules and CODEOWNERS before you blame the scanner. A required review from a team that never looks at dependency bumps is a queue with no consumer.
Why the PMR™ matters for security teams
GitHub reports that Broken Access Control overtook Injection as the most common CodeQL alert, flagged in more than 151,000 repositories. Much of it traces to misconfigured permissions in CI/CD pipelines. An unmerged patch and a misconfigured pipeline token are the same kind of problem: the pipeline exists, and nobody owns its hygiene. For the vulnerability side of this picture, see our patch management statistics.
How Much CI/CD Compute Does It Take to Ship a Change? CI Minutes per Merge (CMM™)
CMM™ — CI Minutes per Merge — is an Axis Intelligence Research metric. It divides GitHub Actions CPU minutes in public projects by pull requests merged in public repositories over the same Octoverse year (September to August).
| Year | Actions minutes (free, public) | Public PRs merged | CMM™ | Source |
|---|---|---|---|---|
| 2024 | 8.5B | 402.7M | 21.11 | GitHub Octoverse 2025; Axis calculation |
| 2025 | 11.5B | 518.7M | 22.17 | GitHub Octoverse 2025; Axis calculation |
Worked example: 11.5 billion ÷ 518.7 million = 22.17 CPU minutes per merged PR. The 2025 reading is 5.04% above 2024.
Actions minutes grew faster than merges, so every shipped change now carries slightly more build, test and scan compute. Including self-hosted runners, GitHub counts 13.5 billion minutes in 2025. We keep the free-public series for CMM™ because only that series has a like-for-like 2024 value.
James Porter’s read: Five percent more compute per merge is not waste by itself. More security scanning in the pipeline should cost minutes. It becomes waste when the extra minutes run full test matrices on docs-only changes. Path filters and job-level if: conditions are the cheap fix. Most workflows don’t use them.
How Many Pull Requests and Deployments Happen on GitHub?
GitHub Octoverse 2025 (year ending August 31, 2025) recorded the most active year in the platform’s history:
| Metric | 2024 | 2025 | Source |
|---|---|---|---|
| Pull requests merged per month (avg) | 35M | 43.2M | GitHub Octoverse |
| Pull requests created per month | 39.5M | 47.5M | GitHub Octoverse |
| Public PRs merged (year) | 402.7M | 518.7M | GitHub Octoverse |
| Commits pushed | — | 986M+ | GitHub Octoverse |
| Repositories with a Dockerfile | 875K | 1.9M | GitHub Octoverse |
Axis Intelligence Research calculates a merge-to-create ratio of 90.95% in 2025 (43.2M ÷ 47.5M), up from 88.61% in 2024. Human-authored pull request flow is healthy. The stall is concentrated in automated patches (PMR™ 35.29%), not in feature work. That contrast is the most useful diagnostic on this page.
Which DevOps Tools Do Developers Use Most?
The 2025 Stack Overflow Developer Survey (24,473 responses to its cloud-development question) shows how far containers have spread:
| Tool / platform | Professional developers | Source |
|---|---|---|
| Docker | 73.8% | Stack Overflow 2025 |
| AWS | 45.9% | Stack Overflow 2025 |
| Kubernetes | 30.1% | Stack Overflow 2025 |
| Microsoft Azure | 27.2% | Stack Overflow 2025 |
| Google Cloud | 24.3% | Stack Overflow 2025 |
| Terraform | 18.7% | Stack Overflow 2025 |
| Prometheus | 12.1% | Stack Overflow 2025 |
| Ansible | 11.2% | Stack Overflow 2025 |
| Datadog | 9.7% | Stack Overflow 2025 |
Docker usage jumped 17 points from 2024 to 2025, the largest single-year gain in the survey. Stack Overflow notes that some of this comes from consolidated technology categories. For source control and planning, professional developers report using GitHub (80.5%), GitLab (36.7%) and Azure DevOps (18.6%).
James Porter’s read: Kubernetes shows up at 30.1% among developers but 82% among container-using organizations. That is not a contradiction. It is platform engineering working as designed. Most developers ship to Kubernetes without touching a manifest.
How Is AI Changing DevOps and Software Delivery?
Google’s summary of DORA 2025 reports 90% AI adoption among software professionals, 14% higher than the prior year, with a median of two hours a day spent working with AI. Some 65% report heavy reliance, and over 80% say AI raised their productivity.
Trust has not caught up. 24% trust AI output “a lot” or “a great deal,” while 30% trust it “a little” or “not at all.” Axis Intelligence Research calculates an adoption-to-trust multiple of 3.75 (90 ÷ 24).
DORA’s delivery finding is the one that matters for operations. AI adoption is now associated with higher software delivery throughput, a reversal from 2024, but it still carries a negative relationship with delivery stability. For deeper AI-specific data, see our AI-generated code statistics and AI coding assistant statistics.
James Porter’s read: More throughput plus less stability is the same shape as the PMR™ curve. Change volume rises, and the verification layer does not. AI writes the diff. Your test suite, your review rota and your rollback path decide whether it’s safe to merge.
How Much Do DevOps and Software Engineers Earn?
The U.S. Bureau of Labor Statistics does not publish a separate “DevOps engineer” occupation. Most DevOps roles fall under software developers or adjacent categories. BLS reports the following for May 2025:
| Occupation (U.S.) | Median pay | Employment 2025 | Projected growth 2025–35 | Source |
|---|---|---|---|---|
| Software developers | $135,980 | 1,717,800 | 10% | BLS OOH |
| QA analysts and testers | $104,300 | 187,600 | 6% | BLS OOH |
BLS projects about 106,100 openings a year across both groups. Axis Intelligence Research calculates 9.16 developers per QA tester in 2025, widening to 9.55 by 2035 on BLS projections. QA median pay is 76.7% of developer median pay. The verification workforce is growing more slowly than the workforce that generates change, which matches the merge and stability data above. For language-level demand, see our programming language statistics.
Frequently Asked Questions
Why do most Dependabot pull requests never get merged?
Axis Intelligence Research’s Patch Merge Rate (PMR™) was 35.29% in 2025: 14.17 million merged out of 40.15 million opened (GitHub Octoverse 2025). Common causes are required reviews from teams that don’t triage dependency bumps, failing CI on stale branches, and no auto-merge rules for patch-level updates.
Is Kubernetes still growing in 2026?
Yes. CNCF’s survey published January 2026 found 82% of container users run Kubernetes in production, up from 66% in 2023. Growth now comes from AI inference: 66% of organizations hosting generative AI models run some inference on Kubernetes.
Does AI make deployments less stable?
DORA 2025 found AI adoption is associated with higher delivery throughput but still has a negative relationship with delivery stability. Teams with strong automated testing and fast feedback loops gain the most.
What share of teams practice GitOps?
CNCF 2025 found 58% of cloud native innovators use GitOps extensively versus 23% of adopters. Axis Intelligence Research calculates a GitOps Maturity Multiple of 2.52.
How much CI compute does one merged pull request cost?
On public GitHub projects, Axis Intelligence Research’s CI Minutes per Merge (CMM™) was 22.17 CPU minutes in 2025, up 5.04% from 21.11 in 2024.
Is the biggest DevOps obstacle technical or cultural?
Cultural. 47% of CNCF 2025 respondents named cultural change within development teams as the top challenge, ahead of training (36%), security (36%) and complexity (34%).
Are DORA “elite performer” multiples still current?
The 2025 DORA report replaced single performance tiers with seven team archetypes. Pages that still quote legacy elite-versus-low multiples as current-year data are citing an older model.
What is the median salary for a DevOps-adjacent engineer in the U.S.?
BLS reports a May 2025 median of $135,980 for software developers, the occupation that covers most DevOps roles. There is no separate DevOps occupation code.
Methodology
Axis Intelligence Research assembled this dataset between September 29 and 30, 2026 from six primary sources: the Google Cloud DORA 2025 announcement, Google’s DORA 2025 summary, CNCF’s Annual Cloud Native Survey announcement (January 20, 2026), GitHub Octoverse 2025 (updated February 28, 2026), the Stack Overflow 2025 Developer Survey, and the BLS Occupational Outlook Handbook (last modified August 27, 2026). Each figure was fetched directly, and each ratio was computed twice, in floating point and in exact rational arithmetic, before it entered the text.
Formulas. PMR™ = Dependabot PRs merged ÷ opened × 100. CMM™ = free public GitHub Actions CPU minutes ÷ public PRs merged. GitOps Maturity Multiple = innovator ÷ adopter extensive-GitOps share. Adoption-to-trust multiple = AI adoption ÷ high-trust share. Developers per QA tester = BLS employment ratio.
Scope. PMR™ covers Dependabot only, not Renovate or other bots, and it combines public and private activity as GitHub reports it. CMM™ covers free public-project Actions minutes, so enterprise and self-hosted pipelines are excluded from the ratio. Survey figures reflect each organization’s own sample. We deliberately did not merge survey percentages from DORA, CNCF and Stack Overflow into one composite, because their populations differ.
About This Dataset
devops-statistics.csv contains 85 rows covering AI adoption and trust in software delivery, platform and Kubernetes adoption, GitOps maturity, cloud native challenges, GitHub pull request, CI and Dependabot activity for 2023–2025, developer tooling usage, and U.S. pay and employment. It also includes every PMR™ and CMM™ input and reading. Each row carries its source organization, document, URL, retrieval date and, for Axis calculations, the formula.
License: CC BY 4.0. Citation line: Axis Intelligence Research, DevOps Statistics 2026, 2026. The dataset is updated when DORA, CNCF, GitHub Octoverse or BLS publish new editions.
Cite This Page
APA: Axis Intelligence Research, & Porter, J. (2026, September 30). DevOps statistics 2026: Adoption, Kubernetes, CI/CD, AI and the patch merge gap. Axis Intelligence. https://axis-intelligence.com/devops-statistics/
MLA: Axis Intelligence Research, and James Porter. “DevOps Statistics 2026: Adoption, Kubernetes, CI/CD, AI and the Patch Merge Gap.” Axis Intelligence, 30 Sept. 2026, axis-intelligence.com/devops-statistics/.
Chicago: Axis Intelligence Research, and James Porter. “DevOps Statistics 2026: Adoption, Kubernetes, CI/CD, AI and the Patch Merge Gap.” Axis Intelligence, September 30, 2026. https://axis-intelligence.com/devops-statistics/.