GTLB · Research
GitLab Is the Checkpoint Agent Code Must Pass. Stock Pay Took 103% of Operating Cash, and Growth Is Sliding Toward 15%
GitLab owns the repository and pipeline every agent-written change must clear, but owner cash was -$16.3M over the last year, agent consumption runs near $15M a year, and the guide points to about 15% growth.
- Owner cash after all capex and stock pay
- -$16.3M
- Stock pay as a share of operating cash flow
- 103%
- Paid consumption run rate from agent work
- $15M
1. Verdict and what this piece answers
Verdict: watch. GitLab holds the place where software becomes real: the repository where code is stored, the merge request where a change is reviewed, and the pipeline that tests it and ships it to production.
As agents write more of the world's code, every one of those changes still has to pass through a checkpoint like this before it can run. That is a strong position.
The problem is that owners are not being paid for it. Owner cash, meaning operating cash flow of $223.3M less $9.6M of equipment, less $230.0M of stock pay, was -$16.3M for the twelve months ended Jul 31, 2026, our calculation [2][17].
Stock pay took 103% of operating cash flow.
Growth is also slowing just as the agent story is supposed to take off. Revenue grew 21.3% in the quarter ended Jul 31, 2026, but the company's own guide for the quarter ending Oct 31 is $281M to $283M, which is less than the $286.3M it just booked and only 15.0% to 15.8% above the same quarter a year earlier, our calculation [3][17].
Agent work is billed separately through GitLab Credits, but paid consumption ran at only about $15M a year in the first quarter, against more than $1B of recurring revenue [7][20].
The question this piece answers is whether GitLab can charge agents for passing through its checkpoint before its seat-based business shrinks under it, and whether owners see any of it after stock pay. Our view: the checkpoint is real and defensible, but the toll is still mostly charged per human, and the cash after stock pay is around zero.
2. Rating metrics
| Metric | Value |
|---|---|
| R&D share of revenue | 28.7%, research and development expense of $274.6M on $955.2M of revenue, fiscal 2026 [1]. Over the twelve months ended Jul 31, 2026 it was $304.2M, 28.8% of revenue, our calculation [2][17] |
| GAAP net margin | -5.9%, a net loss attributable to GitLab of $56.0M on $955.2M of revenue, fiscal 2026 [1]. Twelve months ended Jul 31, 2026: -$52.7M, -5.0%, our calculation [2][17]. Restructuring charges in fiscal 2027 make the recent figure look worse than the run rate |
| API quality | 5 of 5, our scoring: documented REST and GraphQL APIs with self-serve tokens, public tier pricing and documented rate limits, a generally available MCP server that lets outside agent tools such as Claude Code and Cursor act on projects, merge requests and pipelines, an official command line tool and client libraries with a public changelog, and an integrations ecosystem across clouds and model providers [1][18] |
| Toll-gate score | 2 of 4 pass, 2 partial (section 2b) |
2b. Thesis and the toll-gate test
GitLab began in 2011 as an open source project and became a single application covering the full life of software: planning, source code, code review, continuous integration, security scanning, packaging and deployment [2]. Customers run it on GitLab.com, on their own servers, or as a dedicated single-tenant cloud [1].
More than 50 million registered users and about half of the Fortune 100 use it [3]. Its principal competitor, named in the 10-K, is Microsoft, which owns GitHub; Atlassian, JFrog and Harness compete in parts of the cycle [1].
The bet management laid out in May is that software will be built by machines and directed by people [6]. Agents will open merge requests in parallel and trigger pipelines around the clock, and someone has to decide which of those changes is allowed to reach production.
GitLab wants to be that control point: one data model, one set of permissions, one audit trail, and one policy engine that every agent, pipeline and merge request passes through by default [6].
| Test | Result | Evidence |
|---|---|---|
| Owns the record | Pass | GitLab stores the code, the review history, the pipeline results and the security findings for each customer, in one data model [1]. GitLab Orbit, in public beta, links code, work items, pipelines, deployments and production signals into a single graph that agents can query; the company says internal tests showed up to 11 times faster agent responses [3]. |
| Owns the connection | Partial | Outside agents such as Claude Code, Codex and Gemini can be plugged in, and GitLab's MCP server gives any compatible agent access under the user's permissions [7][18]. Code must pass through a repository and pipeline to reach production, but GitHub offers the same checkpoint, and agents can write code anywhere before it arrives [1] |
| Prices close to outcomes | Partial | The core platform is priced per licensed user, by tier [1]. Agent work is sold through GitLab Credits, usage-based, and since June through GitLab Flex, one annual commitment that can be spread across seats, credits and new features [1][3]. That is the right direction, but paid consumption was only about $15M a year in the first quarter [7] |
| R&D on the gate vs features | Pass | R&D runs near 29% of revenue, and the stated priorities are the gate itself: rebuilding Git for machine-scale load, the Orbit context graph, identity, audit and policy as core services, and a secrets manager with the same permissions and audit trail as the code [3][6][7]. |
3. Growth
| Measure | Value | Change |
|---|---|---|
| Fiscal 2024 revenue (year ended Jan 31, 2024) | $579.9M | +36.7% [17] |
| Fiscal 2025 revenue | $759.2M | +30.9% [17] |
| Fiscal 2026 revenue | $955.2M | +25.8% [1] |
| Twelve months ended Jul 31, 2026 | $1,055.1M | Our calculation [2][17] |
| Q1 fiscal 2027 revenue | $264.2M | +23.2% [4] |
| Q2 fiscal 2027 revenue | $286.3M | +21.3% [3] |
| Q3 fiscal 2027 revenue guide | $281M to $283M | +15.0% to +15.8%, our calculation [3][17] |
| Fiscal 2027 revenue guide | $1,129M to $1,133M | +18.2% to +18.6%, our calculation [3] |
| Implied second half of fiscal 2027 | $578.6M to $582.6M | +14.6% to +15.4%, our calculation [3][17] |
| Remaining performance obligations, Jul 31, 2026 | $1.2B | +16%; current portion $744.7M, +20% [3] |
Growth has stepped down every year, from 37% in fiscal 2024 to 26% in fiscal 2026, and the guide points to about 15% in the second half of fiscal 2027 [3][17]. The second quarter beat the company's own $272M to $274M guide by more than $12M, and the full-year guide was raised twice, from $1,099M to $1,118M in March to $1,129M to $1,133M in September [3][4][20].
Even so, guiding the third quarter below the second is unusual for a subscription business. The filings do not explain it.
One possible cause is timing: self-managed license revenue is recognized up front, and the 10-Q warns that revenue timing under Flex depends on how fast self-managed customers convert [2].
New business is improving. First orders, meaning initial purchases by new customers, more than doubled year over year in the second quarter, and the company called gross bookings a record
4. New products and shipping pace
In January 2026 the company released the Duo Agent Platform, which runs specialist agents for planning, security analysis, data analysis and pipeline repair, plus multi-step flows for code review, fixing failed pipelines, resolving vulnerabilities and filtering false positives from security scans [6][7]. In the first quarter it opened the platform to free-tier users with flat-rate agent code reviews and spending caps, deepened its link to Anthropic's Claude models, and signed collaborations with AWS and Google Cloud so customers can run agents on Amazon Bedrock or Google Vertex under their existing cloud commitments [4].
In the second quarter it launched GitLab Flex, a usage-based secrets manager and the Orbit context graph in public beta [3]. Its MCP server, which became generally available in GitLab 18.6, lets outside agent tools act on GitLab data with OAuth sign-in and per-toolset permissions [18].
Adoption is early. Duo Agent Platform was part of 4 of the 10 largest deals in the first quarter, and a pilot at a top-10 US bank showed about 1.5 hours saved per developer per task [7].
GitLab Ultimate, the top tier with security and compliance, made up 57% of recurring revenue at the end of the first quarter and 7 of the 10 largest deals [7].
5. Net revenue retention
GitLab measures net retention as this year's recurring revenue from the customers it had a year ago, divided by those customers' recurring revenue a year ago, including upsells, price changes, user growth, contraction and churn [3]. The rate fell four points in a year and held flat at 117% in the last two quarters.
Gross retention is not disclosed. The company told investors that headwinds persist for about 20% of recurring revenue, from price-sensitive customers, and that it is building lower-cost options for them [7].
That matters because seat-based revenue shrinks if agents let customers do the same work with fewer developers.
6. Free cash flow and owner cash
| Period | Operating cash flow | Equipment | Stock pay | Owner cash |
|---|---|---|---|---|
| Fiscal 2023 | -$77.4M | $6.1M | $122.6M | -$206.1M [17] |
| Fiscal 2024 | $35.0M | $1.6M | $163.0M | -$129.6M [17] |
| Fiscal 2025 | -$64.0M | $3.8M | $185.9M | -$253.7M [17] |
| Fiscal 2026 | $232.9M | $10.8M | $215.0M | $7.1M [1] |
| Twelve months ended Jul 31, 2026 | $223.3M | $9.6M | $230.0M | -$16.3M, our calculation [2][17] |
Owner cash has been negative in four of the last five periods, and the one positive year, fiscal 2026, left $7.1M. The company's own measure, adjusted free cash flow, was $219.6M in fiscal 2026 [20].
Nearly all of the gap is stock pay. GitLab does not show capitalized software as a separate cash line; stock pay is shown net of amounts capitalized, so the equipment line is the full capex we can see [2].
Three adjustments matter. First, fiscal 2025 operating cash flow absorbed a one-time tax payment of $187.7M tied to moving its intellectual property to the US under a bilateral advance pricing agreement [17].
Without it, fiscal 2025 owner cash would have been about -$66.0M, our calculation. Second, our standard strips interest on corporate cash.
GitLab earned $47.4M of interest over the last twelve months on about $1.26B of cash and investments, our calculation [2][17]. Removing it takes owner cash to about -$63.7M.
Third, the restructuring cost real cash: the second quarter showed -$3.1M of operating cash flow, which the company attributes to slower collections and restructuring payments, and severance paid under the plan reached $12.5M through Jul 31
7. Acquisitions
GitLab has made no acquisitions in the last 24 months. The one related-party arrangement worth knowing is a right of first refusal, signed Nov 25, 2025 for $1,000, over Kilo Code, Inc., a coding-agent company whose sole member is GitLab's Executive Chair, Sytse Sijbrandij [10][12].
If Kilo received a takeover offer it wanted to accept before Aug 24, 2026, GitLab could match it. The audit committee approved the deal [12].
No offer has been disclosed, and the right has now lapsed by its terms. We note it because the founder and largest individual holder now also owns a company building agents that write code, the kind of product that sits in front of GitLab's gate.
8. Risks and catalysts
Risks. Seat compression: if agents let customers do the same work with fewer developers, per-user revenue shrinks before credit revenue scales [2].
Competition: Microsoft's GitHub is the principal competitor and is bundled with the largest coding agent ecosystem [1]. Margin: GAAP gross margin fell to 84% from 88% in the second quarter, mostly from $11.0M more hosting cost for GitLab.com and cloud usage, so agent traffic raises costs as well as revenue [2].
Execution: the company replaced its chief financial officer, chief technology officer and chief accounting officer within nine months and is exiting 22 countries during a restructuring [5][13][14][15]. Forecasting: the 10-Q warns that Flex makes revenue timing harder to predict [2].
Catalysts: third quarter fiscal 2027 results, expected around Dec 1, 2026 (unconfirmed; data vendors estimate the date and the company has not announced it as of Oct 31, 2026; the last two releases came Jun 2 and Sep 1); whether the third quarter lands above the $283M top of the guide (pending); the first disclosure of Flex commitments as a share of bookings (pending); completion of the restructuring by Jan 31, 2027 (company target)
9. Valuation anchor and balance sheet
Owner cash has been positive in only one full year, fiscal 2026, and was negative over the last twelve months [17]. Our standard needs three full positive years before a base value is set, so none is set here, and none is likely before fiscal 2029 on the current trend.
The anchor to track is twelve-month owner cash, -$16.3M today, our calculation.
GitLab carries no debt and about $1.26B of cash and investments [2]. In March the board approved a $400.0M buyback, and the company spent $154.7M on 5.8M shares in the first six months of fiscal 2027 [2].
Under our standard buybacks only matter when they change leverage; these are paid from cash on hand, not borrowing, so they do not. But they exceed the owner cash the business produced, so they are a draw on the cash pile, not a return of earnings.
10. Superintelligence angle
Mixed, leaning tailwind. Superintelligence agents multiply the amount of code, the number of changes and the number of pipeline runs, and GitLab sees that directly in its own usage data [7].
Every one of those changes needs review, testing, security scanning, policy checks and an audit record before it reaches production, which is exactly what GitLab sells [6]. The headwind is pricing.
GitLab still earns most of its revenue per human user, and if superintelligence lets a team of ten do the work of fifty, seat counts fall even as traffic rises [2]. The company has moved early with Credits and Flex, and its chief executive wrote that the market is moving from tens of dollars per user per month to hundreds [6].
Until consumption revenue is large enough to show in the numbers, that is a claim, not a result.
11. Leadership and control
Co-founder Sytse Sijbrandij resigned as chief executive effective Dec 5, 2024, and became Executive Chair of the board the same day; the board let his 2021 options keep vesting while he serves in that role [8]. Bill Staples, legally William Staples, became chief executive and a director on Dec 5, 2024 [8].
Staples ran New Relic as chief executive from July 2021 to December 2023, led engineering for Adobe's Experience Cloud, and spent 17 years at Microsoft, most recently as vice president of the Azure application platform [8]. He owned 81,787 shares at Apr 1, 2026 [10].
The rest of the team is new. Jessica Ross, formerly chief financial officer of Frontdoor and a Salesforce finance executive, became chief financial officer on or about Jan 15, 2026, after Brian Robins left in September 2025 [13].
Sivaprasad Padisetty, Staples's chief technology officer at New Relic, took the same role at GitLab on or about Jan 15, 2026, replacing Sabrina Farmer [14]. Ross also became principal accounting officer on Sep 16, 2026, when Simon Mundy resigned [15].
Ian Steward has been chief revenue officer since May 3, 2025 [10]. Godfrey Sullivan is lead independent director
12. Questions for the owner and predictions
Question one: when will stock pay fall below operating cash flow, and does the restructuring get it there in fiscal 2028?
Question two: what share of recurring revenue comes from Credits and Flex, and when will it be disclosed each quarter?
Question three: why is the third-quarter revenue guide below the second quarter's result?
Prediction one: third quarter fiscal 2027 revenue lands above $283M, the top of the guide (65%).
Prediction two: net retention is 117% or lower at Oct 31, 2026 (60%).
Prediction three: fiscal 2027 owner cash, as we define it, is negative (65%).
Prediction four: stock pay for fiscal 2027 exceeds $240M (55%).
Prediction five: the company discloses consumption revenue or Flex commitments as a standing metric by Jan 31, 2027 (35%).
Filing notes. The third-quarter guide sits below the second-quarter result with no explanation in the 10-Q or release [2][3].
The chief financial officer cites sequential acceleration in dollar-based net retention, but the disclosed rate was 117% in both the first and second quarters [3][4]. The 2026 proxy shows the chief legal officer with 60,858 Class A shares while its footnote describes 17,366 shares owned directly, the same figure given for another officer [10]
Bottom line
GitLab owns a real gate: code written by agents still has to pass review, testing and policy checks, and that traffic is rising fast, with pushes up 49% and pipelines up 38%. Owners have not been paid for it.
Stock pay took 103% of operating cash, owner cash was -$16.3M over the last twelve months, and growth is sliding from 21% toward 15%. The agent toll is still small, at about $15M of paid consumption against more than $1B of recurring revenue.
Our view would turn positive if consumption revenue becomes a disclosed, fast-growing line and stock pay falls below operating cash flow; it would turn negative if GitHub pulls the checkpoint inside its own platform. Verdict: watch, conviction 3 of 5.
Stock Profiles & Comparable Companies
GTLB Stock Profile →More Published Research
View All →Diebold Services the Machines Where People Use Cash but Owners Kept 113 Million Dollars
Diebold grew third quarter revenue 5.2%, but owner cash after capex and all stock pay fell to $112.7M over twelve months while buybacks continued.
Asure Holds the Payroll Record for 100 Thousand Small Employers but Owners Kept 1 Million Dollars
Asure grew recurring revenue 16% in the third quarter, but after capex and all stock pay owners kept $1.0M over the latest twelve months while the share count rose.