Razor's Edge: Giga Long Gitlab Time...

Investors generally want a story to be easy. They frame a narrative in their head and then go to work seeking the confirmation points along the way. If the stock works out, they "predicted" what would happen. If they didn't, well there is usually some sort of finger pointing or debate over being early/late i.e. wrong. But if you do this long enough, you come to realize an investment thesis is organic and fluid vs inorganic and statically defined.

When I first got involved with Gitlab, I had not yet spent much time on the biz and was simply of the view that AI coding startups and general SaaS businesses gonna die fears had created a good risk/reward trading opportunity. The work quickly seemed to confirm this. I spent my initial time investigating whether or not Fortune 100 customers were on the verge of all abandoning ship. This isn't the hardest thing to do as you generally assume they are likely not. But not dying overnight is the definition of a short-term rental thesis, and not a long-term investment. My journey as a investor has pretty much reflected that. I've bought the stock $43 or lower and sold $47 and higher 3x since April with the last notable exit over $50. My trading has been more of a function of the SaaS factor headwind capping my upside relative value expectations and the thesis not evolving much.

Consequently, as of a few weeks ago, I had no Gitlab exposure despite it drifting towards to lower end of the recent range. Then my macro views on AI investing started to change, and some micro data points started to support me taking an "investor" level conviction approach to the name.

I think the narrative around this name is now at an inflection point and I'm getting there in real-time by acknowledging just how fluid things have been. And there is no better way to do this than by recapping how a marquee customer's views/expectations have evolved in a very short time.

This Fortune 100 Global Financial Giant has been a Gitlab customer since 2018.

Phase 1: The "Copilot Crisis" (Early-Mid 2025)

Customer adopts Github co-pilot in Summer of 2024 and immediately deems it as critical to their business. Gitlab is behind on this front and thus not only is not benefiting from the new product demand, but now has a customer platform issue due to no integration with competitors product.

By early 2025, you have some existential customer risk materializing:

• The Attempted Exit: They "started a trend of moving away from GitLab and started to use GitHub" for new AI-heavy repositories. There was even a push by some leaders to "focus on migrating GitLab to GitHub" entirely.
• The Technical Wall: They hit a "temporary state of indecision" where they realized business-critical apps couldn't just move.
• The "Band-Aid" Solution: Instead of leaving, they built a "homegrown solution" to sync repos between GitLab (for compliance) and GitHub (for Copilot).
• Spend Forecast: Flat. They expected GitLab spend to stay "pretty flat" as headcount reductions counterbalanced the growth in repositories.

This was a stress-test confirmation moment for me as you had a customer clearly acknowledging in real-time what investors were already rapidly discounting in the share price. This is how a still printing 30% growth and gradually decelerating growth biz experiences the type of multiple contraction that a 30% to 10% growth biz goes through. But as I pointed out then, this was all I needed for a very sized long trade with 15-20% expectations. This is because this customer simply does not fuck around with their tech stack and would have moved to what made the most strategic sense tooling wise if need be. To be clear this wasn't exactly bullish seeing as they had fallen enough behind for this risk to have become real, but as new investor at 5x ev/sales that was my opportunity.

Phase 2: "Death by Devin " (Late Q3 2025)

By late summer/early fall, the I'm a co-pilot laggard survivor thesis had worn off. The conversation shifted to autonomous agents like "Devin" (Cognition Ai) and secular/structural dev seat license perpetual headwinds. This was something I was again willing to discount as more of a unjustified near-term concern by some investors as generally speaking checks on these engineer replacing agent tools remained mix.

But reflexivity being what it is, the customer was in fact having this moment as well:

• The New Toy: Leadership became "more excited about Cognition and [Devin]... than GitLab". They viewed GitLab as more of a legacy utility while Devin was the future.
• The "30%" Cut: The long-term internal target for headcount reduction was revised from ~15% up to "30%... next year" specifically because of the ROI calculation on agents like Devin.
• The Bearish Prediction: The customer explicitly predicted: "I'm sure GitLab outlook will drop". He believed that fewer humans would linearly translate to a lower GitLab bill, predicting that at renewal time (2026), they would be looking to cut spend.

To be clear this was more of an outlier take at the time, but again this is a marquee early tech innovator/adopter MNC customer. The structural persistent seat-based headwind debate clearly was not going to go away amongst investors if this is what this customer was saying now. Yes, consumption based pricing was coming and the product was catching up with DUO, but you can't disprove an expectation till you have some facts that reverse it.

Phase 3: The "Agentic Reversal" (Two weeks ago)

The narrative flipped 180 degrees. They realized that replacing human engineers doesn't reduce license count because agents need identities.

• The Reversal: Instead of GTLB bill dropping with engineer headcount, they now see that "licensing needs with less headcount will actually increase"
• The Mechanism: The expected long-term "30% reduction" in human engineers is being backfilled by agents that require "their own entitlements" for audit compliance (Source: Your Turn 13 Notes).
• The New Forecast: Spend is "set to rise" because they are adopting higher-tier plans (Ultimate) to govern these bots and because they intend to levg Gitlab's Agentic platform going forward as well.

As the number of agents grows quickly, total GitLab seats rise even faster than human headcount falls. They expect a large and fast-growing number of bots, and that the number of bots will far outpace human headcount reduction. Each bot is treated like a human for entitlements/licenses.
2026 GitLab spend is set to rise because overall usage is climbing, and customer is preparing to adopt higher-tier plans (including Duo workflows) as part of a broader push to standardize tooling and support heavier code-generation, testing, and MR volume.

GitLab is winning on enterprise fit and language coverage, especially for Java/Spring Boot, test coverage quality, and structured workflows. It complements Devin in Python/C++ work, and both are now used together for refactoring legacy systems, bug-finding, and integration/load testing across different workloads.

Licensing and operational alignment are in place: GitLab is now their strategic cloud partner, all repos have been moved to AWS/GitLab, bot entitlements mirror human licenses for audit compliance, and a governance framework is in place to administer and scale bot entitlements.

So, we went from Devin killing Gitlab growth to it Accelerating it in 3 months. I'd also emphasize this now where things have actually landed with this customer heading into the contract renewal which is coinciding with DUO GA in a couple of weeks, and that "tech relevance" tone turned on dime.

This all transpired over 9 months at Fortune 100 with over 12k software engineers..

Basically, anyone arguing ABC is going to happen is full of shit as the customers have been figuring this out for themselves in real-time. Yes, its now safe to say that things have been noisy and incumbents are now looking notably better, but until that's clear it is not. Hindsight is always 20/10.

GITLAB IN THE AI ERA: What to Expect?

I'm sure there will be more people calling the stock "Shitlab" because of the fed shutdown headwinds in the guide and no perfect print quarter to declare thesis victory on yet, but that's now looking like the alpha brownie.

I came into this print more conflicted then I usually am on a stock. On the one end, I was certain there would be no results that would trigger a panic buying like some of the consumption infra names and the stock was not value guy earnings cheap yet. But on the other end, I was at 5x ev/sales for biz likely to have grown 700bps faster than Mongo in the quarter and maybe 300-400bps slower next q. Yes, directionally still not clear cut convincing people its gonna cagr nicely for years or reaccel very soon, but its also not 12x-20x sales like those names. And this is with DUO GA around the corner, new hybrid pricing consumption coming, and agentic governance tailwinds in coding/devops now being confirmed as a DBNR catalyst.

Consequently, I been doing a lot of actual work on what's going on product/pricing wise to understand/visualize what's around the corner.

My 2c..

Management actually appears to be playing their cards a little closer to their chest then you would usually expect. DUO is about to be GA and that they have told you they will be selling compute credits in a pooled funds manner going forward, but they aren't exactly handholding you on all the drivers ahead.

Everyone shorting the name has keyed in on pricing tailwinds turning into headwinds in the traditional seat based rate increase manner, which is now going to prove to be a mistake. You even had analysts on the call last night asking about whether they are pondering another price increase based on what some of coding tools are charging. Management pushed back on this with multi-skus and hybrid consumption coming, but to me the big driver around the corner is Ultimate upgrades.

There is no need to push price on premium yet, when AI consumption/demand is a natural pull to Ultimate at 3x the price.

Here are some "Hidden Pricing" PLG Agent based drivers from the GitLab 18.6 release me and the supreme intelligence dug up last week...

The "Security Analyst" Stack

If you want the agent that actually fixes vulnerabilities (the "Vulnerability Resolution" feature), you need the "Triple Crown" of billing. It is not just one upgrade; it is three layers of cost.

• Layer 1: The Tier (Ultimate)
  • Requirement: You must be on GitLab Ultimate ($99/user/mo).
  • Why: The agent relies on the "Vulnerability Report" database, which acts as its memory. This database does not exist in Premium. If the agent can't "read" the report, it can't answer your questions.
• Layer 2: The Brain (Duo Enterprise)
  • Requirement: You likely need the Duo Enterprise Add-on (approx. +$39/user/mo).
  • The Catch: While a basic "chat" might work on lower tiers, the specific "Vulnerability Resolution" feature (where the AI writes the code to fix the bug) is locked to the Enterprise add-on.
• Total Cost: ~$138/month per developer to have full "Auto-Fix" capabilities. This excludes the compute consumption that will come on top of this.

The "Planner Agent" Stack

You can use this on GitLab Premium ($29/user/mo), but this is where the "Dev Count Growth" loophole kicks in.

• The Trap: On Premium you are limited to 1 Service Account per 1 Human Seat.
• The Scenario:
  • You have 10 Humans. You get 10 Service Accounts.
  • You deploy the Planner Agent to 5 repos. You also have existing bots for CI/CD, Terraform, and backups.
  • Suddenly, you need 15 Service Accounts to run your new AI workflow.
• The Consequence: You are blocked. You cannot just pay $5 for a bot. You must buy 5 more "Human" Premium Seats ($29 x 5 = $145/mo) just to unlock the identity slots for your agents.

Summary: The "Make or Buy" Decision

GitLab’s strategy is forcing companies into a binary choice:

Customer convos confirm this driver is literally around the corner from a spend intent basis. The marquee customer I discussed earlier has been at under 1% Ultimate tier licenses after six years. Imagine their engineering team shrinks by 20% total over next two years due to Agentic AI, but they have to give everyone left full stack Agentic capabilities/credentials. What happens to their spend?

Full adoption of Ultimate/DUO licenses for the remaining staff would 3.2x ARR from this customer pre-compute consumption overages. Upgrading now has a serious demand pull catalyst, and as of today I estimate 75%+ of Gitlab licensed human seats are on Premium tiering.

So, yeah a premium price increase again makes no sense at this point. In fact, if I'm a customer I upgrade to Ultimate now and lock in pricing there if I am seeing my usage explode as it is, and understand these tools will be critical for all my remaining devs going fwd.

I think by June of 2026 this is consensus in the investing community.

Judging by the call last night and some of these sell-side notes, its nowhere near there yet.

“We are downgrading GitLab to Hold from Buy as we see the company facing low visibility and heightened execution risk in the face of a business model transition that is being forced by secular changes in the developer tools landscape,” analyst Miller Jump wrote in a note to clients. “At its core, GitLab has been built on a seat-based pricing model, but the widespread adoption of AI coding assistants is beginning to structurally disrupt this approach. Unlike traditional users, AI agents cannot be sold a seat in GitLab’s current model, which undermines the legacy backbone of their revenue. In response, management has indicated plans to incorporate a consumption-based component into the pricing model, beginning in earnest with the general availability release of the Duo Agent platform in late 2025/early 2026.”

Like this note is flat-out WRONG and that's not an opinion.. that's coming from Fortune 100 existing customers mouths...

Btw- Unlike past activist saas short work I've done, I can't fault sell-side analysts with takes that some marquee customers have been providing you just over a month ago. This AI disruption stuff has clearly been fluid, and I'd argue the dust is finally starting to settle. Which is a nice spot to be with a SI this high and Federal Govt shutdown and new CFO guide noise moment weighing on the stock as the monetization engine is entering a substantial new tailwind.

And let's not forget that Gitlab has a huge self-managed base that could deliver quite market surprising upside in lumpy fashion over the coming quarters. I also think mgmt clearly understands the advantage of not being a pure SaaS in the Agentic Ai environment and are starting to see it in their biz, and will get more credit very soon for air gap capabilities and fed appeal narrative wise.

So what's the real problem here?

I think its investors being lazier than usual lately and the extended period of poor of poor long-term narrative visibility. This has led to software spreads blowing out on consumption or security names vs everything else and of course the broader AI sectoral imbalances. Gitlab, IMHO, is an orphan still. A consumption/infra overall tailwind winner w directional headwinds from pricing model flux and DUO roll-out.

Consider that Cursor just raised at $30bl and Cognition over $10bl 3 months ago, and here is Gitlab getting questions on its earnings call on whether it can raise its premium tier price 10-15%. That's a disconnect to me.

GITLAB PRODUCT VALUE IS RAPIDLY INCREASING

Management hit you with some key usage/engagement metrics on the call:

In the first 10 months of 2025, key activity metrics CI pipelines, deployments and releases are up about 35% to 45% year-over-year, similar to what peers are seeing. For customers paying us more than $5,000 in ARR, usage proxies like deployments and CI pipelines on a per seat basis are up 20% to 40% annually. Simply put, more code means more of a need for GitLab. Our 2025 Global DevSecOps report shows that while AI accelerates coding, more code doesn't necessarily mean better outcomes.

So, when you see stuff like this what you should look at is the landscape and ARPU for the software.

DEVIN has an entp seat license + agent compute credit pricing model.

Research and customer feedback indicates that Devin spend will be somewhere between 500-1k a month.

Cursor right now looks like $70-80 month avg w 3x that on heavier coding usage.

Gitlab ARPU is sub $40/month still for 2ml devs in largely Entp/Govt environments. It's basically very underpriced product right now and the disclosures on usage reveal that.

And the whole biz is a $5bl ev...you go do the math on margins and op income if arpu is doubling in the base.

That's a compelling investment opportunity.

I'd also suggest anyone curious start reading user feedback on some of these tools as Gitlab went from not in the convo vs copilot to now getting very positive feedback on the value prop.