Atlassian 66 亿美元 ARR 增长加速的 5 条启示

Atlassian closed its fiscal year on June 30 and reported on August 6. Going in, it was one of the most doubted names in enterprise software. The bear case was simple: AI agents write the code, AI agents file the tickets, and the tools that track the work get squeezed out of the middle. The stock traded around $84 in early July.

Then Q4 landed. Revenue up 28% to $1.766B. Cloud up 31% to $1.213B, accelerating. Subscription ARR of $6.6B, up 23%. RPO up 44% to $4.8B. GAAP operating income of $211M against a $28M loss a year ago. The stock closed at $110.17 on August 6 and opened the next morning on its way to $146, a 35% single-day move.

For a 25-year-old company at $6.6B in ARR, growth going up is the rarest thing on this list. But the more useful learnings are underneath the headline. AI spend is way up, but so is backlog/RPO (revenue to come).

The quick numbers:

• Q4 revenue: $1.766B, up 28% YoY
• Q4 cloud revenue: $1.213B, up 31% YoY (accelerating)
• Subscription ARR: $6.606B, up 23% YoY
• RPO: $4.817B, up 44% YoY
• FY26 revenue: $6.572B, up 26%
• Q4 non-GAAP operating margin: 36%, up from 24%
• FY26 free cash flow: $1.319B, down from $1.416B
• Customers: 350,000+, including 85% of the Fortune 500

#1. Growth Accelerated at $6.6 Billion in ARR. Almost Nobody Had That Priced In.

Cloud revenue growth went from 28% for the full fiscal year to 31% in Q4. Total revenue growth went from 26% for the year to 28% in Q4. At this scale, that is a genuinely uncommon result. Most B2B leaders past $5B are managing decay curves, not reversing them.

What is more interesting is the gap between the fundamentals and the price. Atlassian traded in the low $80s five weeks before the print. The market had substantially written down the entire “system of work” category on the theory that agents would eat it. One quarter of accelerating cloud growth erased that view in a single session.

Two takeaways for founders and operators:

• Multiple compression runs ahead of the numbers, and it reverses just as fast. Atlassian’s business never broke. The narrative did. When the narrative flips, a 35% day is what it looks like. If you are raising, benchmarking, or setting a secondary price against public comps right now, understand that you may be pricing off sentiment that has nothing to do with the revenue underneath.
• “AI will kill this category” is a thesis, not always a finding. So far in software development tooling, more agents has meant more work items, more branches, more reviews, and more governance. Atlassian sells the coordination layer for all of it. Volume went up, not down.

One honest caveat: reported revenue grew 28% while Subscription ARR grew 23%. Revenue is the lagging number here, flattered by contract timing and Data Center renewals. The 23% is the better read on the underlying run rate, and the FY27 guide confirms it.

#2. AI Is Free. The Margin Isn’t. Same Problem as Figma, Different Line Item.

Atlassian is bundling its new agentic capabilities into paid Jira Cloud at no additional charge, including orchestration of third-party coding agents like Claude Code, Cursor, and GitHub Copilot. Rovo is now used by more than 80% of the Fortune 500, and Rovo-assisted actions grew 50% quarter over quarter.

Now look at the guide.

• Q4 FY26 non-GAAP operating margin: 36% (a record, up 12 points YoY)
• Q1 FY27 guided non-GAAP operating margin: 28.5%
• FY27 guided non-GAAP operating margin: 25%
• FY27 guided GAAP operating margin: 4.5%, down from 12% in Q4

Atlassian just posted its best margin quarter ever and then guided away 11 points of it. Gross margin gets trimmed too, from 88% non-GAAP in FY26 to a guided 86.5% in FY27.

Compare that to Figma’s Q2, reported the day before. Figma grew 48% at a $1.5B run rate, the third straight quarter of acceleration, beat, raised the year, and the stock fell 15%. The number that got the attention there was gross margin: 84% GAAP, down about 5 points year over year, because inference is a real per-request cost in a way that hosting a design canvas never was. Figma is also carrying inference on beta products that generate no revenue at all.

Two of the most-watched software companies in the market, one week apart, with the same underlying finding: AI

cost lands in the P&L one to four quarters before AI revenue does. What differs is where it shows up and how it gets communicated.

• Figma took it in gross margin. Credits are a second meter, sold coterminous with the subscription, so the cost sits in COGS and shows up in the quarter you ship. Visible, immediate, and the market marked it as expense rather than investment.
• Atlassian is taking it in operating margin, and in the guide rather than the quarter. Q4 gross margin actually went up, to 87% GAAP from 83%. The spend is going in as R&D, GTM, and inference across FY27, disclosed in advance, with the headline quarter left clean.

The pricing models differ the same way. Figma is metering AI directly. Atlassian is giving the capability away and monetizing the tier: push customers into higher collections carrying roughly 10x the Rovo credits, then test consumption overages and a Flex commitment model on top. Customers using Rovo are reportedly growing ARR at more than twice the rate of non-adopters, which is the number Atlassian is actually underwriting.

Why the market reactions diverged? It was not that one company solved the AI cost problem and the other didn’t. Figma’s stock broke on the sequential shape of its Q3 guide, not on the margin line. Atlassian’s popped because cloud growth accelerated.

If you are pricing AI features right now, the useful question from these two leaders is not whether you charge for it. It is which line you want the cost to hit and whether your buyers will tolerate a second meter. Figma could charge because the AI output is the deliverable. Atlassian gave it away because the AI makes the seat stickier and the collection upgrade is worth more than the credit revenue. Atlassian could afford that trade because it already had 350,000 customers to upgrade into. Most companies have exactly one of those two options, not both.

#3. 85% of the Fortune 500 Uses Atlassian. The Fortune 500 Is About 10% of Revenue.

This is the most useful stat in the entire report.

Atlassian is inside 85% of the Fortune 500. Those accounts represent roughly 10% of total revenue. Total penetration, minimal monetization. Twenty-five years of bottoms-up distribution produced presence almost everywhere and spend almost nowhere.

The enterprise motion is finally catching up:

• Record quarter for $1M, $3M, and $5M deals, all three levels
• 164 customers now above $3M in ARR, up more than 50% YoY
• Customers above $5M in ARR up more than 70% YoY

And it costs money. Sales and marketing spend grew 36% in FY26 against 26% revenue growth, for a company that built its brand on not having a sales team. That is what converting free-and-cheap distribution into seven-figure contracts actually requires.

The lesson generalizes past Atlassian. Bottoms-up adoption is not monetization. If your product is in most of your target accounts and your ACVs are still four figures, the gap is a go-to-market gap, not a product gap, and closing it means hiring the expensive motion you spent years avoiding. Atlassian got 25 years of free land. The expand is now the entire company.

#4. RPO Grew 44%. Free Cash Flow Fell 7%. This is The Reality of AI Driven Growth

RPO is the most underdiscussed number in this print and the one the market actually traded on. Remaining performance obligations, the contracted revenue Atlassian has signed but not yet recognized, grew 44% to $4.817B. Revenue grew 28%. Subscription ARR grew 23%. The current portion of deferred revenue grew 12%.

Four growth rates, all measuring the same business, spread across 32 points. The order matters.