BUSINESS OVERVIEW

Atlassian Corporation

NASDAQ: TEAM · Fiscal year ended 30 June 2026

7 September 2026

Evidence base: Atlassian FY2026 Form 10-K (filed 14 August 2026), FY2025 10-K, FY2021 and FY2016 Form 20-F, Q3 FY2026 Form 10-Q, quarterly earnings releases through 6 August 2026, and the 2025 proxy statement — all held in the analyst's own document archive. Industry structure is drawn from independent primary sources: competitors' own filings and earnings releases, the Stack Overflow and JetBrains developer surveys, GitHub Octoverse, and US Bureau of Labor Statistics data.

This is a business overview for long-term investors. It is not a valuation and not a recommendation. It contains no price target, no multiple, and no view on the shares.

1. Executive Snapshot

2. What the Company Does

Every organisation that builds software faces the same coordination problem: work is done in parallel by people who cannot all be in the same conversation, and someone above them needs to know what is happening without interrupting it. Atlassian sells the system of record for that work. Jira holds the tickets and the workflow states, Confluence holds the documents and decisions attached to them, and Jira Service Management holds the same machinery pointed at IT and internal support requests. These three products generate the majority of revenue (FY2026 10-K); the company does not break out how much each contributes.

The unit is one paid seat, and it is not disclosed

Atlassian charges per user per month, by product and by edition. A customer is defined in the filings as "an organization with unique domains with an active subscription for two or more seats" (FY2026 10-K), which makes the seat the underlying billing unit. The company does not publish how many paid seats exist, what the average one costs, or what net revenue retention is — three of the four numbers an investor would want. What it does publish is the customer count (more than 350,000 at FY2026 year end, up from more than 300,000), Subscription ARR of $6,606 million at 30 June 2026 (up 23%), and the count of customers spending over $10,000 of Cloud ARR: 45,842 in FY2024, 51,978 in FY2025, and 57,334 in FY2026 (FY2026 10-K).

That last series is the closest thing to a unit metric the company gives, and it is decelerating: 13.4% growth in FY2025, 10.3% in FY2026 (inferred). It matters because the same filing says this cohort — roughly 16% of all customers (inferred) — represents "the majority of our Cloud revenue." The customer base is extremely long-tailed. Total revenue divided by the customer floor implies about $18,800 per customer (inferred, and an upper bound because the denominator is a "more than" figure).

Tracing one seat from sign-up to cash

A team adopts a free edition or buys a single product for a handful of users — at a price the filings describe as one that "typically does not require capital budget approval." Most free users never convert, and the company says so plainly. Those that do buy through the website, which handles "thousands of commercial transactions daily" (FY2026 10-K). Small customers pay by card, and Atlassian counts card receivables as cash equivalents because they settle within three days. Larger customers are invoiced, increasingly through a partner: over 50% of FY2026 revenue came through channel partners' sales efforts (FY2026 10-K).

Cash usually arrives before the service is delivered, which is why $2,662 million of deferred revenue sat on the FY2026 balance sheet. How that cash becomes revenue depends entirely on the deployment, and this is the single most consequential accounting fact in the business. Cloud subscriptions are recognised ratably over the term. Data Center term licences are recognised upfront for the licence portion, with only support spread over time (FY2026 10-K). The same dollar of customer commitment therefore lands as revenue at very different speeds depending on which product the customer buys — a mechanic that becomes decisive in Section 4.

What is being sold, and what is being retired

Revenue reports in two lines. Subscription was $6,262.2 million in FY2026, 95% of the total and up 27%; Other was $310.1 million, 5% and up 9%, of which $21.4 million of the increase was Marketplace revenue (FY2026 10-K). The more informative cut is by deployment:

Source: FY2026 10-K, Note 13. Percentages of FY2026 total are inferred from the disclosed dollars.

Atlassian has been deliberately killing its own product lines for six years, and each retirement forced its installed base to re-contract. Perpetual licences stopped selling in February 2021. Server maintenance ended in February 2024, taking Server revenue from $607.8 million in FY2021 — 29% of that year's revenue — to zero in FY2025 (FY2022 Q4 release; FY2026 10-K). On 8 September 2025 the company announced Atlassian Ascend, the end-of-life of Data Center itself: no new customers from March 2026, no expansions or new term licences to existing customers from March 2028, and maintenance ending in March 2029, with an extended period of roughly three more years for certain customers (FY2026 10-K).

The economics of that swap are worse before they are better, and the filing says so: "Revenues recognized from our Cloud offerings are also typically lower in the initial year compared to our Data Center offerings," and "due to the higher fees associated with hosting our Cloud infrastructure, we have seen and expect to continue to see increased expenses and lower margins as customers transition to Cloud" (FY2026 10-K). The company also states it will offer migration discounts, as it did during the Server transition.

One structural change deserves note because it contradicts the model the company was originally admired for. The FY2016 20-F carried a risk factor headed "We do not have a direct salesforce." The FY2026 10-K says: "we have strategically grown our sales team to focus on expanding the relationships with our existing customers, particularly large enterprises," and "We expand the relationships with our largest existing customers primarily through our direct sales force." The low-touch flywheel still feeds the funnel, but it no longer closes the large deals.

3. Industry, Competitive Position and Moat

Atlassian is usually described as competing in one market. It competes in four, and they have materially different economics.

Agile planning and issue tracking is bought by engineering leadership, expanded bottom-up team by team, then consolidated top-down by procurement. Collaboration and knowledge management is the weakest of the four, because it competes against capability that Microsoft 365 and Google Workspace give away inside an existing contract. IT service management is the strongest: it is bought by a CIO against a defined budget line, the workflows encode audited process, and contracts are multi-year. Source control and DevOps is the most commoditised, because Git itself is a free protocol and monetisation has moved to hosting, compliance and AI assistance.

Where the profit actually pools

The layer that holds price is ITSM. ServiceNow booked $12,883 million of subscription revenue in calendar 2025, growing 21%, with remaining performance obligations of $28.2 billion growing 26.5% — contracted future revenue expanding faster than recognised revenue, which is what pricing power looks like on a balance sheet (ServiceNow Q4 2025 release). That is roughly twice Atlassian's entire revenue. At the other end, the source-control layer is thin: GitLab reported $955.2 million for its year ended January 2026, and Microsoft last publicly disclosed GitHub at roughly $2 billion of annual run-rate in August 2024 and has not re-disclosed it since — including in its FY2026 fourth-quarter release, which cites over 30 million paid Microsoft 365 Copilot seats and no GitHub figure at all.

Within Atlassian's own layer, the most useful evidence is the dispersion between vendors of similar vintage, because it shows that "work management" is not one business:

Sources: each company's own fourth-quarter earnings release. Atlassian does not disclose a net retention or net expansion rate.

Asana's 96% dollar-based net retention is the single most important number in this table, because it is 96% among its customers spending over $100,000 as well — so it is not a small-account churn artefact. Its largest customers are net contracting. monday.com at 110% and GitLab at 118% show that this is not yet a category-wide condition. What it establishes is that a seat-priced work-management vendor with an installed base can shrink inside its own accounts without losing them, which is a specific and testable failure mode rather than a general worry.

Which barriers actually bind

The barrier that binds hardest for Atlassian is the marketplace. Its developer site reports over 6,000 apps from more than 2,000 third-party vendors, with $6 billion of lifetime partner sales. That is a commercial constituency with its own revenue at stake in Atlassian remaining the platform, and it creates a switching cost the customer pays for itself — an enterprise that has bought and configured a dozen Marketplace apps has to re-buy and reconfigure them elsewhere. No entrant reproduces that quickly, and it is the one advantage in this business supported by an independent number rather than an assertion.

The barrier most often claimed for Atlassian — data gravity, the idea that years of accumulated tickets make leaving impossible — is not supported. Asana has data gravity too, and it is running at 96% net retention. Data gravity stops customers leaving; it does not stop them buying fewer seats, and the seat count is where the revenue is.

The barrier working against Atlassian is bundling. Microsoft's Productivity and Business Processes segment runs at roughly $140 billion of annual revenue and has already put Copilot on over 30 million paid seats (Microsoft FY2026 Q4 release). A vendor that can attach a capability to an agreement a customer has already signed does not need to win a product comparison. This is precisely the dynamic that has capped standalone collaboration and knowledge tools, and it is the layer where Atlassian sells Confluence.

Where Atlassian actually stands

The structural evidence says three kinds of company win here, in order: the vendor bundled into an existing enterprise agreement; the vendor that owns an audited, compliance-bearing workflow; and the vendor with a third-party economy attached to it. Atlassian is decisively the third, is genuinely in the second through Jira Service Management, and is not the first. It is the standalone vendor competing against a bundle, in a stack where the deepest profit pool sits with a company twice its size.

One disclosure change is worth recording, because it removes evidence rather than adding it. The FY2016 20-F named its competitors: Microsoft, IBM, Hewlett Packard Enterprise, Google, ServiceNow, Salesforce, Zendesk, Slack and GitHub. The FY2021 20-F still named Microsoft (including GitHub), IBM, GitLab, ServiceNow, Salesforce, PagerDuty, Zendesk, Freshworks, BMC, Google, Asana, monday.com and Smartsheet. The FY2025 and FY2026 10-Ks name no competitor at all — only categories such as "large technology vendors" and "AI-native companies and emerging startups." The only companies now named in the filing are partners: Microsoft, Zoom, Anthropic, Cursor, OpenAI, Salesforce, Workday and Dropbox. An investor assessing Microsoft displacement risk gets no help from the filings and must do it from outside them.

4. Growth Engine

Revenue grew $1,357 million, or 26%, in FY2026. Cloud contributed $963 million of that, Data Center $364 million, and Marketplace $30 million (FY2026 10-K). Growth came almost entirely from customers Atlassian already had: over 90% of FY2026 revenue was attributable to accounts existing on or before 30 June 2025, the same formulation the company has used since FY2016. This is an expansion business, not a customer-acquisition business, and its growth rate is therefore a function of what happens inside existing accounts.

The drivers, ranked

• Cloud seat expansion and price increases — structural, with an unquantified component. The FY2026 10-K attributes subscription growth "primarily" to "paid seat expansion from our existing customers and price increases." Cloud grew 28% for the year and accelerated to 31% in the fourth quarter. The company never quantifies the price component, names the products repriced, or gives an effective date, so how much of Cloud growth is more seats versus higher prices cannot be established from the filings.
• The Data Center wind-down — temporary, and larger than it looks. Data Center grew 25% in FY2026 against guidance of roughly 12.5% given in August 2025, raised to 21.5% in April 2026. The peak was the March 2026 quarter — the one containing the cutoff after which no new customer could buy Data Center — when it grew 44%. Because Data Center term licences are recognised upfront while Cloud is recognised ratably, customers renewing or expanding ahead of a deadline convert into revenue immediately. Management describes the FY2027 consequence in its own words as "the mechanical drag on total revenue growth in fiscal year 2027 as we lap the greater upfront term license revenue recognized in fiscal year 2026."
• Edition and Collection upgrades — management-driven. Atlassian has repackaged from standalone products into Collections (Teamwork, Service, Strategy, Software, Product) and made Rovo, its AI layer, available at no additional cost to standard, premium and enterprise Cloud customers. Higher Collections carry "significantly higher pooled Rovo credit allowances." The AI is therefore monetised indirectly, through edition mix, not as a separate SKU. Service Collection passed $1 billion of ARR growing over 30% (Q3 FY2026 release).
• The enterprise motion — management-driven, and it costs more. Over 85% of the Fortune 500 are now customers, against 291 of them in FY2016. Remaining performance obligations reached $4,817 million at year end, up 44% against 26% revenue growth, with the share expected to be recognised within twelve months falling from 74% to 65% — contracts are getting longer and larger. The cost shows in Section 5.
• Acquisitions — material to cash, undisclosed in revenue. Atlassian spent $1,228.9 million on acquisitions in FY2026: The Browser Company of New York for $488.3 million in October 2025, DX (engineering-productivity analytics) for $720.4 million in November 2025, and two undisclosed deals.

Reported growth versus organic growth

It cannot be decomposed, and this is a finding rather than an omission on the analyst's part. Atlassian has never presented pro-forma results for any acquisition. For FY2026 it states that the results of the acquired businesses "were not material," and that "Pro forma results of operations have not been presented for fiscal year 2026 because the effect of the acquisitions individually and in the aggregate would not be material." The same language covered Loom in FY2024 ($885.6 million) and Mindville and Chartio in FY2021.

The claim is plausible but unverifiable. DX brought $25.5 million of assumed deferred revenue against a $720.4 million price, which is consistent with buying capability rather than revenue. Against that, $2.09 billion went out the door on acquisitions across FY2024–FY2026 — 93% of FY2026 free cash flow in that year alone — and the DX intangible valuation was significant enough to be flagged as a Critical Audit Matter. A transaction large enough to be a Critical Audit Matter is nonetheless declared immaterial to revenue. The 26% headline is most likely close to organic; an investor simply cannot confirm it from the disclosure.

What the guidance says about all of this

For FY2027 the company guides total revenue growth of approximately 13.0%, Cloud growth of 25.5%, Data Center revenue declining approximately 17.0%, and Subscription ARR growth of approximately 18.0% (6 August 2026 release). The headline rate halves. The Cloud rate barely moves. The difference between those two facts is the entire Data Center effect, made explicit by the company itself, and it is the reason a reader should treat FY2026's 26% and FY2027's 13% as describing the same underlying business at two points in a transition rather than as a deterioration.

5. Margin, Cash and Capital Allocation

Gross margin was 84.8% on a GAAP basis in FY2026 and 88% non-GAAP, up from 82.8% and 85% respectively. Software delivered over the internet costs almost nothing to reproduce, so the interesting question is not the level but the direction, and the direction is downward by design. Cost of revenue now explicitly includes "third-party hosting fees, including expenses related to our Cloud infrastructure and AI inference incurred through our customer's usage of our AI tools" (FY2026 10-K), and hosting fees paid to third parties rose $87.0 million in FY2026. The company guides gross margin down to roughly 84.5% GAAP in FY2027 and states that providing AI tools at no or low cost "may further increase our hosting costs without corresponding revenue increases."

Where the money goes

Research and development is the largest expense line in every year of the company's public life, at 50% of revenue in FY2026 and 51% in FY2025, against 45–46% through FY2021. Over half of employees work in R&D. General and administrative has fallen steadily, from 18.7% of revenue in FY2016 to 11%. The line that has moved against the model is marketing and sales: 20% of revenue in FY2024, 22% in FY2025, 24% in FY2026, growing 36% in a year when revenue grew 26%. The filing still describes a "frictionless flywheel with an emphasis on self-service," and that is still true of how customers arrive; it is decreasingly true of what it costs to keep them growing.

The gap between the two profit numbers

Atlassian reported GAAP operating income of $10.4 million in FY2026 and non-GAAP operating income of $1,996.5 million. Of that $1,986 million bridge, $1,605 million — 81$279.5 million) and intangible amortisation ($101.5 million) making up the rest. SBC was 24.4% of revenue, of which R&D alone accounted for $1,144 million. On a GAAP basis the company lost $53.8 million; on a non-GAAP basis it earned $5.85 per diluted share. Which of those an investor treats as real determines almost everything they conclude about this business, and the answer turns on whether the shares issued to pay employees are recovered by the buyback — which is a capital allocation question, addressed below.

Cash conversion is deteriorating

Sources: FY2023 10-K (FY2021 restated to US GAAP), FY2025 10-K, FY2026 10-K. Comparability caveat: Atlassian re-domiciled from the UK to Delaware on 30 September 2022 and moved from IFRS to US GAAP with the FY2023 filings. FY2021 above is the restated US GAAP series; as originally reported under IFRS the same year showed operating income of $101.6m against $141.4m restated, and a net loss of $696.3m against $579.0m. Non-GAAP operating margin was not compiled on a comparable basis before FY2023. SBC percentages for FY2021 are inferred from restated dollars.

Free cash flow fell in absolute dollars in FY2026 — from $1,415.5 million to $1,319.1 million — while revenue rose $1.36 billion. FCF margin has now declined every year since FY2021, from 36.3% to 20.1%. The cause is not stock-based compensation, which is non-cash and rising; it is working capital. Accounts receivable consumed $484.5 million of cash in FY2026 against $150.0 million the year before, and the balance rose 63% against 26% revenue growth. Deferred revenue, which used to fund the business, contributed only $153.3 million against $552.3 million in FY2024. Contract assets went from $3.3 million to $53.5 million.

These three movements have one explanation. As revenue shifts from cards settling in three days and small annual-in-advance subscriptions toward invoiced enterprise and channel contracts recognised on longer schedules, the float that made this business self-funding drains away. RPO growing 44% while deferred revenue grows 7.3% says the same thing from the other side: the commitments are real and growing, but they are increasingly unbilled. This is a mix effect rather than a deterioration in demand, and it is the reason FY2026 free cash flow sat $677 million below non-GAAP operating income when in FY2025 it sat $128 million above it (inferred).

What management did with the cash

Ranked across FY2024–FY2026: share repurchases $2,975 million, acquisitions $2,091 million, capital expenditure $112 million, strategic investments $51 million, dividends nil. Capex is 0.5% of revenue — this business consumes almost no fixed capital, and every allocation decision is therefore a choice between buying companies and buying back stock.

FY2026 was aggressive on both. Atlassian spent $1,800 million repurchasing 19.1 million shares at an average of $94.31, having paid $196.02 the year before, and $1,229 million on acquisitions — $3.03 billion against $1.32 billion of free cash flow. Net cash fell from $1,949 million to $250 million. Goodwill and intangibles reached $2,735 million, 2.58 times shareholders' equity (inferred), with no impairment taken.

The buyback arithmetic is the part worth doing carefully. Shares outstanding fell 9.6 million, or 3.64%, to 253.4 million — the first material shrink in the company's history. But the buyback retired 19.1 million shares, which means roughly 9.5 million were issued, equal to 3.62% of the opening count against 2.47% the prior year (inferred). Gross dilution accelerated; it was simply outrun. And it was outrun at $94.31 a share. The same 19.1 million shares at FY2025's average price would have cost $3.74 billion (inferred), nearly three times free cash flow. The ability of this buyback to absorb stock-based compensation is a direct function of the share price, which is not a mechanism an investor should treat as durable.

Two further points on allocation behaviour. The FY2026 acquisitions carried $345.9 million of retention consideration on top of the $1.21 billion of purchase price — $105.2 million of restricted stock for The Browser Company, $201.8 million of restricted stock plus $38.9 million of deferred cash for DX (inferred sum) — which is expensed through compensation rather than capitalised, and is part of why gross issuance stepped up. And the FY2026 10-K contains no subsequent-events note and discloses no pending transaction as of 14 August 2026, so nothing in this section post-dates the reported figures.

6. Cyclicality, Constraints and What to Monitor

Atlassian is not cyclical in the way an industrial is. Revenue is recurring, no customer exceeds 5% of the total, and geography is spread across the Americas (48%), EMEA (41%) and Asia Pacific (11%). What it has instead is exposure to a single variable: the number of people its customers employ to do knowledge work. Seats are bought for headcount, and the risk factor concedes the transmission mechanism directly — economic uncertainty "could cause customers to reduce the number of personnel providing development or engineering services," and "we may see declines in expected spending from new customers or renewals and reductions in paid seats from existing customers" (FY2026 10-K). The filing frames this as cyclical. Section 7 takes up whether it is only cyclical.

Where the business sits right now

On three different measures, FY2026 sits at three different points in its own history, and reading them together is the point.

Revenue growth of 26% is a mid-cycle figure that had re-accelerated from the FY2025 trough of 20%, against 43% in FY2016 and 34% in FY2022. It is guided to 13% in FY2027 — the lowest in the company's public life. Non-GAAP operating margin of 30% is an all-time peak, up from 23% in FY2024 and 25% in FY2025, and it is guided down five points to 25%. Free cash flow margin of 20.1% is the trough of the available series, having fallen from a 36.3% peak in FY2021.

A peak reported margin sitting on a trough cash margin, with the growth rate about to halve, is a specific and unusual combination. It resolves once the composition is understood: the margin peak is flattered by upfront Data Center revenue and by a restructuring whose $285.4 million of charges is excluded from the non-GAAP figure, while the cash trough reflects the working-capital drain of the same enterprise mix shift. Neither number is wrong; they are measuring different halves of a transition.

Management's own commentary carries no caution. The August 2026 release contains no mention of budget scrutiny, elongated sales cycles or macro pressure; the CFO describes enterprises "deepening their commitment." Every hedge lives in the risk factors: elongated sales cycles, budget freezes, pricing pressure from larger customers, and the observation that "some customers may also view our offerings as discretionary purchases." The gap between the two registers is normal, but an investor should read the 10-K rather than the release for demand conditions.

Durable versus borrowed

Sources: FY2026 10-K; 6 August 2026 earnings release; Atlassian developer site (Marketplace figures).

What to monitor, and where it is published

7. Risks, Unknowns and Questions for Deeper Work

The risks below are those that cyclicality does not capture. They are ranked by how much they compound with each other rather than by likelihood.

• The seat model and AI point in opposite directions, and the filings do not engage with it. Atlassian charges per person. AI tools raise output per person: the JetBrains 2025 survey of 24,534 developers found 85% regularly use AI for coding and one in five saves eight or more hours a week. If firms hold output constant and slow hiring, seat counts flatten without a single customer leaving — and net retention drifts toward 100%, which is precisely the pattern Asana is already reporting at 96%, including among its largest accounts. The FY2026 10-K nowhere addresses whether AI teammates substitute for human ones. Its only relevant sentence — "an increased prevalence of AI may impact the work practices of software teams… and therefore our market opportunity" — names no mechanism and no direction, and is carried over verbatim from FY2025. Management's stated theory runs the other way: AI increases work volume, and Atlassian is the "orchestration layer" for a "blended workforce."
• The counter-evidence is real and should be weighed. The Stack Overflow 2025 survey of over 49,000 developers found that 69% do not plan to use AI for project planning and 76% reject it for deployment and monitoring — resistance is highest in exactly the workflow layer Jira occupies. US Bureau of Labor Statistics projections published in August 2026 still show software developer employment rising from 1.72 million to 1.89 million by 2035. Where headcount reductions have been publicly attributed to AI, the confirmed cuts have been in customer support rather than engineering. The seat-compression thesis is plausible and unresolved, not established.
• Platform dependency on companies that also compete, governed by nothing. Atlassian's value depends on integrating with third-party applications "including the applications of software providers that compete with us," and the filing states plainly: "To date, we have not relied on long-term written contracts to govern our relationship with these providers." It enumerates the failure modes itself — a provider discontinues API access, changes its fees, "establishes more favorable relationships with one or more of our competitors," or "develops or otherwise favors its own competitive offerings over ours." This is the Microsoft risk, and because the 10-K names no competitors it is never stated as such.
• Over half of revenue reaches Atlassian through parties who owe it nothing. Channel partners generated over 50% of FY2026 revenue. Their agreements are non-exclusive, they may sell competing products, and they "may also cease marketing our offerings with limited or no notice and with little or no penalty." Atlassian also carries their credit risk and their conduct risk. A distribution channel that is simultaneously the majority of revenue and terminable without penalty is a structural exposure that has grown quietly — it was roughly 25% of revenue in FY2016.
• The marketplace that is the moat is also the attack surface. Third-party apps execute inside customer tenancies, and the filing concedes they "may not meet the same quality standards that we apply to our own development efforts," that "our ability to mandate security standards and ensure compliance by these third parties may be limited," and that third-party apps "have caused disruptions affecting multiple customers" in the past. Separately, Atlassian discloses that it has experienced breaches resulting in unauthorised access to customer data, and that it has been "a target of security threats, including from state actors." A serious Marketplace-originated breach would damage the ecosystem and the product at the same time.
• Concentrated control with no realistic check. Class B shares carry ten votes to Class A's one. As of 30 June 2026, Class B holders controlled approximately 86% of voting power, and the two co-founders hold substantially all of it — 42.59% each as of the 2025 proxy. The filing acknowledges this "will limit the ability or preclude the ability of our stockholders to influence corporate matters." Vanguard, the largest outside holder, controls 1.42% of the vote.
• Leadership has turned over almost completely during the transition. Scott Farquhar stepped down as co-CEO in August 2024. The President left in December 2025, the CTO in March 2026. James Chuong became CFO in March 2026, replacing Joe Binz; a new Chief Product and AI Officer was appointed in FY2026 and a new Chief Product Officer for Enterprise in August 2026. Roughly 10% of the workforce was eliminated in March 2026. Executing a dated end-of-life of 28% of revenue, an AI repositioning and an enterprise sales build simultaneously is demanding with a settled team.

What the sources could not answer

These are gaps in the disclosure, not in the research, and each is a question an investor would need resolved before forming a thesis.

• Net revenue retention is not disclosed, in any period. Neither is a paid seat count, a price per seat, or gross churn. For a per-seat business, the absence of a net expansion rate is the most consequential gap in the filings — and it is the exact metric that would test the seat-compression thesis above.
• Migration progress is unquantified. There is no disclosure of how many Data Center customers remain, how much Data Center ARR is still to convert, or what proportion has migrated — only the calendar dates and a 17% FY2027 decline guide.
• AI monetisation is undisclosed, and a metric was dropped. The FY2025 fourth-quarter release cited 2.3 million AI monthly active users. The FY2026 release does not restate or update it, substituting a narrower figure — one million monthly active users of the MCP server and Teamwork Graph CLI. The two are not comparable. No AI revenue, Rovo seat count or attach rate is given anywhere.
• Price increases are cited as a growth driver in both FY2025 and FY2026 without magnitude, product or effective date, so the durability of that contribution cannot be assessed.
• Product-level revenue does not exist. Atlassian reports as a single operating segment and discloses only that Jira, Confluence and Jira Service Management together are "a majority" of revenue. The relative health of the three cannot be observed.
• Gross margin is not disaggregated. There is a single cost-of-revenue line, so Cloud and Data Center unit economics cannot be separated — which matters when the company states Cloud margins are structurally lower.
• Neither the shareholder letters nor the earnings call transcripts were available in the archive, so all management commentary in this memo comes from press-release text. The Q&A, where demand nuance usually surfaces, has not been read.

8. Investor Takeaways

• What this business really is: the system of record for how software and IT teams coordinate work, sold per seat, embedded deeply enough that over 90% of each year's revenue comes from last year's accounts.
• The core economic engine: near-zero marginal delivery cost against per-seat pricing, expanded inside existing customers rather than won from new ones — with the caveat that half of every revenue dollar now funds R&D and a further quarter funds sales.
• The main growth lever: Cloud seat expansion and edition upgrades. Everything else in FY2026 — the Data Center pull-forward above all — is a transition effect that reverses next year, which is why 26% becomes a guided 13%.
• What could break the story: a per-seat model meeting a technology that reduces the number of people needed per unit of work. The evidence is genuinely unresolved, the company does not engage with it, and a peer is already reporting 96% net retention among its largest accounts.
• What to monitor: Cloud revenue growth once Data Center stops helping, the >$10,000 Cloud ARR customer count, and free cash flow margin against the receivables build — three numbers that together say whether the seats are still expanding.