Atlassian (TEAM) Q4 FY26: Data Center Migrations Contributed Nothing to the $49 Million Cloud Beat
The headline numbers were large enough to move a $29 billion company 37% in two sessions. Total revenue of $1.766 billion, up 28% against a $1.66 billion consensus. Adjusted EPS of $1.87 against $1.50. Cloud revenue of $1.213 billion, up 31%, with growth accelerating rather than decaying. GAAP operating income of $211 million against a $28 million loss a year earlier. None of that is the number worth anchoring on.
The number is $49 million, and it is the composition of the cloud beat rather than its size. Asked on the call where the upside came from, CFO James Chuong volunteered that Data Center migrations were in line with internal expectations and contributed none of it. The overperformance came from cross-sell into the Service and Teamwork Collections and from seat expansion inside existing Jira and Confluence estates. Migrations, he added, are running at mid-to-high single digits of contribution to cloud growth for the full year.
That decomposition attacks the standing bear case directly. For four years the objection to Atlassian has been that cloud growth is a bookkeeping artifact of forced migration. Server was killed, Data Center is being killed, and the same customer paying the same money moves from one revenue line to another while the company books it as growth. On that reading, cloud growth has a terminal date: when the migration pool empties, the growth rate collapses to whatever the underlying business was doing, which was not 31%. If migrations carry seven points of a 31% cloud growth rate, the organic component is running near 24%. That is the figure that governs FY28 and beyond, and it is the first quarter where management has separated it cleanly enough to be checked.
The FY27 guide then reads very differently from how the tape initially processed it. Total revenue growth of roughly 13%, down from 26%, looks like a deceleration bad enough to override the print. It is mostly an accounting shadow. The September 2025 announcement that Data Center reaches end of life in March 2029 triggered two effects simultaneously: term license renewals pulled forward out of future periods into FY26, and ASC 606 recognized a larger share of those licenses upfront rather than ratably. FY26 Data Center revenue was inflated by both. FY27 Data Center revenue is guided down 17% against that inflated base, and the drag lands on consolidated growth even as cloud compounds at a guided 25.5% and rises to roughly 74.5% of the total. Management expects total revenue growth to re-accelerate in FY28 as the comparison laps.
Management’s response is to redirect attention to Subscription ARR, which finished at $6.606 billion, up 23%, guided to 18% for FY27. The redirection is analytically correct and self-serving at the same time. ARR does normalize the ASC 606 timing distortion. It also happens to be the metric that looks best, and the 18% guide is still a five-point deceleration that no accounting mechanic explains away. The honest read is that both numbers are understating the business and the gap between them is where the argument lives. Remaining performance obligations of $4.817 billion, up 44%, grew at nearly double the ARR rate, which is the strongest forward signal in the release and also the most easily misread. RPO expands when contract duration lengthens, and duration lengthens precisely when Data Center customers sign multi-year cloud commitments to cover their migration window. Some meaningful part of that 44% is tenor rather than demand. It is still difficult to construct a version of 44% RPO growth that resolves into 18% ARR growth without management having sandbagged.
The competitive position is the part of the story that has actually changed, and it is not the product list. Atlassian’s defensible asset is the Teamwork Graph, the accumulated relational record of how work moves through an organization: which ticket blocks which, who owns what, which document justified which decision, what happened the last time this broke. Management quantified the advantage as up to 44% more accurate answers on 48% fewer tokens, which is the correct way to frame it. Context that already exists is cheaper to retrieve than context that must be reconstructed by a model at inference time. Net revenue retention north of 120%, Rovo adopters expanding ARR at more than twice the rate of non-adopters, $3 million ARR customers up over 50% and $5 million ARR customers up over 70% are all downstream of the same mechanism. Cannon-Brookes framed it as context being the edge, hard to build and impossible to hire. That is true and it is also the most durable thing the company owns, because a competitor can clone Jira’s feature set in a quarter and cannot clone eleven years of a customer’s issue history at any price.
The complication is that Atlassian is now exposing that graph through an open protocol. Its MCP server and Teamwork Graph CLI passed one million monthly active users and more than doubled in a single quarter, and the company signed or expanded partnerships with Cursor, Claude and GitHub. Read one way, this makes Jira the substrate that every coding agent has to call, and the graph becomes infrastructure. Read the other way, standardizing access to your moat through a protocol anyone can implement turns a proprietary interface into a commodity endpoint, and the value migrates to whoever owns the agent rather than whoever owns the data. Both readings are live. What tilts it is that the graph itself stays proprietary regardless of how it is accessed, and the access layer was never where the pricing power was.
Pricing power in AI is the one place the quarter is quiet. Jira Coding Agent and the Claude Code, Cursor and Copilot integrations ship at no additional cost to paid Jira Cloud customers. Rovo adoption is disclosed in usage metrics, over 80% of the Fortune 500, assisted actions up 50% sequentially, service calls up over 400%, and never in dollars. There is no AI revenue line because AI is currently being spent rather than sold. It functions as a retention and expansion vector, which is defensible while the seat model holds and becomes a problem if agentic development compresses engineering headcount at customers priced per human. Atlassian’s answer, that agents consume seats too, is untested. The quarter provides no evidence either way.
Margins support the patience. Non-GAAP operating margin expanded twelve points to 36%. GAAP operating margin reached 12% against negative 2%, with FY27 guided to stay positive on GAAP. Free cash flow was $475 million in the quarter at a 27% margin and $1.319 billion for the year at 20%. FY26 still closed with a small GAAP net loss of roughly $54 million, a 79% improvement, so the profitability turn is one quarter old at the consolidated level.
On the stock, this is a recovery from a low base rather than a breakout. TEAM traded as low as $81.43 on July 24 and closed Thursday at $110.28, down about 32% year to date and roughly 36% over twelve months. Friday’s close at $151.08 puts the market capitalization near $39.7 billion, roughly 5.4 times guided FY27 revenue of $7.42 billion and about 30 times trailing free cash flow. Consensus targets clustered between $134 and $140 going into the print and are now below spot; Jefferies moved to $200 on backlog acceleration and relative cheapness, against a standing low near $95 to $100 from Wolfe. The base case has cloud compounding near 25%, the Data Center comparison lapping, and FY28 revenue growth returning to the high teens, which supports $150 to $180 without multiple expansion. The bull case requires the 18% ARR guide to prove as conservative as prior guides did and GAAP margin to clear the mid-teens, which gets to $200 to $230. The bear case is not company-specific at all: enterprise SaaS derates as a cohort on agentic seat-displacement fears, Salesforce and ServiceNow and Monday drag the group’s multiple down together, and TEAM revisits $95 to $110 having executed perfectly.
The line to watch is inside the guide itself. Data Center revenue is forecast to decline about 4% in the September quarter and about 17% for the full fiscal year. The easy comparison is the one being reported first, and the decline has to steepen into the low twenties across the remaining three quarters to reach the annual figure. Anyone underwriting the FY28 re-acceleration is buying three consecutive quarters of worsening optics to get there.