Atlassian’s stock jumped roughly 30% on Friday May 1 after the company reported Q3 FY26 revenue of $1.79 billion, up 32% year over year, against a sell-side estimate of $1.69 billion. Adjusted EPS of $1.75 beat the $1.32 consensus by 33%. Cloud revenue grew 29% to $1.13 billion. Data Center revenue hit $561 million versus an expected $515 million. Full-year cloud and data-center guidance was raised. Going into the print, the stock was down about 45% year to date and 84% off its 2021 peak.
Six weeks earlier, on March 11, the same company cut 1,600 employees — about 10% of its 16,000-person workforce. The press release used a specific phrase: the layoffs would “self-fund further investment in AI and enterprise sales.” The restructuring carried a $225–236 million charge, with most of the cost landing in the same Q3 the market just rewarded with a 30% rally. The print’s net loss widened to $98 million from $71 million a year earlier; the beat happened with the layoff bill embedded.
Two things are happening at once inside this company, and they are pointed in opposite directions on the hypothesis.
The bear case showed up in the call. CFO James Chuong told investors that Data Center customers actively transitioning to the cloud are “moderating their seat expansion versus historical trends” — the cohort closest to the AI surface area is buying fewer seats than the same customers used to buy at this stage. That is the exact mechanism this blog has been tracking since The Seat Compression, and it is now visible at the canonical seat-based vendor on a quarter where the headline beat consensus by six percent.
Now the falsifying side. Atlassian’s agent product, Rovo, shipped credit-based AI on top of the seat SKU. CEO Mike Cannon-Brookes said on the call that customers using Rovo are “growing their ARR at roughly two times the rate of customers not using Rovo,” that Rovo credit usage is up more than 20% month-over-month, and that customers on the Teamwork Collection bundle — which ships with ten times more Rovo credits than the standalone subscription — use more than three times as many credits per user and run more than twice as many active agents as customers on the standalone tier. Two-times ARR growth for the cohort using the new SKU is the strongest internal monetisation signal a seat-based incumbent has produced in this cycle. It is also the answer to the question The Foolish Quarter left open about SAP: whether the consumption layer can grow fast enough to matter while the seat base is still compressing.
The mechanism here is not “AI compresses seats” or “AI doesn’t compress seats.” It is more specific. A seat-based application vendor cut human op-ex by 10% at peak revenue, used the proceeds to fund an agent SKU sitting on top of the seat product, and produced a quarter where the seat compression is real for the migrating cohort and the agent SKU is compounding at 20% a month. The bifurcation The Great Sorting located across companies is now visible inside one company, on one income statement, in one quarter.
The market repriced in both directions across 51 days. The 45% year-to-date drawdown going into the print was the phantom repricing — investors front-running a thesis the actual numbers had not yet confirmed. The 30% snap-back is the partial unwind once the thesis met the quarter that had absorbed the worst of the costs and still grew 32%. Neither price was wrong about the mechanism. They were wrong about the speed.
This complicates the hypothesis without falsifying it. Software value is being lost to AI at the seat layer where Atlassian is most exposed — exactly as the hypothesis predicts. Software value is being captured at the credit layer where Atlassian shipped first — which the hypothesis does not predict, because the original framing treats AI as the destroyer rather than as the SKU on top of the destroyed thing. The vendor that survives this is the vendor that runs both lines on the same P&L, with the cost structure sized to the new mix. Atlassian funded the transition by removing 1,600 jobs. The market noticed the funding worked.
Two falsification risks remain visible in the print. The first is that Q3 carried unusually clean tailwinds — the layoff cost sits in operating expenses where it is partially offset by the headcount reduction itself, and the FY guidance raise is small enough that the next two quarters could give it back if Rovo credit growth normalises. The second is that the moderation in Data Center seat expansion is described as a transition effect for the cloud-migration cohort. If that moderation extends to net-new cloud cohorts in Q4, the bear case stops being a Data Center story and becomes a cloud story, which is most of the revenue.
What to watch. HubSpot reports Q1 on May 7, the first quarter to land after the company moved Customer Agent to $0.50 per resolved conversation — the cleanest test of whether outcome SKUs land in the headline numbers when seat-based revenue is still the base. Rovo’s 20%-month-over-month credit growth is not a number that holds for many quarters at any scale; the next print will show whether it has decayed to a sustainable rate or compressed faster than expected. And any Q4 update to the seat-expansion language — particularly whether it stays scoped to migrating Data Center customers or extends to cloud-native customers — will determine whether Atlassian’s path through this is a template or a one-off.
The hypothesis tracker moves a small amount on this print. Verifying evidence: seat compression is now visible at the textbook seat-based vendor on a beat-the-quarter print. Falsifying evidence: the same vendor shipped an agent SKU that is monetising the transition fast enough to push headline growth higher, not lower, while the company self-funded the pivot through a cut deep enough to register on the income statement. The mechanism the hypothesis tracks is correct. The shape of the outcome is more company-specific than the hypothesis assumed.
Sources: Atlassian Q3 FY26 earnings — CNBC; Q3 FY26 earnings call transcript — Motley Fool; March 11 layoff coverage — CNBC; Restructuring charge detail — Computerworld; Atlassian’s March announcement.