Atlassian's collapse isn't really about Atlassian
When beating earnings by 67% still tanks your stock, something fundamental has shifted in how markets price software companies
The numbers don’t lie (they just confuse)
Atlassian posted its best quarter in history. Revenue hit $1.59 billion. Cloud revenue crossed the $1 billion mark for the first time. Earnings per share came in at $1.22, against analyst estimates of $0.73.
The stock responded by losing half its value.
I’ve seen plenty of earnings misses that cratered stocks. I’ve seen revenue warnings send share prices into free fall. Those make sense. The cause-and-effect is obvious.
But I can’t recall the last time a company beat EPS estimates by 67%, crossed two major revenue milestones in the same quarter, and then watched $20 billion in market cap evaporate over the following weeks.
If you’re confused, good. You should be.
The story here is bigger than Atlassian. An entire sector is getting repriced in real time, and Atlassian happens to be standing closest to the blast radius. It’s the canary. The coal mine is the entire enterprise software industry.
Atlassian beat every estimate Wall Street had for it. The market punished it anyway. The disconnect tells you everything about where investor sentiment has shifted.
To understand how we got here, we need to look at what Atlassian actually reported, why the market didn’t care, what the insiders were doing while all this unfolded, and what a trillion-dollar software selloff tells us about the next decade of enterprise tech.
What a record quarter actually looks like
Atlassian’s Q2 FY2026 earnings, reported February 5, were genuinely impressive. Revenue came in at $1.59 billion. The $1.22 EPS figure crushed the $0.73 consensus estimate by 67%.
Cloud revenue, the metric investors have been told to watch for years as Atlassian shifts from on-premise to subscription, crossed $1 billion in a single quarter for the first time. That’s a 26% year-over-year jump. Remaining performance obligations (RPO), which measures contracted future revenue, hit $3.8 billion, up 44% year-over-year. Net revenue retention stayed above 120% for the third consecutive quarter, meaning existing customers are spending more, not less.
The company crossed $6 billion in annual run rate. Over five million users were actively using Rovo (myself included), Atlassian’s AI assistant, which didn’t even exist eighteen months ago.
RPO accelerating to 44% growth while revenue grows at 23% is the kind of forward indicator most SaaS companies would kill for. It means the sales pipeline is strengthening, not weakening.
By every metric the industry has spent the last decade telling investors to care about, Atlassian had an exceptional quarter.
So why did the stock drop 6.31% after hours?
So why did it tank?
The after-hours drop on February 5 was just the appetiser. By the end of that week, TEAM was down 20%. Then it kept falling.
The surface-level explanations circulated quickly. Data Center revenue, the on-premise product line that Atlassian has been deliberately winding down, showed signs of faster-than-expected decline. Management used the phrase “seat compression” on the earnings call, referring to customers buying fewer per-user licenses. And despite the massive EPS beat, Atlassian remained GAAP unprofitable, posting a net loss of $42.6 million for the quarter, with share-based compensation eating up roughly a quarter of revenue.
These are valid concerns. But they’re also concerns that existed last quarter, and the quarter before that. Data Center decline has been telegraphed for years. GAAP profitability has been the company’s white whale for a decade. Seat compression is a fancy way of describing what happens when companies get better at managing their software licences.
None of this explains a 50% share price collapse.
The market wasn’t looking at what Atlassian did. It was looking at what AI might do to it.
The earnings call could have been the best in Atlassian’s history. It probably was. But the market had already moved on to a different question: does Jira still matter in a world where AI can manage projects?
And that question wasn’t being asked about Atlassian alone.
Welcome to the SaaS-Pocalypse
On February 4, one day before Atlassian’s earnings, Reuters reported that Anthropic’s announcement of a Claude-powered legal tool had triggered roughly $830 billion in losses across global software stocks in just six trading sessions.
Eight hundred and thirty billion dollars. Gone. Because an AI company showed it could do legal research.
The logic, if you can call it that, ran like this: if an AI tool can replace parts of what legal software does, then AI tools can presumably replace parts of what all enterprise software does. Project management, collaboration, customer relationship management, HR software, accounting platforms. Everything becomes a target.
Fortune reported on the emerging “SaaS-Pocalypse” on the same day. The S&P 500 Software & Services Index dropped 20% year-to-date. More than 90% of software components were trading lower.
Nine out of ten software stocks, down. Not because they had bad quarters. Most of them, like Atlassian, were posting solid results.
Citi’s analysts put a finer point on it. In a downgrade that knocked TEAM to $118.55 on January 16, they specifically flagged “AI solutions and private code generation” as existential threats to Jira and Confluence. The argument: why pay per-seat licensing for a project management tool when an AI agent can coordinate tasks, generate documentation, and track progress at a fraction of the cost?
In hindsight, the dominoes fell in a predictable order.
It’s a reasonable question. It’s also, right now, almost entirely speculative. No AI agent is currently replacing Jira at any meaningful scale. The companies buying Atlassian’s products aren’t ripping them out and replacing them with ChatGPT. Enterprise software transitions take years, sometimes decades. (Ask anyone who’s tried to get a Fortune 500 company off SAP.)
But markets don’t price the present. They price the future. And the future, according to the great SaaS unbundling thesis, looks like AI doing most of what traditional SaaS does, without the per-seat pricing model.
The SaaS-Pocalypse wiped nearly a trillion dollars from software stocks. The trigger wasn’t a product failure or an earnings miss. It was a demo.
Whether that thesis is right or wrong is almost beside the point. The repricing is happening now.
The insiders saw it coming
While the market was digesting the AI threat thesis, the people who know Atlassian best were selling. A lot.
Co-founders Mike Cannon-Brookes and Scott Farquhar each sold 866,145 shares over the six months leading up to the crash, pocketing roughly $134 million apiece. Farquhar’s selling pattern was particularly striking: ten consecutive identical sales of 7,665 shares at $94.81, executed like clockwork between January 12 and February 6.
On February 18, CFO Joe Binz announced his retirement, effective March 30. His replacement, James Chuong, was named the same day.
The next day, February 19, the CFO, CTO, and CRO all sold shares.
Now, insider selling doesn’t automatically mean insiders expect the stock to collapse. Executives sell shares for all sorts of reasons: diversification, tax planning, buying a third house in Noosa. Pre-arranged 10b5-1 trading plans exist specifically so insiders can sell without it looking like they’re dumping stock based on inside knowledge.
But the optics are brutal. When nearly $300 million in insider sales precedes a 50% stock collapse, retail investors notice. And they get angry.
Angry enough that Pomerantz LLP announced a securities fraud investigation in late January. I should note: Pomerantz specialises in exactly this kind of ambulance-chasing class action. They announce investigations the way fast food chains announce new menu items. The legal merit of the investigation is, at this stage, entirely unproven.
The institutional investors were less dramatic but arguably more telling. UBS Asset Management slashed its position by 76.4%, dumping 8.76 million shares. Artisan Partners exited completely. So did WCM Investment Management. And Coatue Management. Sands Capital cut 94.8% of its holdings.
These are institutional investors with research teams and quarterly portfolio reviews, not retail traders panic-selling on Reddit. When five major institutions independently decide to bail on the same stock within weeks of each other, that’s a signal.
Five major institutional investors independently decided to exit or drastically cut their Atlassian positions. Less panic, more coordinated reassessment of the entire sector.
Wall Street says buy, the market says sell
And yet.
As of mid-February, the analyst consensus on Atlassian was 21 Buy ratings, 4 Outperform, 8 Hold, and zero Sells. Not a single analyst on Wall Street had a Sell rating on a stock that had lost half its value in eight weeks.
The mean price target sat at $177.85, implying 134% upside from the February 21 close of $75.98. Morgan Stanley still had a $290 target, calling the current price an “attractive entry point.” The Motley Fool ran an article asking whether Atlassian was now at its cheapest valuation ever and whether it was time to buy.
But every single analyst also cut their targets. Oppenheimer dropped from $275 to $150, a 45% reduction. BTIG went from $220 to $140. Macquarie from $250 to $150. TD Cowen cut to $140 specifically citing AI concerns. BMO Capital sat at $135.
Analysts are saying, in effect: “The company is doing great, the stock is cheap, you should buy it, and also we’re cutting our price targets by 30-45% because we have no idea where the floor is.”
The disconnect between analyst conviction and market behaviour is either a screaming buy signal or evidence that the traditional analyst model doesn’t know how to price AI disruption risk. Probably both.
Morgan Stanley at $290 on a stock trading at $69. Less a price target than a declaration of faith.
The timeline of a slow-motion crash
Zooming out, the destruction happened in stages. Each one felt survivable in isolation. Together, they paint a picture of a stock in genuine free fall.
Atlassian started 2026 at around $143. Already down from its 2021 all-time high of $483, but still a $37 billion company.
January 16: Citi published its downgrade, flagging AI disruption risks. The stock dropped to $118.55.
February 3: Pre-earnings, trading around $113. The market was already nervous.
February 5: Earnings after hours. Despite the blowout quarter, the stock fell to $98.41.
February 10: End of that catastrophic week. Down to $91.
February 17: New 52-week low at $82.60.
February 21: $75.98.
February 23: $68.81.
That’s a 52% decline from the start of the year. An 80% decline from the all-time high. Market capitalisation shrank from roughly $60 billion at peak to about $20 billion.
For context, $20 billion is roughly what Atlassian was worth in 2019. Five years of growth, five million AI users, $6 billion in annual run rate; all of it priced away as if it never happened.
(I keep coming back to that $6 billion run rate figure. It’s not a projection at all. It’s actual contracted revenue. The company is collecting that money right now. And the market is saying it’s worth three times annual revenue. For a company growing at 23%.)
What this suggests
The median SaaS enterprise value to revenue multiple has compressed from 18-19x to 5.1x. Atlassian, at around 3x price-to-sales as of its February 23 close, is trading at its cheapest valuation in its history as a public company.
This is a sector-wide repricing, not a single-company meltdown. Atlassian’s fundamentals didn’t change between January 1 and February 23. The products didn’t get worse. The customers didn’t leave. Revenue kept growing. What changed was the framework investors use to value software companies.
For the last fifteen years, the story went like this: SaaS companies sell subscriptions, subscriptions are sticky, per-seat pricing scales with customer growth, switching costs are high, therefore SaaS companies deserve premium multiples. It was a good story. It was mostly true.
The new story goes: AI can replicate core SaaS functionality, AI doesn’t charge per seat, switching costs drop when the replacement is an API call, therefore SaaS multiples need to come down. A lot.
Both stories contain truth. Both also contain a lot of assumption.
The old story assumed switching costs would stay high forever. They probably won’t.
The new story assumes AI can replace enterprise software quickly and cheaply. It probably can’t; not for years, and not completely.
But the market doesn’t wait for nuance. It reprices now and figures out the details later.
The market has gone from pricing SaaS companies at 18x revenue to 5x revenue in less than a year. That’s a regime change, not an adjustment.
For Atlassian specifically, the question comes down to this: is Jira more like a utility or more like a feature? If Jira is infrastructure that teams build their workflows around (like email, or Slack, or Git), then the AI disruption threat is overstated and the stock is absurdly cheap. If Jira is a feature that can be replicated by an AI agent for a tenth of the cost, then the current price might still be too high.
I don’t know the answer. Neither does the market. That uncertainty is precisely what’s being priced in.
What I do know is that the reaction to Atlassian’s best-ever quarter tells us something important about where we are in the AI hype cycle. The market isn’t evaluating software companies on their current performance anymore. It’s evaluating them on how exposed they look to a technology that, in most enterprise applications, is still closer to demo than deployment.
Atlassian may recover. It may not. But the SaaS-Pocalypse, ridiculous name and all, represents a genuine structural shift in how investors think about software. The companies that survive this repricing will be the ones that prove AI makes their products more valuable, not less. Atlassian’s five million Rovo users suggest they’re trying.
Whether trying is enough is the $40 billion question.
I write about the forces reshaping the software industry regularly. If this kind of breakdown is useful to you, consider following so you don’t miss the next one.