Atlassian's "No Sales Team" Story Hides How It Really Makes Money

In 2015, when Atlassian filed to go public, it wrote down something almost no software company would dare admit on the way to an IPO: 'We do not have a direct salesforce and our sales model does not include traditional, quota-carrying sales personnel.'⁴ The line became legend — a software company that sold itself, no reps, no cold calls, just a product so good engineers bought it on a credit card. It is a wonderful story. It is also the wrong place to look. The genius of Atlassian was never the missing sales team. It was where the money quietly compounds while everyone admires the empty sales floor.

The official story is that Atlassian is a product-led-growth company that wins by being too useful to need selling. The real story is that Atlassian is a lock-in machine — it lands a team cheaply, lets the team's dependence on it deepen, taxes the ecosystem that grows around it, and is now deliberately demolishing every door a customer could use to leave. The absence of salespeople was never the moat. The structure underneath them was.

Ninety percent of the revenue arrives whether or not anyone sells

Start with the shape of the money. In fiscal 2024 Atlassian booked $4.358B in total revenue, and $3.924B of that — about 90% — was subscription revenue, growing 34% year over year.¹ That is not the income statement of a company that lives off new logos. It is the income statement of a company that lives off the same customers spending more, every year, almost on autopilot. The company says so itself: over 90% of FY2024 revenue came from customer accounts that already existed before the year even began.² You don't need a quota-carrying rep to grow a number that grows from the inside.

This is the first layer of the machine, and it is the boring one. A five-person team adopts Jira to track bugs. The team becomes thirty. The thirty adopt Confluence to write things down. The workflows ossify, the integrations multiply, the institutional memory moves in — and the seat count, the product count, and the price per seat all drift upward without anyone being sold anything. The proof is in the high end: 45,842 customers now spend more than $10,000 a year in Cloud, up 18%, and the count of customers spending over $1M a year grew 48% in a single year.³ Land small, expand relentlessly. The flywheel runs on dependence, not persuasion.

The toll booth on an ecosystem it doesn't build

The second layer is the one most people get backwards. Around Jira and Confluence has grown a vast bazaar of third-party apps — the Atlassian Marketplace — and the gross revenue flowing through that ecosystem is estimated at roughly $1.8B a year across 1,800-plus partners.⁷ Atlassian writes almost none of those apps. It owns the storefront. So it takes a cut, and that cut shows up in its filings as the modest 'Marketplace and other' line of $273.6M.¹ Here is the move people miss: those are two completely different numbers. The $1.8B is what the ecosystem earns. The $273.6M is Atlassian's tax on it.⁷ Confuse them and you either overstate Atlassian's storefront business sixfold, or you miss the leverage entirely.

And the tax rate is rising. Beginning in 2026, Atlassian's commission on Connect apps climbs from 15% to 20%, then to 25% by mid-year; the rate on its newer Forge apps creeps from 15% toward 17%.⁸ Notice what that requires. You can only raise the toll on a road no one can leave. The app vendors are as locked in as the customers — their entire business sits on top of Jira's user base — so when the landlord raises rent, they pay. The Marketplace isn't a side hustle. It is the clearest evidence of what Atlassian actually is: a platform that profits from the gravity it has accumulated, not from the apps themselves.

Why Atlassian is bricking its own software on purpose

The third layer is the most aggressive, and it explains the company's behavior better than any PLG slide. For years, Atlassian sold software you could run on your own servers — a perpetual-ish license, your hardware, your control. That was the escape valve: a customer who hated price hikes could simply stay put on self-managed software. So Atlassian is closing the valve. Server support ended entirely on February 15, 2024. Data Center, its remaining self-hosted option, is on a countdown: new licenses for new customers stop in March 2026, existing customers can no longer expand from March 2028, and on March 28, 2029, the products go read-only.¹⁰ There is no version four. There is only the door marked Cloud.

This is the why behind the why. Every self-managed seat is a seat Atlassian doesn't fully control — it can't raise it on a subscription cadence, can't bundle it, can't compound it. The deployment table shows the prize: Cloud was $2.699B in FY2024, Data Center $1.208B, and dying Server just $177.6M.¹ By engineering a hard end-of-life, Atlassian converts every one of those self-hosted seats into higher-value, recurring Cloud ARR on a fixed schedule, whether the customer wants the move or not. It is the rare company with the leverage to issue its installed base a migration deadline and make it stick. Same customers. Higher math. No choice.

But didn't it really sell without a sales team?

The fair objection is that the PLG story is true and I'm being cynical: Atlassian genuinely grew on the strength of products people sought out, and that's a real, rare achievement. Granted. But 'no sales team' was always a careful half-truth. Atlassian leaned on a network of 400-plus channel partners to do the actual enterprise selling; as former president Jay Simons described it, the company didn't have its own deal-hunting reps — partners handled those engagements instead.⁹ The selling happened — it just happened off the balance sheet's most obvious line.

And the present is even less ambiguous. By FY2024 Atlassian was spending $877M — roughly a fifth of revenue — on marketing and sales, with internal teams and 'enterprise advocates' courting its largest accounts.¹ The myth describes a 2015 company; the 2024 company is something else. There is also an honest accounting caveat worth naming: the business press calls Atlassian 'profitable' on the strength of $1.014B in non-GAAP operating income, but on a GAAP basis it posted a $117.1M operating loss, the gap largely explained by paying its people in stock.¹ None of which weakens the thesis — it sharpens it. A company that compounds expansion revenue, taxes its ecosystem, and times its own product obsolescence doesn't need a heroic sales force. The structure does the selling.

Atlassian makes its money the way a landlord does once the tenants have moved in, hung the pictures, and rerouted their whole lives around the address. The empty sales floor everyone admired was never the achievement; it was the thing you noticed instead of the architecture. Land a team for almost nothing, let dependence compound, tax everyone who builds on top, and then — when the lease is up — quietly inform the tenants there is only one building left to move into. The story was 'a company that doesn't have to sell.' The truth is a company that arranged things so it never has to.