In 2001, Mike Cannon-Brookes sent an email to his classmates at the University of New South Wales asking if any of them wanted to start a tech company after graduation. One person replied. Scott Farquhar wrote back.
That was the entire founding story, more or less. Two students at a Sydney university who had separately decided they did not want corporate jobs, responding to each other’s interest in building something instead. Their stated ambition at the time was not to build a billion-dollar company or change the enterprise software market. It was to match the starting salary of a corporate graduate job, which in Australia in 2002 was around $48,500, without having to work for someone else.
What they ended up building over the next two decades was one of the largest enterprise software companies in the world. Atlassian hit $4.1 billion in revenue in fiscal year 2024, serves over 300,000 customers including most of the Fortune 500, and listed on the Nasdaq at a $4.37 billion market cap in 2015, making Cannon-Brookes and Farquhar Australia’s first technology startup billionaires. The company did all of this, for the first several years, without a traditional sales team. They didn’t have a pitch deck. They barely had a credit card.
Actually, the credit card was the only funding they had. $10,000 charged to a personal card. That was the capital Atlassian started with.
Supporting Other Companies and Hating the Bug Tracking Software
The first version of Atlassian was not Jira and it was not Confluence. It was a services business: Cannon-Brookes and Farquhar were providing customer support to other companies. They were taking calls at all hours, tracking issues and requests for clients who needed help managing their customer service operations.
The software they used to track their own work was Bugzilla, an open source bug tracking tool developed by Mozilla. Bugzilla was functional but not easy to use, and for two people running a small operation and trying to keep up with client issues across multiple accounts, it was not meeting their needs. They kept bumping into its limitations.
So they built their own.
The tool they built was named Jira, a reference to Gojira, the Japanese name for the 1954 Godzilla film. The naming had no particular meaning; it was a play on Gojira because the founders thought it sounded memorable. Jira 1.0 launched in 2002 as an issue and project tracking tool built primarily for software teams. It was faster than Bugzilla, more configurable, and had a cleaner interface.
After building it, they made a decision that reoriented the entire company: they realized they were better off selling Jira than providing customer support services. The product was more scalable than a services business, the margins were better, and it solved a problem that was genuinely widespread among software teams. They stopped answering other companies’ support calls and started selling bug tracking software.
The pricing decision mattered as much as the product decision. Atlassian set Jira’s price at $800 per license for small teams. At a time when enterprise software routinely sold for tens of thousands of dollars per seat through lengthy sales processes, $800 was a price that any software team could justify buying without a formal procurement process. You could put it on a company credit card. You didn’t need to get the VP of Engineering to sign off. You didn’t need to sit through a demo from a salesperson. You just went to the website and bought it.
Their goal in the early days was to sell one license per week. If they could do that, they could cover their costs and grow.
They did it without a single salesperson. Every customer found Atlassian through the website, evaluated the product themselves, and paid online. This was the model that Farquhar would later describe, in his farewell letter when stepping down as co-CEO in 2024, as the start of what is now called product-led growth. They didn’t have a name for it in 2002. They just didn’t have money for salespeople and built a product good enough that people would pay for it without being sold to.
The American Airlines Fax
The story that best illustrates what was working became part of Atlassian’s internal mythology.
One morning in 2003, Cannon-Brookes and Farquhar came into their small Sydney office and found an order from American Airlines on the fax machine. A large US airline had seen Jira on their website and bought a license without speaking to anyone at the company. Nobody at Atlassian had called American Airlines. Nobody had sent them a proposal. The sale was entirely inbound.
They looked at each other and said, roughly: did you talk to them? No, did you? No. So American Airlines just found us on the internet and bought software from us?
It confirmed something important. The target market for Jira was overwhelmingly in the United States, which was a problem and an opportunity simultaneously. They were a two-person operation in Sydney trying to sell to software teams in San Francisco and New York. Flying there to demo software was not viable. Building a US sales team was not affordable. But if American Airlines could find them through a search engine and buy without any human contact, so could other companies.
They invested heavily in the self-service purchase experience instead of salespeople. Clear product documentation. An easy trial process. Transparent pricing on the website with no “contact us for pricing” obfuscation. A support system that let customers get help without picking up the phone. The approach was unusual enough that it attracted comment. Most B2B software companies of the era assumed you needed a sales force to close deals above a certain size. Atlassian was proving otherwise, one fax at a time.
Confluence, the New York Office That Didn’t Last, and ShipIt
In 2004, Atlassian launched Confluence, a wiki-style team collaboration platform. Where Jira was for tracking work, issues, and bugs, Confluence was for documentation, knowledge sharing, and team communication in written form. A software team might use Jira to track what they were building and Confluence to document how it worked. The two products were complementary in a way that made each more valuable when used together.
Confluence followed the same pricing model as Jira: low cost, available on the website, no sales team required. The customer base grew across both products as word spread through developer communities and technology teams.
With most customers in the US and virtually none in Australia, the founders decided they needed a physical American presence. They opened a small office in New York in 2005. Within a year they had realized it was a mistake. The office was too expensive for what it was producing, and the team was stretched thin maintaining operations across two locations separated by fourteen time zones. They closed the New York office and acknowledged it had been an overextension.
They’ve told this story publicly over the years as an example of making a mistake, recognizing it quickly, and not letting sunk cost push them to compound it. The lesson they took was that making a hard call sooner was better than delaying it. They eventually did establish a US presence, this time in San Francisco, which worked better.
In 2005, Atlassian started ShipIt, a quarterly hackathon where employees spent 24 hours building whatever they wanted with whoever they wanted, with no constraints on the direction. The results were presented at the end. The format was designed to create dedicated space for experimentation outside the normal product roadmap.
Some of Atlassian’s most significant products came out of ShipIt sessions. Jira Service Management, which became Atlassian’s fastest-growing product for several years running, started as a ShipIt project. The hackathon format became one of the more copied elements of Atlassian’s culture as the company became better known.
Eight Years Without Outside Money
Atlassian ran for eight years without taking money from outside investors.
The $10,000 credit card debt from 2002 was gone within the first year of selling Jira. After that, the company was funded by its own revenue. They hired slowly. They grew by adding customers who found them online, upgrading to bigger licenses, and spreading Jira through their own organizations. A development team at one company would adopt Jira, that company would grow, the Jira deployment would expand, and Atlassian’s revenue from that account would grow with it.
By 2010, the company had passed $100 million in cumulative revenue. The revenue model was generating enough cash to fund the business without outside capital, but the founders decided it was time to bring in a major investor, not because they needed the money but because they wanted access to advice and networks for the next phase of scaling.
In July 2010, Accel Partners invested $60 million in Atlassian. Notably, most of that investment was a secondary transaction: Accel bought shares from the founders rather than putting fresh capital into the company. The founders took some money off the table after eight years of building. The investment valued Atlassian at around $400 million.
By the time Atlassian filed for its IPO in 2015, annual revenue was approximately $320 million and growing. The company had reached $100 million in annual revenue by around 2012, growing from $255,000 in 2007. The trajectory was consistent and built almost entirely on the back of a product that sold itself.
The IPO and “Australia’s First Tech Startup Billionaires”
Atlassian listed on the Nasdaq on December 10, 2015, under the ticker TEAM. The market capitalization at IPO was $4.37 billion, making it one of the larger software IPOs in Australian history and one of the more notable of the year globally. The New York Times, in a piece that Atlassian employees remember with some amusement, called the company “very boring software” because of its focus on bug tracking and project management rather than consumer-facing applications.
The boring software had made Cannon-Brookes and Farquhar billionaires. They were, by wide acknowledgment, Australia’s first tech startup billionaires: people who had built a technology company from zero to billion-dollar valuation rather than inheriting wealth or making money through real estate or mining, which had historically been where Australian fortunes were made.
At IPO, the two founders owned approximately 30% each of the company they had started. The IPO had no VC-driven liquidity event structure forcing a quick sale or complex ownership cleanup. They had taken one outside round from Accel in 2010 and had kept their ownership largely intact through to a public listing that gave investors access to a company with thirteen years of operating history, consistent revenue growth, and a product that was genuinely embedded in the workflows of software teams around the world.
The IPO was the moment Atlassian got properly noticed outside of the developer and enterprise software communities it had served for over a decade. The story of two guys with a credit card and no sales team building a multi-billion dollar software company from Sydney attracted attention. The boring software was suddenly interesting.
Trello, HipChat, and the One Bet They Got Wrong
After the IPO, Atlassian had the currency to make acquisitions at scale. The first significant one was Trello, bought in January 2017 for $425 million.
Trello was a visual project management tool built around boards, lists, and cards. Where Jira was powerful and complex, Trello was simple and visual, aimed at teams doing general project tracking rather than software development. It had nearly 19 million registered users at acquisition and was growing fast. Atlassian’s logic was that Trello served a different user than Jira, primarily business teams in marketing, HR, and operations rather than software engineers, and that bringing Trello into the portfolio extended Atlassian’s reach beyond its technical core.
The acquisition worked. Trello retained its product identity within Atlassian, kept growing, and opened the platform to users who would never have been Jira customers.
The bet that didn’t work was the attempt to compete in enterprise chat.
Atlassian had acquired HipChat in 2012, a team messaging tool that was popular among developers. When Slack emerged as the dominant player in the category after 2013, Atlassian tried to compete with a rebuilt product called Stride. In 2018, after two years of trying, Atlassian admitted it wasn’t working. They sold the intellectual property for both HipChat and Stride to Slack in exchange for a small equity stake in Slack and got out of the chat business entirely. The decision was pragmatic and expensive in terms of the resources spent, but the alternative of continuing to invest in a losing position would have been more expensive. They took the Slack stake, shut down the products, and moved on.
The HipChat exit illustrated something about how Atlassian made strategic calls. They were willing to try things that didn’t work and willing to stop doing them quickly when the evidence was clear. The New York office. HipChat. Getting out when the data said to get out was a consistent behavior across both good markets and bad.
The No-Sales-Team Model and Who It Actually Worked For
Atlassian’s approach to selling software became more interesting after the IPO precisely because the company started to make more money and people studied how it worked.
The model was simple in concept. Build a product that solves a real problem. Price it accessibly. Make it easy to try and easy to buy online. Let the product spread within organizations as people who like it tell colleagues or switch jobs and bring it with them. Use a marketplace of third-party apps and integrations to extend the product’s functionality and lock customers in through integration depth.
The customers Atlassian described themselves as targeting were the “Fortune 500,000”: not just the largest companies in the world but any company that needed software for teams to track work. By keeping prices accessible and the buying process frictionless, they could land customers from a startup with five engineers to a multinational with 100,000 employees. The multinational might take years to become a large account, but they would often get there through organic expansion rather than an enterprise sales relationship.
By FY2017, Atlassian had over 89,000 customers. By FY2024, that number had grown to over 300,000. The revenue per customer had grown steadily as companies bought more products, added more seats, and moved to cloud subscriptions that included ongoing support and upgrades.
The Atlassian Marketplace, where third-party developers built and sold apps and integrations on top of Jira and Confluence, had passed $500 million in cumulative sales by 2021. More than 5,700 apps in the marketplace meant customers could customize and extend the platform without Atlassian having to build every feature themselves. Customers who had integrated their Jira instance with dozens of third-party tools were not going to switch to a competitor easily; the switching cost was the sum of every integration they would have to rebuild.
The Cloud Transition and Loom
The most significant operational challenge of Atlassian’s post-IPO decade was a deliberate and difficult transition: moving the business from selling perpetual licenses for server-installed software to selling cloud subscriptions.
For most of its history, a company buying Jira would download it, install it on their own servers, and pay a one-time license fee plus annual maintenance. Atlassian called this the Server product. In 2020, the company announced it would stop selling new Server licenses in February 2021 and end support for existing Server installations in February 2024. Every customer needed to move to either Cloud, the hosted SaaS version, or Data Center, a self-managed product for organizations that needed to run their own infrastructure.
Forcing customers to migrate is operationally painful. Some organizations had run Jira Server for ten years, had deeply customized installations, and resisted moving. Atlassian built migration tools, offered transition pricing, and extended timelines where necessary, but the shift involved years of friction with a significant portion of the customer base.
The long-term logic was sound. Cloud subscriptions generate higher margins and more predictable revenue than perpetual licenses. A customer on a cloud subscription is more likely to expand their usage and buy additional products than a customer on a server installation who only interacts with Atlassian at renewal time. By FY2024, cloud revenue represented approximately 70% of Atlassian’s total revenue and was growing at around 30% year over year.
In October 2023, Atlassian acquired Loom for $975 million. Loom was a video messaging tool that let people record short videos of their screen and face instead of sending a long email or scheduling a meeting. For remote and distributed teams, it was a way to communicate complex information asynchronously. Atlassian integrated Loom with Jira so that video recordings could be attached to issues, and built Loom AI features that could turn a video transcript into a Jira task automatically.
The Loom acquisition was the company’s largest to date and reflected a thesis about remote work: that asynchronous communication tools would become as important to distributed teams as project management and documentation tools. Atlassian had been operating as a fully distributed company itself since announcing its “Team Anywhere” policy in 2020.
Farquhar Steps Down and What $4 Billion Looks Like
In April 2024, Scott Farquhar announced he would step down as co-CEO at the end of August, leaving Mike Cannon-Brookes as the sole CEO. Farquhar had been co-CEO alongside Cannon-Brookes for twenty-two years, since the day they founded the company. The co-CEO structure was unusual in the technology industry and even more unusual for lasting as long as it did.
Farquhar’s farewell letter, filed publicly as an SEC exhibit, described the company’s history with candor about what they had gotten right and what had been hard. He noted that Atlassian had “started what is now known as product-led growth by selling business software online with no salespeople” before the term existed, and that twenty-three years later the company was operating as one of the largest remote-first enterprises in the world.
He also wrote something that captured the absurdity of what the company had become: rockets don’t launch into orbit without Atlassian’s software. NASA’s Jet Propulsion Laboratory uses Jira. The fax from American Airlines in 2003 had scaled, somewhat improbably, to teams managing space missions.
In FY2024, Atlassian reported approximately $4.1 billion in revenue, serving over 300,000 customers across more than 190 countries. The market cap, which had peaked above $100 billion during the 2021 software boom and fallen with the broader market, was in the range of $40-50 billion in 2025. The founders, who each owned roughly 20-30% of the company they had started with a credit card, remained among the wealthiest people in Australia.
In March 2026, Atlassian announced it would lay off around 10% of its workforce as part of a pivot toward AI and enterprise sales, a signal that even the company that had built its identity on not having a traditional sales force was beginning to reconsider what the next phase of growth required.
The product-led model that Cannon-Brookes and Farquhar invented by necessity in 2002, because they couldn’t afford salespeople and Australian venture capital didn’t really exist yet, had become a widely imitated approach to building software companies. The name had been given to it retrospectively; the practice had been invented in a small Sydney office by two people trying to match a graduate starting salary without working for anyone else.
Leave a Reply