How Atlassian built a $45 billion software company far from Silicon Valley’s ‘froth and bubbles’

The combination of an Australian headquarters, a Nasdaq listing, and customers around the world means that Mike Cannon-Brookes usually has his first coffee of the day before he opens his laptop at 5 am. But Atlassian’s co-founder and CEO sees his location as one of the reasons why his enterprise software company has grown to a $48 billion valuation, making him one of Australia’s wealthiest entrepreneurs.

Straddling the Pacific Ocean can be both a source of frustration and differentiation, Cannon-Brookes says. Sydney’s diversified economy makes it “a great microcosm,” where Atlassian can test whether ideas hatched in California’s Bay Area will work in less tech-forward cities around the world.

Even far from the “froth and bubbles” of Silicon Valley, Atlassian has been caught up in US investors’ mood swings over whether AI will hurt or help software companies like his. Atlassian’s stock has leapt 230% from an April low, helped by stronger-than-expected earnings, but is still 60% below its late-2021 peak.

Cannon-Brookes says the evidence he’s seen so far suggests that customers will spend more with Atlassian as they use AI more heavily. “The most AI-centric, the most AI-native, those are actually the customers that are expanding the fastest for us.”

Here’s how he explains what it’s offering them, and why he’s advising other CEOs to make big bets on AI, even if they won’t all pay out.

This interview has been edited for clarity and brevity.

What’s your routine in Sydney?

I start most days at 5 am. I work 5 to 7, then kids 7 to 9, then work 9 to 3. I’m not one of these early-risers that gets up, does some journaling, drinks some water, goes to the gym, that sort of thing. It’s more like [I wake up at] 4:45 am, make an espresso, and get to the computer, which is less healthy than all the journaling and meditation kind of stuff.

I’m interested in the challenge of building a company this big from Sydney. You’ve straddled the US and Australia for a long time, but it sounds like you think the time zone is not a disadvantage.

I travel an awful lot. When you’ve got a global customer base, at some point it doesn’t matter where you are [based]. But I think one of Atlassian’s biggest hidden advantages is having one foot in Silicon Valley and one foot squarely outside of Silicon Valley, in a large traditional corporate market. Sydney is a big city, with banks, insurance companies, manufacturing companies, and grocery companies. We’re fast adopters of technology, but we’re not super technology-driven.

I always say you get the froth and bubbles of Silicon Valley. That is amazing and naive at the same time. This amazing “I’m going to change the world” attitude has produced some of the most incredible businesses ever. And at the same time, sometimes there’s a lot of technology that’s applicable between San Francisco and San Jose, and nowhere else in the world. We take the ideas from Silicon Valley, and we try to make sure those ideas apply in Sydney, which is a proxy for every other big city in the world.

Atlassian is going to be 25 years old next year. How do you describe the company’s place in the market now?

We help companies run business processes, and collaborate around those processes in a variety of different areas that are critical glue to how a company actually runs on a daily basis. We don’t handle their accounting system, their email, or their calendar. We handle all of the glue in between the systems for their business teams, their technology teams, their service teams, and their leadership teams, and try to bring that together into repeatable business processes. Which sounds naff, but a great company is just actually a collection of processes that [also allow people to] innovate and change and collaborate. People aren’t robots. The intelligence and creativity of the people is what makes it a beautiful mess.

You’ve been talking about ‘context’ as the great AI bottleneck. What exactly is the context problem that your customers are dealing with?

AI models are intelligence, and you can rent intelligence from lots of different model providers. It’s a fantastic battle [between them] that we’re all benefiting from because the intelligence power that we’re being given is going up continually. But models lack wisdom and knowledge. They have a lot of generalized knowledge. Ask any model in the world where Einstein was born, and they’ll give you the answer. If you ask, ‘When is this project due?’ they have no idea because it’s not in the model; it’s in your company. So you have to bring that knowledge from the company with the intelligence of the model. Knowledge is not exactly context, because context requires you to have that wisdom of your company. What we have, for many customers, is 20 years’ worth of business processes, 20 years’ worth of workflows, and 20 years’ worth of decisions they made, and the outcome of those decisions.

Have you gone through the cycle many other companies have gone through of encouraging your software developers to go wild on the latest AI models and then realizing it’s costing you a fortune and reining them back in?

I’m a bit more sanguine on this. I think there’s a logical learning path that the world is on. If every staff member is an AI novice, and you want them to become AI-native, how do you do that? You have to encourage them to use all these tools. You have to encourage them to play, to spend tokens, and to learn. So, I think it’s natural that companies are going to say, “Use all these tools.” Then, as people start using them, they’re suddenly going to be like, “Wait, they’re all really expensive.” I think that’s not a bad thing. You’ve learned how to [use them]. You’ve convinced people that they’re valuable. You’ve changed their mindset. Now you [just] have a problem of budget management. What you need to understand is how your token use relates to performance and [returns on investment].

You advise companies not to outpace their customers. At the same time, you’re in the business of innovating, and trying to get customers to upgrade. How do you strike the right balance?

Your understanding of the technology should be ahead of your customers’ at all times. The challenge we have at the moment is we can also ship technology far faster than our customers can consume it. I’ve had a number of customers say, “You are building things too quickly for me to absorb and change my organization to use those tools.” So you also need to be increasing the customer comprehension of what [those tools are] for. If we’re building things that they’re not ready or able to use, we need to spend time increasing their comprehension and their ability to use what we’re making. Why is this going to change their world?

You say CEOs should make big bets on AI rather than pivoting with each model release. Why?

You need to work out what bets your company is placing. What are the things that aren’t going to change? You’re not going to get it 100% correct. But you need to make some multi-year bets that are a durable strategic pillar your company can build on top of, that is not going to be quickly worn away by AI’s pace or competitors’ pace in your industry. Those are going to be different for every business, [depending on their] core competitive advantages. AI makes you rethink your business from first principles, which many businesses haven’t had to do for a long time. And first-principle thinking is quite hard for people to do. It’s more luck than skill if you don’t deeply understand the technology.