Good afternoon, and thank you for joining Atlassian's earnings conference call for the first quarter of fiscal year 2027. As a reminder, this conference call is being recorded and will be available for replay on the investor relations section of Atlassian's website following this call. I will now hand the call over to Martin Lam, Atlassian's Head of Investor Relations.
Welcome to Atlassian's first quarter fiscal year 2027 earnings call. Thank you for joining us today. On the call with me today, we have Atlassian CEO and Co-Founder, Mike Cannon-Brookes, and James Chuong. Earlier today, we published the shareholder letter and press release with our financial results and commentary for the first quarter of fiscal year 2027. The shareholder letter is available on the investor relations section of our website, where you will also find our other earnings-related materials, including the earnings press release and supplemental investor data sheet. As always, our shareholder letter contains management's insight and commentary for the quarter. During the call today, we will have brief opening remarks and then focus our time on Q&A. This call will include forward-looking statements. Forward-looking statements involve known and unknown risks, uncertainties, and assumptions.
If any such risks or uncertainties materialize or if any of the assumptions prove incorrect, our results could differ materially from the results expressed or implied by the forward-looking statements we make. You should not rely upon forward-looking statements as predictions of future events. Forward-looking statements represent our management's beliefs and assumptions only as of the date such statements are made, and we undertake no obligation to update or revise such statements should they change or cease to be current. Further information on these and other factors that could affect our business performance and financial results is included in filings we make with the Securities and Exchange Commission from time to time, including the section titled "Risk Factors" in our most recently filed annual and quarterly reports. During today's call, we will also discuss non-GAAP financial measures.
These non-GAAP financial measures are in addition to and are not a substitute for or superior to measures of financial performance prepared in accordance with GAAP. A reconciliation between GAAP and non-GAAP financial measures is available in our shareholder letter and earnings release on the investor relations section of our website. We would like to allow as many of you to participate in Q&A as possible, so out of respect for others on the call, we will take one question at a time. With that, I will turn the call over to Mike for opening remarks.
Thank you all for joining us today. As you've already read in our shareholder letter, we had a great start to fiscal 2027. Total revenue grew 24% to $1.8 billion. Cloud revenue grew 31% to $1.3 billion, and subscription ARR grew 23% to $6.8 billion. We also delivered GAAP operating margin of 11% and free cash flow of $463 million, an increase of 304% year-on-year, representing a free cash flow margin of 26%. These results reflect strong customer demand and execution across our big strategic priorities of enterprise, AI, and the system of work, our continued focus on driving durable, profitable growth. Customers are bringing more people, more agents, and more workflows onto the Atlassian platform, and they're making larger and longer-term commitments.
This quarter, Infosys, SAP, Workday, and Westpac are just a few examples of global enterprises that expanded greatly with us, standardizing how their teams collaborate, operate, and execute business processes on the Atlassian platform. You can see that in our RPO, which grew 44% year-over-year to $4.7 billion, and customers spending more than $1 million a year grew well north of 30% year-over-year. In the cloud, we saw continued strong seat expansion in Jira and Confluence, along with strong cross-sell into our collections. More than 3,000 customers have now upgraded to the Teamwork Collection for expanded Rovo credits and platform capabilities. That's a 7x increase year-over-year. The Service Collection has now surpassed 70,000 customers, with ARR continuing to grow over 30% as we continue to take market share. What makes our system of work valuable in the AI era is a unique combination of context, collaboration, and confidence.
With context, yes, models are getting smarter, but to be effective at work, they need to understand an organization's goals, knowledge, people, decisions, and history. The Teamwork Graph now connects more than 250 billion objects and relationships and extends across customers' third-party tools to more than 80 out-of-the-box connectors. That context translates into real value for our customers. Customers with a richer Teamwork Graph are twice as likely to become deep Rovo adopters, and deep adopters grow ARR more than twice as fast than as customers with light Rovo usage. Teamwork Collection customers are running twice as many agents while expanding their Jira and Confluence seats much faster than standalone customers. Next, to collaboration. At Team '26 Europe, we introduced our Agentic Multiplayer Protocol, or AMP, as a collaboration layer built across the Atlassian platform, not as a separate product that customers have to buy.
It helps people and agents work together with shared context, clear identity, attribution, and governed authority. More than 10 million human-agent collaborations now happen on our platform every month, double the level of six months ago. To learn more about AMP, I just shared a Loom on our investor relations website. Finally, to customer confidence. As agents take on more work, enterprises need visibility, security, and governance of where that work is happening. We continue to build those controls into our platform so customers can scale their AI responsibly. More than 87% of our enterprise cloud customers are now using Rovo, and our larger, longer commitments show how deeply customers are investing in this foundation. Overall, we're executing extremely strongly our strategy from a position of strength. We're seeing this play out again in our results. I'm super excited about the year ahead.
Connecting agents and humans on the Atlassian platform will continue to unlock huge value for our customers, and we are extremely proud to be building for that future. With that, I will pass the call to the operator for Q&A.
We will now begin the question- and- answer session. If you have a question, please press star followed by the one on your phone. If you would like to withdraw from the queue, please press star followed by the one again. Your first question comes from Ryan MacWilliams from Wells Fargo. Please go ahead.
Hey, thanks for taking the question. Really strong result on the Data Center side. Then the full year guide was pretty similar. We have talked to some of your customers, some of your largest customers, and it seems like they are prioritizing the migration to cloud in the coming years due to agentic coding and some of the new features that you are offering. So, love to hear just what drove the strong Data Center result in the quarter, and then maybe how your customers are thinking about migrations for the rest of this year and at this point. Thanks.
Yeah, thanks for the question, Ryan. Yeah, I think maybe just starting on the overall performance on the Data Center side, as we continue to see customers migrate over to the cloud right now, we are now seeing some of the largest, most complex customers. One of the things that we are seeing in the quarter right now is really strong retention. A big part of that is obviously the cloud roadmap that we have, but also the commitments that we are making, for example, in areas like the public sector opportunity. We talked about submitting for IL5 or Impact Level 5 authorization six months early, ahead of schedule for July 2028, FedRAMP High authorization submission in January 2029 to follow that, and then submitting for Rovo and AI on our Atlassian Government Cloud platform in June 2027.
I think what we are seeing is that, in that commitment that we are making to our customers here, we are really being rewarded by some of the high retention rates that we are seeing, and that is outpacing even our expectations here. Really pleased with what we are seeing on that side of it. I will also note that right now, Data Center migrations to cloud are contributing approximately mid to high single digits to cloud revenue growth. But in terms of one of the areas where we want to make sure we meet our customers where they are commercially is to make available for them hybrid deployments, right?
In particular, in Q2, there were a number of very large customers in the quarter who, as part of their migration journey, signed cloud migration commitments through these hybrid deployments that enabled them to migrate from Data Center to cloud over a multi-year period, right? This is a great sign. Customers are deepening their commitment to Atlassian. They see cloud as the final destination, and because we do not charge customers twice on these hybrid deals, hybrid deployments did split revenue between Data Center and cloud, all right? As a result, that meant slightly more revenue on Data Center and slightly less on cloud this quarter. But if you zoom out, the key takeaway here is that the largest, most complex customers are moving ahead with their multi-year journeys to cloud, where they can really unlock all the benefits of the Atlassian platform.
Yeah, Ryan, I guess I will just add on a little customer color from my point of view. We have spent nigh on a decade now managing these migrations between server and then Data Center to the cloud. Having had more than 100 customer conversations in the last quarter, well over 100, I would say, the customers are very excited about what we have built in the cloud, right? They continue to be so. They see the value of AI. It is a big catalyst beyond all the enterprise and compliance and infrastructural improvements that James talked about to driving that momentum forward. We landed some very big migrations during the quarter, and we continue to work on them for the quarters ahead.
I would say the customer feeling and drive towards the cloud remains incredibly strong as they see how powerful, what a huge upgrade the cloud platform is from where they are in Data Center.
Your next question comes from Fatima Boolani from Citi.
Good afternoon. Thank you for taking my question. Mike, this one's for you. It's a little bit of a bigger picture question. As organizations shift to more of an AI software development life cycle away from the predecessor agile movement, just based on agentic coding becoming the de facto standard on how software engineering is being done, can you talk about how Jira and Confluence benefit from what is essentially, I would argue, becoming a diminished coordination or collaboration tax between human engineers and more presumably seamless collaboration between agents? Then just some of the commercial and monetization trajectory implications for these core products where the functional reason for their existence is to drive that collaboration. So, would just love to hear your perspectives on that. Thank you.
Thanks, Fatima. Yeah, great question. Look, we've often agreed on things in the past. I'd have to say we disagree on this question personally. The value of Jira and Confluence for coordination in the AI era is even stronger than it has been in the past. I think this is often mistaken, right? In fact, we shared an example at Team '26 Europe where Morgan Stanley, somewhere on the call, announced they're moving their Jira and Confluence to the cloud at large scale and rolling out Rovo to thousands of users.
The main reason to turn even more so to the Atlassian platform than they have been before as a longtime customer, is for their software development and to get a better view of what's happening across their teams and to connect their technology teams with their business teams, which is almost exactly out of our system of work playbook, right? I think there are multiple factors happening simultaneously when you talk about this. Firstly, it's about bringing all of that agentic work into a collaboration platform. The human role is incredibly important in intent, in applying intuition and understanding, in guiding and handling exceptions, in initiating work. So, we shipped a whole series of features about agentic sessions coming into Jira at Team '26 Europe, and they were incredibly positively received. There's no doubt the amount of work happening is going to go up, precipitously so.
The coordination tax of that then becomes even more so. If you are trying to understand 10x as much work, you need to understand more coordination, more connections than you have been beforehand. The context that Jira and Confluence provides of your history and your future are two of the most essential ingredients. We can see that. We know we are in the top handful of MCP servers of all of the major foundation labs, not just for reading, but for writing as well, as organizations continue to try to work out how to manage that across their systems. Lastly, as I said, it is about connecting those technology teams to business teams. Increasingly, that code writing, which is incredibly exciting, is being done by people in finance, by employees in marketing and HR and other areas.
You have even more of a coordination challenge, a larger opportunity for us to connect technology and business teams than we have ever seen before. I believe you can see that showing up in our numbers today. Our customers would say that. You can see that in our seat expansion numbers being incredibly strong this quarter. Our RPO number, again, accelerating to 44.5% this quarter, reflecting just increasing longer and larger commitments from customers are going out multiple years as they see the value of the Atlassian platform and the context we have in the Teamwork Graph and that collaboration happening across multiple teams, be that in the Service Collection or be that in the Teamwork Collection or the Software Collection.
We are incredibly bullish on our role in the AI future, and that is before we even get to AMP, which we have talked about, and I recommend watching the Team '26 keynote or the Loom that I just shipped today.
Your next question comes from Matt Hedberg from RBC Capital Markets. Please go ahead.
Great. Hey, guys. Thanks for taking my question. One thing I noticed is you are expanding your FDE motion. It was interesting. ServiceNow noted something similar last week. I guess I am wondering, can you expand upon that motion? Is this indicative kind of where customers are at in their AI decision-making? Just kind of trying to plan things out here, just kind of given how quickly technology is moving. Just color on that would be helpful. Thanks, guys.
Thanks, Matt. Look, I can talk to our FDE motion. Firstly, I would encourage, as I encourage all customers and I encourage all investors, all FDEs are not built the same. Our FDEs are fully qualified AI and ML engineers who are writing full production customer code and product code at the customer site. So they are trying to help customers to get their AI motions moving, to get workflows moving onto the platform. That is different to some other places where they are rebadged customer success staff, et cetera. It is extremely hard to scale that program fast. As you can well appreciate, those are very rare individuals. Our FDE program, though, is going extremely well, couple of quarters in now. We are working with many, many large customers on helping them get more value from AI, right?
They bring processes and improvements that they want to see, and our FDEs help them connect the Teamwork Graph. They help them explain to engineers and staff at the customer site how they can continue to roll the Atlassian platform out across more and more of their world, how they can continue to build out their ontology into the Teamwork Graph, and how that yields better results with Rovo, with coding agents, with service agents, et cetera. So, still early in the program for us. The goal is delivering AI usage and leaving the customer with a trained group of people who can then continue that rollout by themselves, who have confidence in Forge as an extensibility platform, in the Teamwork Graph, in our connectors, in everything that exists in the Atlassian platform, and it has been going very, very well. Again, our goal is not building a service business.
These are not professional services folk. They are rarely ever charged for, so it is pretty different to what other people are doing in terms of professional services and customer success. Lastly, I will say that philosophically, we are seeing great speed-ups in our ability to build product with FDEs. So increasingly, we are doing more and more core software engineering at the customer site. That is very unique. It is an ability of our platform and our really world-class R&D organization to be doing this. It helps us speed up the customer request to delivery loop incredibly fast. It is something we are very bullish on. We will probably talk more about that in the coming quarters as we keep accelerating that program.
Your next question comes from Adam Wood from Morgan Stanley. Please go ahead.
Hi. Thanks for taking the question, and also congrats on a good start to the year, and thanks for the shout-outs. Maybe if I could ask about AMP. I think you're one of many companies that are claiming the right to execute against owning and managing agent identities, permissions, and audits. Could you just talk a little bit about why you think you have the right to win here? Then I think you said you're not monetizing this directly. Do you always see it being that way? If so, where do we see that drive the revenues for you coming into the business, please? Thank you.
Thanks, Adam. How would I answer that? Look, a couple things. I think the, quote, unquote, "the right" to sort of orchestrate, manage compliance agents, et cetera, that's a very broad term. The reason you're seeing a lot of companies kind of say things like this is it means a lot of different things to a lot of different customers. I can only talk about what it is that we are doing. If you talk about what, I don't know, an Okta or a Workday or someone else is doing, it's very, very different, even though the words may sound the same, and part of the problem is just everyone saying agents are lost. We're trying to show in two different places, I think, incredible differentiation and uniqueness. The first I would point to is AMP.
We are trying to bring agents into customers' workflows, structured or unstructured workflows, into their collaboration. That collaboration happens an awful lot on the Atlassian platform, inside Confluence or inside service management. It happens an awful lot off the Atlassian platform. That may be in a partner app like Figma. It may be in a command line tool like Codex. All of that is contributing to your collaboration and your workflows. What we're doing in AMP is bringing you a set of experiences, really highly designed experiences, which are truly unique in the world today, as well as bringing all the technological standards, so your MCPs, your A2As, all the technology standards that are underneath AMP, and bringing them that in a very human-centric way so that you can quickly roll out the agentic use cases in ways that are understandable.
And whether your staff are operating on or off the Atlassian platform, you have understanding about what is happening. So when we talk about traceability and accountability, it is saying that people can see where a human has done something, where an agent has done something on behalf of a human, and if so, who it was on behalf of, if it is just agents operating by themselves, which does happen quite a lot. And if you need to pull other SaaS applications or other agents in, no matter whether that is on or off the Atlassian platform, it is understandable. It was incredibly well received by customers at Team '26 Europe because it is trying to bring the humanity, the design and experience, and the power of the agentic world together into understandable experiences across Jira, Confluence, Loom, et cetera. We are super bullish about that.
It is very, very unique and very different to what is happening elsewhere. We do have a lot of agentic governance features focused more on agents in the Atlassian platform. So when you are building Rovo agents, customers need to be very clear on the controls, the cost measures, the commissioning, and the compliance of those agents. We continue to improve that. We ship non-human identities, for example, and where those agents can get to our world with MCP servers and the CLI. Again, there is a lot of controls, but that is not so much about control at the orchestration layer overall. You can see that is going really well, I would say. More than 10 million human agent interactions, MCP servers past 2 million now, 15 million tool calls.
All of the numbers are going in incredibly rapid pace in the right direction, but with our goal of helping you collaborate across those business processes everywhere that you speak to. And we really think it expands our opportunity as we continue to win more and more workflows and business processes across all the types of teams, from leadership teams to service teams and everyone in between. How that shows up in the experiences for users is going to be incredibly important, I believe, and letting us harness the best of all of those models. So, it is a pretty unique viewpoint, I think, but it is very customer-centric and customer-driven.
Your next question comes from Gregg Moskowitz from Mizuho. Please go ahead.
Great. Thank you very much. Two related questions, if I may, because I think this is really affecting the share price right now with the investor concern being one of potentially slowing growth outlook at Atlassian. First, the cloud revenue guidance for Q2 reflects about 4 points of deceleration. That is quite a bit more than the guidance from 90 days ago, and it would just be helpful to understand why, if we could. Second, your Q1 subscription ARR growth was really strong, but you are forecasting a meaningful decline over the next three quarters. I know the compares get a little tougher, certainly due to the Data Center pull-forward and to DX in the back half of the year. Beyond that, are there any other factors we should be aware of that could weigh on the growth rate? Thank you.
Yeah, thanks for the question, Gregg. Yeah, look, I think from the starting point of what we saw in Q1, we were really pleased with the performance that we saw across that cloud business. As I mentioned a little bit earlier, sustaining north of 30% year-over-year growth, and really strong NRR, once again here. On what we are seeing here and what we are anticipating is the continued momentum on cross-sell. We saw that Service Collection in Q1 against cross 70,000 customers at $1 billion of ARR, growing north of 30%. In terms of seat growth from competitive displacements, that was up over 40% year-over-year. So, I think we feel really good about what we are seeing there.
As it relates to the Teamwork Collection, over 3,000 customers, 7x year-over-year now, and again, still remains the best way for us to help customers adopt and unleash AI across the entire Atlassian platform. When it comes to the guide here, I think we are still early in the year, and we are going to continue to take a thoughtful and prudent approach here, balancing the momentum that we are seeing in the business with RPO at 44%, CRPO at 28%, and balance that with the uncertainty around the evolving dynamics in the macro environment, fiscal policy, geopolitical landscape. So again, going to continue to take a prudent approach here. In Q2, we will begin to lap the DX acquisition, and in the second half, we will begin to face some tougher comps given the acceleration that we saw in the second half of fiscal 2026.
Of course, we are still scaling our enterprise go-to-market sales motion to unlock the massive opportunity that we have here. We are in 85% of the Fortune 500, and that represents really just about 10% of revenue. So, a ton of opportunity still ahead of us, and for us to continue to invest and scale the capacity on the enterprise sales side. If you look at our million-dollar customer plus cohorts, it has continued to grow north of 30% year-over-year. At the same time, as the business continues to mature, especially on the enterprise side, I do expect to see a little bit more predictability in the business and the forecast, and that is just a natural evolution as we mix shift towards more annual and multi-year deals.
But all up, feel really good about the state of the business, coming off a quarter where we posted really strong RPO subscription ARR growth of 22.5%, and cloud revenue growth still north of 30%.
Your next question comes from Allan Verkhovski from BTIG. Please go ahead.
Hey there. Thanks for taking the question. Mike, you said you don't want Atlassian to be a centralized control tower, but rather a highly connected subway station. So, as enterprises work through their broader AI strategy, what feedback are you getting on Atlassian's approach? What is resonating most with customers? What parts of that vision are maybe taking longer to land? And how is all of that affecting sales cycles and deployment decisions? And then James, just a quick follow-up for you, because I think it's important. Can you quantify the downside impact to cloud revenue in the quarter from the hybrid deployments you mentioned relative to expectations?
Thanks, Allan. Great question. There are a lot of things I would say here. Let me try to put it into a concise answer for you. Overall, the feedback we're getting from customers ranges from excitement to queries about how we are doing so well, which is a great question to get from customers. Rovo continues to massively outperform other AI systems or services that they are using, and we hear that consistently from customers. To credit to our engineering team, all of the layers from the platform through the Teamwork Graph, through the search, and Rovo and chat capabilities, and the integration into our products, our agentic capabilities, so when you're building agents with Studio, these are truly world-class and fantastic and deliver amazing results. And we hear that over and over again from customers.
When I say skepticism, the question is more like, "How are you doing this when X can't?" Or, "I tried Y and it doesn't work as well as Rovo. Why is that?" I love those sort of questions. The answer is we have a kickass engineering team. We've invested very heavily in the fundamentals and the architecture. We have a great ability to use models and harness them, and we have amazing organizational context. We have the best enterprise context graph in the world, and it continues to grow and get stronger. As you saw, we just added assets and agentic sessions and more and more inference that we run on top of that graph to fundamentally give better answers to agents. I think we're seeing that in the pull-through from customers. Our AI numbers are fantastic.
Agentic automation is up 9x year-on-year, now more than 10 million a month. 15 million tool calls from MCP a day. We just shipped only a few weeks ago a V2 of our MCP server, which greatly reduces token usage, has hundreds more tools and availability. The ability to create work, to manage work on the Atlassian platform, to handle business processes and workflows from incident management and service to HRSM, all the way through software teams and strategy, people managing their goals and processes. Agents are able to use the Atlassian platform, and humans are able to understand the agents that are using the Atlassian platform and interact back and forth with them, and that's what we want to be. I totally think the subway station analogy or being part of the subway network, that's where Atlassian has philosophically done well in the past.
It is where we build deep and meaningful customer relationships. Our goal is to help you get things done as a customer. It is not to be in control of your world. It's your world. I think that's landing really well, and you can see that in overall customers using Atlassian's AI technologies. The deep Rovo users are growing twice as fast as those that aren't. Obviously, we're adding to the pile of customers that are using AI heavily on a continued basis. Part of the reason we feel incredibly bullish about the future. I will say, it's very easy to claim these things about AI. So I tell every single customer, "Put us to the test. Take our Teamwork Graph, take our context graph, assess us against other people. I'll send you FDEs, I'll send you people to help. We will win that head-to-head battle.
Test us." Secondly, don't believe everything you read in marketing and from any vendor at the moment. Test it. There's a lot of claims being made about agents and this and that, and we're asking customers to prove to themselves that we can do what we say we can do. We stand behind all the engineering and the deep architectural and R&D investments we've made. This stuff is really hard. This stuff is not simple to build, certainly not to scale to incredibly high levels of throughput when you add compliance and governance, hundreds of millions, if not billions of objects being processed simultaneously to try to give great answers. We have a truly unique AI platform, and it goes all the way up to the design and experience of how users use it.
I will say I feel incredibly bullish, and I would tell every customer, "Test us. We'll stand up against anybody.
Allan, just to answer your question there, we haven't broken out that color, but I would just characterize it as modestly less than what we had expected in terms of impact. I will just point you again to overall Data Center migrations to cloud contributed that mid to high single digits to cloud revenue growth this quarter.
Your next question comes from Keith Bachman from BMO Capital Markets. Please go ahead.
Yes, thank you very much. I want to return to Gregg's question, James, to you. You actually had a reasonable beat on the quarter, maybe a little less cloud and a little more Data Center, but you are raising the full year guide by less than the beat. That is unusual. The message is you are less confident about the next three quarters. Is there anything that you can respond to that? In particular, was there any pull forwards in the quarter, or is there something on mix that we are not appreciating? If you could just speak to why isn't the beat being passed into the year?
Yeah, thanks, Keith. If you look at the cloud revenue beat of $20 million, we are flowing through slightly more than that through the entire year there. The mix that you are seeing, marketplace is also flowing through as well. It is really the Data Center side. If you recall, the Data Center beat in Q1 is $33 million, but again, there is some timing effects related to the pull forwards of ASC 606 on the rev rec. That is why you are not seeing the full total revenue sort of flow through from the Q1 beat through the year. It is really sort of timing mechanics related to that.
Your next question comes from Koji Ikeda from Bank of America. Please go ahead.
Yeah. Hey, thanks, guys. Thanks for taking the question. With AMP, do you see the Teamwork Graph evolving from primarily a context layer into a broader coordination and interface layer for humans and agents, and how does that expand Atlassian's long-term opportunity? Thank you.
Hey, Koji. I am going to guess that I am taking that one. I see them as sort of two layers. It gets a little nuanced and detailed here. The Teamwork Graph is the fundamental context layer underneath AMP and the Atlassian platform, and increasingly a lot of customer usage off the Atlassian platform. That is taking all of your knowledge, your work. We are now semantically indexing the entire code base if you want us to give all of your agents fantastic answers and help them get the right information, getting high-quality answers, significant token savings. We are continuing the Teamwork Graph to do more and more inference on your knowledge to pull out objects that exist. So we can now infer things like projects and teams that are not actually specified as objects, but we can see people working together. We are looking increasingly at inferring other things.
For example, we showed demonstrations of taking any agentic session that is in the Teamwork Graph and pulling out learnings to contribute to the very next session. That is all about helping any agentic use case, any search use case, whether that is human search or agent search, if you want to think about it that way. The Teamwork Graph is a large-scale knowledge graph, an ontology of everything your organization brings, highly customizable through Forge, through Assets, through lots of other technologies to get that data layer in a very AI-specific way with an incredible search engine and an incredible graph query engine, which are a little different, but both give great answers to agents, and agents are able to use them. AMP, think about it as more at the interface level for humans. It is more about experiences and designs.
How do we bring all the agents that you are using on and off the Atlassian platform into workflows in ways that humans can cooperate with? Underneath AMP are a lot of technologies. The Teamwork Graph, yes, MCP, A2A, so agent protocols to other tools. We are trying to bring that all together into a package. Again, maybe an interesting analogy, when you used the first iPhone, you did not care about GPRS and GPS and 5G, third-generation Wi-Fi, 2.4 GHz. All the technology standards were hidden inside the phone and just make a simple experience for you to use between Safari, between your phone, between mail, et cetera. AMP is about bringing all those agentic experiences into our products and off our products, and dealing with the modalities of how agents and humans are going to come together.
Think about it more as an experience layer that is across the Atlassian platform, but also helps inform third parties and how others will use our platform as well. It is built on top of the Teamwork Graph. Oh, long-term opportunity, you asked about. I am sorry about that. I think it continues to expand our long-term opportunity greatly. I think we are, again, durable, profitable growth is about continuing to deliver for customers in every single one of our teams, leadership teams, service teams, technology teams, business teams, product teams, showing how we can deliver incredible value across more workflows, more business processes, structured and unstructured across your organization, and between the Atlassian platform and all of the other technology tools that you use. We are seeing that in customers as they understand how to utilize agents.
Again, we have incredibly high AI usage on our platform because of things like AMP. Customers see that, and then they bring more workflows onto our platform. Many of the customers I spoke to this quarter are massively increasing the number of workflows they run through the Atlassian platform, right? In the Service Collection space, moving off legacy vendors onto the Service Collection because of our AI features, and not just sort of, "We will run your service tickets cheaper," but the way your agents are going to interact with that platform, how it all flows through, and how service is integrated into the broader Atlassian platform to connect to your technology, business teams, and others. I think it is obviously we are investing very heavily in it, both the Teamwork Graph and AMP, because I think it shows a large and multi-year expansion opportunity for us.
Your next question comes from Rob Owens from Piper Sandler. Please go ahead.
Great. Thank you for taking my question here. I wanted to ask on Rovo pricing and some of the thought process with the changes that are coming in December. If we look at current consumption rates, is there any way you can somehow quantify what could be incremental here? Thank you.
Hi, Rob. Let me talk to that. The headline I would say is, as we've said repeatedly, and I hope it's coming through, the Teamwork Collection, and slightly less, the Service Collection, are the primary ways that we think about monetizing AI today. You get a largely increased token usage or credit usage, so you get 10x as many credits in the Teamwork Collection. It's now over 3,000 customers, up 7x year-on-year. Those customers are running twice as many agents. They're growing their seats significantly faster than non-Teamwork Collection and non-AI deep usage customers. So, that is our major motion, we believe, for delivering customer value, and then getting that value back for Atlassian in terms of monetization. That is really, really important that is understood, that continues to be our way.
We are shipping a bunch of usage-based pricing, or UBP features, with enforcement beginning in early December. We were talking to customers about that for a long time. That's about aligning value delivered, especially for extreme customer usage. It is not a near-term revenue driver. We've been very clear about that. That's in our guidance. We do not expect that to be a large revenue generation activity. We are focusing our monetization efforts on continuing to grow to the previous question, to Koji's question, growing your long-term opportunity of more and more workflows running on the Atlassian platform through Teamwork Collection and Service Collection. UBP is about meeting customers where they're at.
We are getting increasing questions around pricing and packaging, around controls, around understanding what that usage is, how they understand their credit usage, who is using it, which agents, which staff members, how that is flowing through, and where they are getting value for that. We continue to see that in lots of different places, in automation, in Rovo credits, in index objects, in Teamwork Graph queries. There is a lot of different meters inside usage-based pricing. Again, we are not assuming any meaningful UBP revenue in FY 2027 at the moment. We are obviously continuing to work with customers, help them with their understanding. We have been very happy with the controls and offerings that we are showing today.
Again, our goal is to have very generous credit usage. Two other things I wanted to say on the back of that. I think I would give huge credit to our finance and engineering teams. Our gross margin is continuing to grow despite massively increased AI usage. That is no mean feat. That is an incredible achievement from those teams. Up to 85.2%, 3% year-on-year, with AI usage greatly expanding.
That is our platform investments paying off. That is very strong collaboration between our finance and engineering teams, our infrastructure teams, on how we are blending models, how we are helping improve the token efficiency of AI features that we are delivering to customers, improving quality and cost simultaneously. Our AI Gateway continues to be there. We continue to show incredibly high quality in Rovo Chat, despite cost per chat coming down on a continual and ongoing basis.
That is a huge credit to the team to do both of those things simultaneously, having the long-term opportunity to grow customers in their workflows while expanding gross margins and delivering a significant AI value to customers in the core packages of Teamwork Collection, Service Collection, et cetera. Those customers continue to grow really strongly, as I have said.
Rob, I would just maybe add to that. As Mike mentioned a little bit earlier, from an AI monetization perspective, what we hear from our customers is that they are looking for predictability, and they are looking for the ability to load balance those tokens. I think we sort of see this in the customer journey, right? As they upgrade from standard to premium to enterprise editions, they are receiving 2x-3x more Rovo credits at each turn. Ultimately, as they upgrade to Teamwork Collection, getting 10x more Rovo credits, right? So, we are seeing that in our ARPU uplift. We are seeing that in the seat expansions as well, as customers continue to consolidate on that platform. Again, important to meet customers where they are, but we really like the traction that we are seeing here.
Your next question comes from Arjun Bhatia from William Blair. Please go ahead.
Perfect. Thank you. Mike, I have a market structure question for you in terms of just how you think about growth going forward. Obviously, you've delivered a lot of innovation on the platform. I remember, I think in the past, in the enterprise, there used to be some maybe holdouts that wouldn't move to Jira that had custom-built systems. I'm curious how you're seeing that evolve with how quickly AI is moving and how that's changing sort of the DevOps landscape, and just how do you think about the net new customer opportunity from here versus expanding with the customers that you already have?
Thanks, Arjun. How would I think about that? Look, I would say at the headline, there's a lot of parts to that question. Let me talk to the headline. I think demand is extremely healthy. We're seeing strong demand at the SMB level and at the enterprise level. We continue to see strong seat expansion that's very broad-based, business teams, software teams, service teams, which is, I think, excellent and reflects, maybe not the thinking about growth as you asked, but where that demand and growth is coming from. Again, we're a very patient and durable company in terms of building out real deep customer value over time. You can see that, subscription ARR at 22.5%, cloud growing north of 30%. As James mentioned, net retention again staying extremely strong, well over the 120 mark, and obviously RPO at 44%.
Very strong future growth indicators, I would say, and demand continues to be healthy. I think we're seeing extremely strong. I'm not sure whether we talk about net new customers. Again, we have an awful lot of customers. Our 10K customer number and overall customer numbers continue to grow strongly. We're seeing very strong momentum, for example, in both collections. We talked about Teamwork Collections, over 3,000 customers, up 7x, and Service Collection now past 70,000 customers, with ARR growing north of 30% and strongly taking share. Not necessarily net new customer opportunities, but existing customers continuing to grow strongly. There are many, many, many, many workflows and processes within our existing customer base that are not currently on the Atlassian platform, and that's where those customers are expanding and bringing more workflows onto the platform.
That continues to be an extremely healthy demand flywheel that we continue to deliver on features to bring in there. Customers north of 3 million growing at over 50% year-on-year rate, and those at 5 million level growing at 70% year-on-year rate. Certainly, at the enterprise end, I think you can see very strong, healthy customer growth resulting in monetization as they bring more and more workflows on. We continue to improve, I will say, like how we do that. You are seeing all of that in lots of different places. I think Jira's role is incredibly important in that world going ahead. Structural workflows are very important.
I think you are seeing more and more customers adopting the Teamwork Collection to get Loom and Confluence, to get projects and goals and some of the platform applications as those workflows increasingly become unstructured as well, as they bring agents into those workflows. I am incredibly bullish on our opportunity for growth going forward. I hope you can feel that and hear that.
Your next question comes from Ittai Kidron from Oppenheimer. Please go ahead.
Thanks. Appreciate the opportunity here. Guys, I wanted to ask about the Data Center. James, you kind of listed before, earlier in your comments, a timeline for government certifications, Impact Level 5, FedRAMP, Rovo, and those kind of seem like carrying to late 2027, 2028, even touching on 2029. I guess it is obvious at this point that you are going to have a non-significant part of your Data Center customer base remain in Data Center beyond the March 2029 end-of-life notice. I guess my question is, one, is there a way for you to roughly quantify what percent of your Data Center business right here, right now, do you expect will remain, quote-unquote, "longer term" beyond the deadline of March 2029, number one. Number two, is that figure greater than what you have potentially expected a year or two ago?
I'm just trying to think if the conversion, the Data Center to Cloud conversion opportunity might be smaller, insofar what's remaining here, given that a not insignificant part of your Data Center business is just likely to stay on that platform for longer beyond March 29. Appreciate it.
Yeah. Thanks, Ittai. I would just say that and reiterate that our migrations continue to be on pace. Our customers are incredibly excited about the value they can unlock there, right? We talked about it earlier. Rovo threaded throughout Teamwork Graph, all of the Collections, and we see that in customers that move from Data Center to Cloud, right? Three years hence, they're increasing their spend 1.75x. We feel really good about the migration pace that we're seeing right now. But again, important to make sure that we're meeting our customers where they are. And I also reiterated earlier that for the quarter, mid to high single digits in terms of contribution to that Cloud revenue.
All up, I think the announcements that we're making around the public sector side is just a recommitment to our customers and making sure that they have the visibility that we're going to be able to unlock all of that value for them as they continue to migrate.
Ittai, if I might just add, and thank you to the amazing people at Atlassian who fed me the numbers. I've seen more than 100 customers in the last quarter in more than 11 different countries, in four different continents. I have a lot of customer detail and knowledge about these types of questions, right? I think one of the things that give me the most strength on Data Center, we talk a lot about customers, sort of talking about when and not if. I also think our history of delivery is incredibly impressive for those customers. Many customers who I saw a year ago or two years ago see the progress we've made in Isolated Cloud, in Government Cloud, in European data residency in various ways. I've been in Singapore, I've been in India, I've been in America, lots of countries in Europe.
Whether that is in financial organizations, IT, telecommunications, large-scale automotive, all of those customers see the progress we have made, our delivery on the platform. It is great that they see that. It is great that they are heading in the right direction. I think, importantly, it is their faith in the delivery going forward that is important and the realization of how difficult that is to do. Delivering large-scale AI solutions that are European data resident with amazing design and experience on top, connected to the best Teamwork Graph, and doing that all with bring your own key encryption, for example, and meeting various different standards, that is incredibly hard. I think that is a big moat for us. It is a big challenge for others to build at that scale and delivering the set of workflows that we are delivering.
That is why customers are adopting more and more of the platform as they move more and more workflows onto that. So, that is all coming through. Fundamentally, you can see it. Look, the 44% RPO number is incredibly strong, continued to grow from last quarter, and we have had a series of strong quarters there, and 28% in current RPO as well. So, that does not happen without recognizing that the value of that overall platform to continue to deliver for all of those customers around the globe. It is very exciting.
Thank you. That concludes our question- and- answer session. I will now turn the call over to Mike for closing remarks.
Thanks, everyone, for joining the call today. Thanks for those who came to Team '26 Europe a couple of weeks ago. I appreciate all the thoughtful questions and continued support. We will continue get cracking through fiscal 2027. So, thanks to all the Atlassians out there, and everyone else, have a kickass day and kickass weekend.