Good to go here. We're very excited here. First and foremost, I'm Thomas Blakey, infrastructure software, and this is Andres Sheppard-Slinger. We're excited to have the Head of Investor Relations, Martin Lam, here with Atlassian, which we think is an exciting buying opportunity. As a segue there, I'd ask Martin to maybe share with the uninitiated, what is it that, Atlassian is trying to solve for customers.
Yeah. Hey, Martin Lam. Thanks for having me. Y ou think about Atlassian, we're a collaboration company. Our ticker symbol is obviously Team. If you look at our products, it's really to plan and manage work. It's to help share knowledge at scale. Especially when you think about that future in an AI world, none of that really changes, right? You still have to plan, manage, and track your work, whether it's humans or agents. You still have to share knowledge across the organization, whether with AI or with humans. All that continues to be very relevant. That's, I think, why we're really excited about our position in an AI era.
One of the genesis of Atlassian was with developers.
Yeah.
As IT in general took over organizations, you were there in terms of managing those teams. Why don't you just maybe address the boogeyman in the room in terms of AI, will we need less developers?
Sure.
How this is going to affect the Atlassian model maybe?
Yeah. I think it's interesting, right? We have a strong heritage with developers. That's where we got our start initially. That represents a small part of our business today. If you actually look at even Jira's core user base, 50% of their users are technical. That includes basically developers, engineers, and IT. The other 50% are non-technical or general business users. We already have a very diversified kind of user base. Everyone's focused right now on the development and how that continues to evolve. We think of this world where humans and agents are gonna work side by side. Humans are always gonna have to be in the loop. When you think about humans are always gonna have to have agency of directing what do we want these agents to do? What do we want them to build?
What do we want them to accomplish? The humans will have to have checkpoints of, did that agent actually accomplish what we wanted it to do? There continues to be this role of humans and agents in the loop, especially even in software development. Again, tracking back to the point I made earlier, the role that Jira plays, I don't think that changes anything. That level of criticality continues to rise because you need to be able to say, "Okay, did that agent actually build what we wanted it to? How do I track and manage all these different agents running across my business? Are they doing what we wanted them to do?" Jira has always been this orchestration layer of tracking work. Who's working on what? What is the output? Does that actually ladder up to our company strategy?
In a world where there are agents in the mix, are those agents doing that? How do I track this increased complexity across my organization? Again, I think that's a really exciting opportunity for us.
I think, along those lines, people are worried about seat growth. You commented on your last quarter, it continues to expand. Yo u even invoked, I guess, the AI 50. Many of them are using your product and customers are expanding their seats even faster in that vertical. Just maybe talk to us about, you said you're expanding outside of developer. Y ou're staying outside of the developer community. Talk to us about what's driving the seat growth.
Yeah. Look, I think we continue to grow with those technical users and non-technical users, so across both vectors, which is great to see. I think two things I would first share is that, one, we looked at those customers that are using these code gen tools. Again, try to address some of the fears out there. When we look at those customers that use a Cursor, that use a Claude Code, VM, Copilot, across thousands of customers, so not an insignificant number, those customers are driving 5% more tasks through Jira. It's leading to actually 5% greater monthly active usage because to your point, more people have to collaborate to manage this increased complexity or throughput of work. Ultimately, it's driving those customers to expand their seats 5% faster than those customers that don't use these code gen tools.
It's really interesting, and again, kind of backs up what we've always believed in, but we thought it was important to share with market. Then number two, I think we get questions about, okay, what about these hot new AI startups? We took the non-subjective list, the Forbes AI 50, didn't want to cherry-pick some stats. I just threw that over to my data science team and said, "How many of the hottest startup, AI startups in the world are, our customers today?" 60% are already our customers today, and they're all growing their seat counts much faster than the overall corporate rate, which is probably not too surprising.
We feel good about our kind of position in an AI world, especially even with these hottest companies that investors may have questions about, "Hey, do those companies work in a different way?" It seems like we're well-positioned even with these companies.
Going back, shifting back to the core, the non-AI 50, you guys have been making a concerted effort for a long time now to move upmarket with the enterprise. We saw RPO jump over 40% in the last quarter. Can you just maybe give us an update in terms of the, I guess, the positive report card?
Yeah.
In terms of those moves with the 40% growth?
Well, in this moment in time, like I think some people would say, okay, are enterprises maybe taking a beat to evaluate their investment options or evaluate kind of the vendors? If anything, we're seeing a lot of these enterprises vote with their wallets in this moment in time. We saw enterprises grow their seat counts. They sign for larger dollar amounts, and they're signing for longer contract duration with showing the commitment to the Atlassian platform, showing that they want a trusted vendor to run their agentic workflows on, and a reliable platform that has context behind the type of work that they're trying to do. The enterprise efforts are going great. That joint report card that you referenced is the result of three things, I think. Talked in more recent times about three company priorities.
It's serving enterprises, delivering AI through what we call the Atlassian System of Work, which is moving more towards the this collection strategy or holistic solutions as opposed to individual point products. Those three strategic initiatives are coming together, and that's kind of the combination culminating in the report card that you just referenced, and it's driving great results right now. We wanna continue to push ahead on those three strategic priorities, and I think those are gonna be important ones as we drive more durable growth into the future.
It's good to see the buy-in via that RPO statistic. Sorry to keep addressing the gloomy. I think it's important for RPO.
Yeah, of course.
We've seen the pressure on software in general as well as Atlassian stock. That's why this vector is a little bit more leaning this way. From an AI orchestration and that orchestration layer that you talked about with regard to the System of Work and even the horizontal movements that you've had across organizations and successfully in recent years, you've heard some providers of maybe some of the apps you connect to talk about the orchestration layers that, and it's not just Atlassian, it could be ServiceNow, it could be Microsoft, becoming parasites, if you will, in terms of trying to connect to our data. Could you maybe not to be provocative, but maybe address how does Atlassian get some of its data?
How would you view like Atlassian in that type of market where maybe there's an instance where one of the applications shuts off the connection to some of its data?
Yeah. I think a couple things to start on for those less familiar. Atlassian has 350,000 customers. We have tens of millions of users, and we power hundreds of millions of discrete business workflows throughout the platform. That platform investment that we've made over the past several years is really important. It provides what we call the Teamwork Graph or a knowledge graph of understanding how relationships and kind of the context behind those relationships, what third-party tools your individual team is utilizing. That's actually how we're able to surface, I think, very intriguing or compelling actions or use cases with our AI. We've always had an open ecosystem philosophy. That's no different in an AI era.
Like we've always tried to be, this coordination layer or a single pane of glass to coordinate all your different workflows. Even when you use third-party tools, 'cause we realize that you're not always gonna be sitting in our platform. We've intentionally tried to stay away from getting too far into specific dev tooling or specific technologies because what's lasting is kind of the human problems of like, okay, human problems are gonna be the more lasting problems as opposed to technologies which come and go. We've always believed in integrating with best-of-breed third-party tools. Now, those come in and out of flavor depending on which kind of, segment that you're looking at. But it's very important to pull, tie into these third-party tools.
Whether you're in Microsoft Teams or Slack or Salesforce products, we want to be able to understand how those teams are working 'cause that's the basis of the Teamwork Graph. In this new era, I think where you're poking at. We still continue.
I'm always poking in.
We continue to believe that the open ecosystem approach is an important one. If customers want to tap into Atlassian products, like that's gonna be an important one because they should have that freedom to do so. We want to encourage that. What's important when we look at the MCP data, customers are writing and editing within our products. That's an important aspect rather than just purely reading, because that means that we're driving more and more workflows through the platform, and that begets more and more users to collaborate on those workflows. That also drives upsell opportunity or upgrade potential. We encourage that type of access into our platform. It is an area that we continue to monitor as we look at how the industry continues to evolve about how to maybe monetize kind of access to your platform.
Again, that open philosophy is a really important one for us, and philosophically, that's something that we believe in.
Certainly doesn't sound like you're seeing any issues in the market. Thank you for addressing that. Atlassian has a number of drivers to growth. You talked about seat growth. You talked about even sub-segment growth, whether it be the Forbes AI 50. Another growth driver you have is moving folks from your Data Center segment, which is on-prem to the cloud. Maybe give us an update there in terms of, what we should expect in terms of tailwinds and we just had some tailwinds recently and maybe some future headwinds as we kind of move through this, moving folks from Data Center to the cloud.
Yeah. Today we have two deployment models. We have our SaaS offering or cloud offering, and then we have what we call the Data Center offering, which is an enterprise-grade on-prem deployment model. We announced the end of life of that Data Center deployment, and so those customers will have to move off of that by March 2029, so a little over three years from now. During that time, we'll continue to migrate those Data Center customers to the cloud. That's an important aspect for us because cloud is where all the innovation. As earlier I talked about the platform, that's where the lion's share of the investment has been. That's where you have AI capabilities. That's where you have analytics. That's where you have automation capabilities. None of that's available on Data Center.
It's really important that we migrate those customers to give them the best experience that they can get, which is gonna be on the cloud platform. These are our largest customers, the Data Center cohort, right? Like these are the biggest of big enterprises, many of them who have been Atlassian customers for over a decade. Many of them have 50,000 users or more. It requires a little more handholding to bring these Data Center customers over. They have customized kind of footprints on their Data Center instance. Part of the challenge and opportunity, of course, is how do we get them to rebuild some of that customization onto the cloud platform?
That's why we know it's gonna be a multi-year journey for some of these customers because they have to rebuild some of the customization on the cloud instance. How do we hold their hand through it with more kind of white glove service? Like these are all the different avenues that we've looked at. We have a program now called Fast Track, which is giving them a dedicated R&D resource to help them roadmap out, the path to get to the cloud. That has accelerated customers' migration plans quite considerably. That's something that we're rolling out with larger scale today.
Let me ask maybe a two-point question, though. What does that calculus look like? That business is large, just under $2 billion. If we fast-forward, say five years from now, what does that business look like when it's on the cloud? Is there any type of calculus or 2:1, 1.5:1 in terms of the dollars?
In terms of the dollars?
Yeah, the dollars.
I think a couple of things to keep in mind is, again, the Data Center cohort is a much smaller pool of customers, but with much larger seat counts behind them. That's the first thing that I would point out about it being a multi-year journey ahead to move these customers over to the cloud. You get varying levels of price uplift that you're pointing at depending on how many users the customer has. On the low end of town, it could just be a cost-neutral or 5% uplift. On the high end of town, it could be upwards of 2x the price. It really depends on kind of the number of users that a customer has. We've been hesitant to give kind of a direct ratio to apply because of that varying level of complexity.
I think that's something we're thinking more about how do we help investors understand the level of uplift. I think the commonality between all these, though, is that because these are the largest customers, we do have a more traditional enterprise negotiation process with them. We want to ease them over to the cloud. In those instances where they have more significant uplift, we wanna smooth that path out for them. In year one, it may be less prohibitive and therefore make it more cost neutral for them. More of a one-to-one kinda ratio in year one. Then years two and three, you start to peel back that level of discounting. You slowly ramp them to that full list price over time. You kinda actualize that price benefit within years two and three.
Very helpful. Talking about some of the benefits of moving folks to the cloud, and you get a lot more technology. You've rolled out just nine months ago the Teamwork Collection. Maybe a lot of AI capabilities in there as well. This is a segue for the audience to just that you've moved to the cloud, you can buy things like Teamwork Collection. Maybe we have a brief definition of that. What are you kinda seeing there in terms of drivers for this? I think you reached over 1 million seats in less than nine months.
Yeah.
There's demand there. What are folks doing on that front?
Yeah. The Teamwork Collection, as you talked about, it's only been in the market for about three quarters now. As we rolled out Rovo, our AI capabilities across our platform, you get a certain number of AI credits with your standard subscription. With Teamwork Collection, you get 10x the amount of AI credits relative to your standalone Jira or standalone Confluence subscription. It's very appealing for customers as they think about their kind of AI journey, because when you talk to our customers, they don't want consumption-based pricing, at least today for our products. I think it makes sense for certain other models, but for our products, they don't really wanna pay on a consumption basis.
Today, if you use AI credits above and beyond your allotment on your standard Jira subscription or Confluence subscription, then you start to pay us on a consumption basis. But again, if you upgrade to Teamwork Collection, you get 10x the amount of AI credits. It's a way to kind of forego that worry or monitoring of AI credits today. That's been the number one driver of why customers want to upgrade or choose to upgrade to Teamwork Collection. A lot of success early on. It's exceeded our expectations. 1,000 customers have upgraded to Teamwork Collection. It equates to 1 million seats sold. Those customers that have upgraded Teamwork Collection have expanded their seats versus their standalone footprint on Jira by double digits or by more than 10 percentage points of seats expansion, which is great to see.
It drives additional kind of attach of Confluence and Loom, and then ultimately the 10x amount of AI credits. There's a lot of compelling kind of package within Teamwork Collection.
Super exciting. Did you wanna maybe share with us what you've seen some of your customers using with regard to AI? You talked about the product side, and that's Rovo. Just, cause I think a question that investors have are, what are folks from a stickiness perspective, from a growth perspective, what are folks doing on the AI on the Atlassian platform with AI?
Yeah. We have over 5 million monthly active users of Rovo today, which is great to see. That has grown considerably over the past year. When we look at kind of the agentic invocations, as we call them, or kind of business process automations, like that's been really encouraging because it's basically speeding up people's daily business processes. We talked about there with tens of millions of business processes flowing through the platform, and to the extent we can speed those up, that's great to see. I think a really interesting one is of those agent invocations, 40% are within Jira Service Management. There's a clear strength within Jira Service Management of those service management Rovo automations.
Okay. Any questions in the audience by chance? Martin? Last two. Sir.
Can you tell me, maybe I have one more question. I'm reaching that year where you start talking about stock-based compensation and wondering, are you still taking? Do you have some indications you can provide on that?
Yeah. That's a great and very relevant question. T he question was in regards to stock-based compensation, how do we think about that? Yes, I think we've been thinking about this very seriously for over six months now, where there's a lot of energy around this. We take it very seriously. I think if we look at our stock-based compensation as a percentage of revenue, we are on the high side. I think we recognize that. We need to bring that down over time. We're looking at different ways that we can perhaps modify our stock comp program. How do we distribute stock comp? 'Cause it is an important part of compensation, but in perhaps a more disciplined way. Do we look at different levels?
Like these are things that we, I think, have to think about as we move forward. Yes, like it's a priority for us to kind of accelerate that profitability. We hear you and we think about this quite a bit. The question was like, how do we think about timing on this? I won't provide a specific timeline. I would say there's different actions that we look at and think how do we think about. Again, part of the balance is of course retaining and hiring key talent. That's the balance of this all that we have to contemplate. I think two things that I would, I guess, point out.
Number one is part of what's driven that is the significant hiring that we've done over the past several years, both in terms of investing in AI as well as building out our platforms. That's been central to our three strategic priorities. I think we've cracked the back of a lot of those platform investments, and so you're actually seeing us moderate the pace of headcount growth, and you're actually starting to see that play out. We're able to reallocate some of that R&D talent that's been exclusively focused on platform initiatives into other product specific areas. That rate of hiring, as that starts to moderate and slow, that should help in some regards. The other aspect as we move to the enterprise that Tom poked at earlier is we've talked about R&D as a percentage of revenue coming down.
Meanwhile, sales and marketing will continue to tick up slightly. Structurally, that should help. We'll always be R&D and product led first, so those will never cross. If they come down and come closer together, that should structurally help from the stock-based compensation perspective, since R&D tends to be where that's concentrated. On sales and marketing, you tend to not have to give as much.
That's helpful. It maybe jogs the question about what do they say? Drinking your own champagne or eating your own dog food. Well given you have these great AI tools and from a workflow management perspective and the new coding tools coming out, talk to us about where you are in the journey of benefiting from AI efficiencies internally.
Yeah, it's a top priority of ours, both across the company. For R&D, I think it is where a lot of the focus is. Even within areas like HR, marketing, even within finance, we're looking at ways of how do we continue to drive, productivity with AI, the adoption of AI tools, our own products, as well as third-party products. Actually really pleased with the adoption across the company, but you always can do more. Some of it is just change management, right? Like there's a lot of exciting capabilities available. I have a lot of agents today at my disposal that I built, but there's always opportunity for more.
It's taking the time and thinking kind of creatively about where I can apply agents into my daily work life and kind of speeding up processes which would have taken me longer from a manual perspective.
Interesting. You have your own agents. I like that. I mean from a journey perspective, are we early in this?
I think it's still quite early.
I guess everybody's early.
Yeah, I think again, a lot of it is just thinking about how do you reinvent certain workflows to utilize or leverage AI. I think like one of the interesting ones is we get asked a lot about our R&D team, right? Are you using some of these code gen tools? I think what people miss is coding is only about 30%. We do surveys, you can look at third-party surveys. Coding is only 30% of a developer's time today, hands-on keyboard development time, coding time. W here we try to apply our AI capabilities is on that other 70% as well. Like, how can you speed up some of the other 70% of what they do in collaborating? How do you figure out what we're trying to build?
On the whole, we use Cursor, we use Claude, we use a lot of tools, Rovo Dev in the mix. Kind of the net net of that is our developers are about 25% more efficient, which I think shows that there's a lot of opportunity for efficiency there.
You addressed the 70% too, so that can be your secret sauce. You mentioned like shifting some dollars from R&D to sales and marketing. That's part and parcel to the enterprise initiatives that you've started and are being successful early on with. Talk about some of channel endeavors too. I think 'cause Atlassian's a multifaceted growth story to me. There's the Data Center to cloud, there's the new channel initiatives, there's the enterprise movements, there's the seat growth even still that nobody wants to talk about. This would be the fourth or fifth, I don't know what number we're on, the fourth rung of growth opportunities for Atlassian.
Maybe again, share some history with the team and the audience about how dependent you used to be on the channel and where your kind of initiatives are.
Yeah. The channel plays an important role for Atlassian's go-to-market efforts. Today the channel touches about 50% of our revenues. Many of them have grown up alongside us. However, many of them have also driven a lot of their business purely by just reselling Atlassian products. Part of the evolution that Tom was poking at is we're trying to evolve the channel from just purely reselling Jira seats, adding Jira licenses into the mix, to more value-added services. We've been trying to paint the opportunity for our partners of if there's actually a significant dollar opportunity, size just as much dollar opportunity and services.
Some of it is actually pretty compelling from a change management perspective, AI adoption perspective that we just talked about. How can they help our customers be more successful, especially as we push more into the enterprise. Some of that relationship may be more direct with us as we grow our quota carrying through the sales force.
I think I've asked you this before, Martin, so I apologize. Has there been like any like, size of or appropriate. Can you put a number on what you think that the benefit to like the longer term sustainable growth of this company could be from those channel initiatives? Because it is a large piece of your business, almost 50%, the channel space very highly viewed. Is there a calculable number that you think that can be added to the sustainable growth of Atlassian.
How our business, yeah.
From moving some of these from account ownerships in-house and managing the reselling of these cloud services.
This evolution with the channel is a multi-year journey as well, right? Some of these channel partners are a little more resistant to the change. Some partners are much more on the forefront, and they've been adapting actually even ahead of us announcing some of these channel initiative changes. This will be a multi-year journey with them, again, as they more move into value-added services. I think this is this benefit that you're pointing at, as kind of a multi-year process as well.
Just a multi-year process.
Yeah.
Any more questions from the audience?
Just sort of, do you, I guess the need for different types of applications or anything like that in the market?
The question was, do we see the need for different type or new products into the mix to evolve into this agentic world? I think we'll look hard at how our customers are using products. I think like actually one exciting development with Jira that we just announced, I think two weeks ago now, is the ability to assign work to agents directly in Jira. That includes not only our first party Rovo agents, but third party agents as well. For example, if we're working on something and I say, "Let's spec out and draw up the product specs," you can assign that to a Figma agent today. You can assign that to a Rovo agent today.
That I think is an exciting step back to that point that we talked about earlier of you still need to track and manage where all this work is happening. Even if third-party agents are doing some of that work, you still need a place to track all that complexity, especially as this continues to grow. I think that's an exciting aspect of Jira's place as kind of this pane of glass of tracking where all those agents are doing and what are they actually working on, and are they actually delivering the results that we wanted? Yeah.
I know I only have 15 seconds left, but I'd like to maybe ask about the M&A strategy. We get a lot of questions from investors relatively acquisitive at the end of last year. Some relatively cool companies like The Browser Company. Maybe just kind of update us on the M&A strategy and where there maybe are some more holes in the portfolio.
I don't know if I can do that in 15 seconds.
I know.
We did two sizable acquisitions last year for us, DX and The Browser Company of New York. DX is a product that is kind of a natural product to cross-sell into our software team or Software Collection as we call it, where you can measure the ROI of different AI tools that your teams are adopting. You can see where kind of the bottlenecks start to appear within the development process. It's a very valuable tool, especially in this moment in time. Even beyond this, like understanding, especially as teams get more efficient at perhaps the coding aspect, where is the next bottleneck in their development process? The Browser Company is a very different type of product and very different strategic approach.
I think most of you probably do most of your work in your browser today. We can envision a world where you can offer a lot of compelling AI added to features into the mix. Especially when you think about all the different SaaS applications that you have in the tabs across your browser. How can you string together different contexts across that? How can you provide compelling AI capabilities by pulling in data from those various tabs in the browser that most of us sit in all day? That's typically the kind of first port of entry into a lot of our daily work. I think that's actually pretty compelling possibilities that we can give to our customers.
Yeah. You did that well in a short period of time. Thank you so much, Martin. Thank you for coming.
All right. Thanks everybody for making it today to this panel and to the Cantor conference. We really appreciate it. Quantum panel number one. We have another one tomorrow. We're lucky to have a variety of panelists here. We got Subodh from Rigetti, Yuping from Quantum Computing, we got Joe from Horizon, and we've got Christian from Xanadu. I want to quickly start with this audience interaction question here. Who in the audience thinks quantum's gonna get commercialized in the next three years? How about five years? Who thinks it never gets commercialized? All right, no skeptics. It's crazy, all the executives are in the three-year camp and the investors are in the five-year camp. Thanks, guys. How about let's just start with a quick introduction.
Give me like a 1-2-minute story about your company and then explain what Rigetti does and where this kind of goes.
Thanks, Roy, for having us. I'm Subodh Kulkarni, the President and CEO for Rigetti Computing. Rigetti is based in Berkeley, California. We belong to the camp that is called superconducting gate-based quantum computing modality. We build superconducting circuits using silicon technology, so essentially like semiconductor-type technology. We create quantum states, and we build quantum computers. We are a full-stack company, so we do everything from chip design to chip fabrication. We own our own fab all the way through the various hardware layers or software layers. Increasingly, we are one of the teams. Along with us, there are bigger companies like IBM, Google, the government of China, and many other companies doing superconducting gate-based quantum computing. We differentiate ourselves with our open modular architecture where we can incorporate partnership solutions.
Specifically, we are partnered with companies like AWS for cloud, NVIDIA for distribution layer software, Quanta Computer for control systems. Our model is more partnership model. We focus on chip design fabrication, and within chip design fabrication, Rigetti has been around for a while. We have almost 300 patents. There are various areas, but primarily the main differentiator comes to our belief in chiplet technology. We are focused on using chiplet technology to scale up the number of qubits while we continue to improve fidelity. That's a quick summary of Rigetti.
Yeah. Perfect. Yuping?
Everybody, Yuping Huang from Quantum Computing Inc. Quantum Computing Inc. is a quantum optics and nanophotonics company. Our headquarters is just across the Hudson River in Hoboken, New Jersey, and we have one facility. Great. Can you guys hear me better now?
Much better.
Okay, thank you. We now have two facilities in Boston. Another facility in New Jersey. We have a facility in Santa Barbara, and then we just acquired a quantum communications company in Chicago. We have a small scale quantum nanophotonics lab in Tempe, Arizona. At our core is our nonlinear integrated nanophotonics based on what we have started over the past I would say 15, 16 years.
Based on this performance technology platform, we are actually commercializing not only quantum computing, but quantum sensing, quantum communications, and photonic AI with the hope that we can significantly reduce the energy consumption of neural networks. Because we really don't believe that building nuclear power plant next to a data center is a sustainable solution.
Thank you. I'm Joe Fitzsimons. I'm the founder and CEO of Horizon Quantum Computing. We're based in Singapore. That's where our headquarters are. We've operations in Ireland as well, in Dublin. What we're focused on is basically building the software infrastructure to enable broad quantum advantage. By that, what I mean is we're trying to build the software development tools to allow the domain experts in all of the fields in which quantum computing stands to have a real impact, to be able to harness computers that actually solve real world problems. I would like to tell you that that's where the field is today, that we're all out there solving all of these hard problems for industry, but we're not. Q uantum computing is still relatively early. We saw the hype up early about commercial applications.
I would say that probably aligns with the general feeling of the field. What we're doing though is we're trying to make it more like programming conventional computers to program quantum computers. There's a real challenge there because ultimately, if you want to take advantage of a quantum computer, you need an algorithm that's specially designed to take advantage of a quantum effect called interference. If you're not doing that, there's no advantage. What we're focused on doing is trying to get a path from the point of code written for a conventional computer to automated acceleration on a quantum computer. Now, that's a pretty challenging path. We've a long way to go on that.
What we have had to do on this path is build up what I view as some of the most capable programming tools in the space, in that we have built up to a very high level of abstraction, so we can build programming languages that allow the user to start to define objects that are not necessarily that are to represent different kinds of information, matrices and vectors and so on. So that as they start to program, it no longer requires them to have deep knowledge of quantum computing.
We can compile this all the way down into a portable form that we can execute on a really wide range of hardware systems, be they superconducting systems, be they trapped-ion systems, be they neutral atom systems or hopefully in the near future, photonic systems, because they each have their own interesting properties. We want to make sure that we are able to harness as wide a variety of these systems as possible, get to a real advantage for the developers so that their software is creating real value for their end users, and do it in a way where the developer does not need to make a bet on which hardware they're working on. That's us. Pass to Christian.
Christian?
Hi everyone, my name is Christian Weedbrook, Founder and CEO of Xanadu. We're based in Toronto, and our mission is to build quantum computers and make them useful and available to people everywhere. I guess we're characterized by like two aspects. We use light or photons to build our quantum computers and also our software stack, which is called PennyLane, one of the most widely used software stacks in the industry is dynamic. It runs on all the major hardware providers out there.
One of the big achievements from a photonics perspective was a major publication last year that showed how to network together quantum computers with our latest quantum computer called Aurora, which showed how to network four server racks together and could scale up to an indefinite number of quantum computers or quantum server racks, provided we get the noise under control.
Perfect. All right. Thanks, guys. W e got hardware and software here. I'm gonna start with hardware first. There is about five. I like to say there's five different hardware methodologies for quantum. We've got two of them represented on the panel here, superconducting and photonics. Can you guys just talk about maybe the benefits and the disadvantages? We'll start with photonics, both of you. We'll start with Christian, and then we'll do superconducting second.
Maybe I can start off with the challenges. The challenge for any quantum computer is how to deal with errors. This noise causes errors and depending on the type of platform, it'll manifest in a certain way. For photonics, it's loss. Whenever you shine light through a medium, whatever it may be, free space or even on integrated chips, not all the light makes it from the start to finish. There's loss, lots of photons. To correct that, you use error correction, and also you make the chips better, as well. That's our biggest challenge for photonic systems. The benefits are really pretty awesome. We can work with large-scale foundries. In fact, we already work with some of the largest in the world, UMC, Tower, GlobalFoundries and others.
It's definitely CMOS compatible, as they phrase it. Most quantum computing architectures will have to network to scale up, and we've solved the networking challenge. Also, we didn't have to invent the laser or fiber optics. We could stand on the shoulders of the telecommunication industry, which helps lower costs and helps us move faster as well. The other big one is, apart from a small part of turning on the computer or initializing it, the rest of our computation is at room temperature. Literally room temperature means no cryogenics and no laser cooling at all. They're the kind of pros and cons of our approach.
Yuping anything to add ?
Yeah. I wanted to add that, I guess a disadvantage using the photonics approach is that the storage of quantum states is harder than or the memory time is shorter. To counter that, a significant advantage is that, so because the computing is really performed at the speed of light, so you can repeat it many, many times and with a very low cost. That actually gives rise to opportunities that we can jump outside the conventional wisdom of having to get very, very high gate fidelity, having to do a very good job in the error correction. Because there are opportunities that you can construct a optical loop so that you can repeat or you can evolve certain operations over a very short amount of time.
Another thing is that, just like Christian said, so we can leverage the manufacturing on the CMOS side and really look at making the nanophotonic chips to host the quantum functions that can work at room temperature. That with the possibility of co-designing and co-hosting photonic-based QPU next to CPU, GPU, so that so we can really have very powerful systems that with different part to handle quantum, with different part to handle classical for the real problem solving. This is where I'll stop.
All right. Subodh?
Sure. As you mentioned, there are two modalities here. Superconducting, which is superconducting gate-based modality, which is what we are dedicated to along with companies like IBM, Google, the government of China, to name a few. There's photonics, there's trapped ion, neutral atoms, spin and other things. There's pros and cons of different technologies, which is why they coexist. The big advantages of superconducting, which is why many of us are investing in it, is scalability and speed. We are fundamentally using semiconductor technology. We know that once we get the chip defined properly, we can scale it up using five decades of semiconductor industry experience. We feel very good about that part of how do you go from hundreds of qubits to thousands and tens of thousands, leveraging all the five decades of semiconductor experience.
Speed, we are dealing with electrons, so we can move commensurate with CPU, GPU speed. When you look at numbers from us or IBM or Google, we're talking about tens of nanoseconds gate speed or clock speed. The downside for us is because we are dealing with man-made chips, we deal with errors and fidelity issues. We intrinsically get fidelity challenges no different than classical semiconductor, but we are not quite good enough to invoke error correction and bring it all under control. Our main disadvantage is fidelity, but the strengths are speed and scalability. When you compare that to some other modalities, particularly not photonics as much as trapped-ion or neutral atom, they're physically moving ions and atoms, which are much, much bulkier than electrons.
You'll typically see trapped-ion and neutral-atom speeds tens of thousands of times slower than where we are. When we report tens of nanoseconds gate speed, you'll see trapped-ion or neutral-atom companies report hundreds of microsecond gate speed. We have this huge advantage in speed and scalability, but what we trade off is nature makes perfect atoms and ions, and so they're typically, their fidelity is better to begin with. Those are the pros and cons we deal with.
Yeah. Very perfect. All right. I'm gonna ask Joe this question because I think you'll be unbiased, but, are we headed towards one dominant quantum architecture, or is there gonna be a heterogeneous ecosystem where we have hybrid platforms?
I mean, I think this is really interesting. Actually hybrid platforms get you some way towards overcoming some of the issues that you've mentioned about storage times, because you can go to vapor cell memory and things like this. Going to hybrid systems can certainly help. The reality is there are, at least in my mind, at least four leading approaches to quantum computing at the moment. You have superconducting systems, you have trapped-ion systems, but photonic systems have been really making a lot of progress, as have neutral-atom systems. You have four distinct paths that are all pushing towards getting to scalable quantum computing.
At least for me, well, for our business, that's certainly beneficial because ultimately the way we are approaching things, when you program through our tools, you're not having to pick which hardware platform you're targeting because we can compile and produce really efficient code on a very wide range of hardware platforms. H aving the opportunities to act as that hedge for the developer against shifts in technology is certainly useful. At least for me, I think it's really hard to predict which technology is going to get to fault-tolerant quantum computing first because we're seeing so much progress. I think photons are absolutely perfect qubits. Everything about them is perfect except for the fact that you can't build a Lightsaber. If you shine two photons at