Autodesk, Inc. (ADSK) Earnings Call Transcript & Summary

Okay. Excellent. Well, hey, good morning, everyone. Welcome to day 1 of the Barclays Tech Conference. My name is Saket Kalia. I cover software here. Really honored to have the team with us here from Autodesk. We've got Mike Haley, SVP of Research; as well as Simon Mays-Smith, Head of Investor Relations. We've got about 30 minutes together. Let's take the first 20, 25 minutes to go through -- let's go through some fireside chat for the first 20 or 25 minutes, which I know is going to be fun and then we'd love to make it interactive. So any questions that you've got, just pop up your hand and we'll get a mic around. So with that, Mike and Simon, thank you so much for being with us here today. Simon, why don't you kick us off with the safe harbor?

We may make forward-looking statements during the course of this presentation. Please refer to our SEC filings for information on risks and other factors that may cause our actual results to differ materially from these statements.

Absolutely. Now we can talk. Well, listen, guys, thanks a ton for coming. A lot of great stuff to talk about here with Autodesk. But maybe, Simon, just to level set for us before we dive in with Mike, can you just recap a few of the most important points you wanted us to know coming out of the last call, just so that we're all on the same page.

Yes, and perfect segue. It's really sort of 3 things. First, the business is performing pretty well, both in absolute terms and compared to almost all of our competitors. Secondly, we gave some talking points really, but what things to think about for next year and those are basically divided into 3 buckets. One is that we will continue to think about macro and potential disruption from sales and marketing optimization when we set guidance in February. Secondly, that we will have headwinds from the new transaction model to margins, which is why we said the path to fiscal '29 goals would be nonlinear. And thirdly, we reminded people just on the math of the new transaction model -- sorry, the switch to annual billing transition has given us a tailwind to billings growth free cash flow growth in fiscal '25 and '26. And obviously, that process is pretty much done by the end of this year. So growth in billings and free cash flow will be normalized -- more normalized in '27 than it has been in '25, '26, all those things we said before, so that's the second thing. And then the third thing is just banging the drum that we are years ahead of our competitors in AI, thanks for the work that Mike and others in the company have been doing, and we're pretty excited about it.

Absolutely. That's a good summary and there's a lot to dig into there as well. But Mike, maybe to start with you. This is your first Barclays tech conference with. Hopefully, the first of many.

First of many.

Yes, absolutely. But maybe for those of us that are meeting for the first time, can you just give us a little bit of a brief background on yourself and your mandate in your current role?

Yes, sure. Sure. So in my role, I lead up Autodesk's research. So Autodesk research is around about a 300-person organization in the company. Who -- we are the only organization inside Autodesk fundamental mandate is to say 5 to 10 years ahead of the company, right. So our job is to derisk technologies in a safe, quick rapid environment and to find new investment opportunities. So we are doing everything you can imagine from AI to robotics, to new forms of user interfaces and across all of our industries, architecture, manufacturing and media, right? So we have -- that's why we have a significantly sized organization. Now coming to Simon's point, coming out of that work, we started doing AI work in my lab, in 2009, and we've been gradually expanding and growing that. In 2018 is when we created our AI Lab, which is kind of like our own deep mind kind of organization, right? Where we're doing very, very deep work, not on generic AI that everybody else is working on, but on AI that is specific to geometry, to shapes, to structures, to products, to buildings, in other words, the things our customers are building. So that's where we've been laser-focused now for 7 years. So what you're now beginning to see from us is the other part of my job, which is leading our AI strategy across the company. So I lead all of our generative AI work. We have a large cross company investment in this right now. We've been heavily focused on taking this technology that we developed in the lab and bringing it to market right now. We announced a bunch of that at our Autodesk University Conference this last year, which I'm happy to share later. But it's a really exciting moment for us. What I'll also just mention is that the cycle time these days between research and product is shrinking. I've been doing research for a long time. It could often take a decade for this new technology. It can be a matter of 3 to 6 months now. So the role of research in actually developing product and getting it out to market is becoming different, which is a pressure I'm feeling and -- but it's a good thing.

Two things to add, sorry, which are important. That time to market is a process and that muscle takes time to build. And actually just putting that together, building on that expertise is one of actually our barriers to entry and what it's going to take time for our competitors to replicate it because building that research breadth and depth and then making it work faster is hard to do. And second one is that centralization is critical. A lot of companies are letting go 1,000-flowers across the company, We're doing very efficient AI invest, not focused investment as well. So having a centralization of focusing the resources behind it is very important.

That is. That is, That cycle time, in particular, something that I hadn't thought about. Maybe just to build on that a little bit, though, Mike, and maybe to dig into some of the individual areas of Autodesk's business. I want to start with the architecture piece. Of course, anchored by your leading tool Revit, maybe the high level question I want to ask is, how is generative AI going to change the daily life for an architect in your view, if at all?

Yes. No. Look, I think it's going to be a radical change. Is it all going to arrive at one night? No, it's not. But I would break it into sort of 3 categories of change. So the first one is we've been pushing this idea of sort of outcome-based design in architecture for a while, which is really about thinking of what you want the ultimate product to be at the end of the day, I want it to be low cost, I want it to be low carbon. It has to exist within this existing city infrastructure. There's all of these broader mandates that are known at the beginning of a project, but it's hard to actually compute had to actually use those in any meaningful way at the beginning of a project. It's much easier to simulate and everything later on to make sure it complies to those and that's often too late. So AI is able to help people early in conceptual design, right, that the beginning of creating a new structural building or infrastructure to understand the impact and the goals that you want to go after. So that's the first thing. The second thing is the actual way you operate and interact with each other in the architecture space. This might be an architect working with a customer for example, who is trying to understand what a building might look like a new foyer in a hotel, what's this going to look like? Or it might be working with engineers and they're sharing information with each other. The whole architecture and building practice is renowned for it being difficult to communicate with people. It might be due to different representation of information or different skill levels of the people you're talking to. AI is beginning to eliminate that. It's beginning to be able to produce rendering instantly of what the architect is talking about or take the structural engineers information and automatically transform it into the architecture information. So now they too can talk to each other much like a language translator. That's the second one. The third one we see is really around what we call automation. So at least 70% to 80% of any design work. And this is not just true for architecture. This is true for manufacturing as well. It's not creative stuff. It's doing pretty much repetitive kind of tedious work, like, for example, producing plans, like 2D actual blueprints. You go to a construction site, there's still somebody walking around with size rolls of paper, right? Nevermind the iPads. So producing those is not a creative endeavor because the 3D model already exists, but it's tedious. It takes weeks to produce those. Let AI produce that for you. So there's a whole set of workflows that are going to -- just going to literally become automated overnight because they just don't -- they're not creative banks. The time wastes effectively. So just those 3 alone, you can imagine how those play out across architecture.

Yes. That's super interesting, Simon, maybe to follow up for you on that point. I mean in application software, we've all sort of debated about kind of seats and value, particularly this year. Maybe to Mike's point, as Autodesk sort of enhances that productivity of an engineer of an architect, how do the -- how do you think the economics here could change? I mean could there be fewer seats, but maybe more value that Autodesk earns? Curious just in general how you think about that?

Yes. So the way we're sort of thinking about it is on the sort of seat size, we think that the seat-based model will be around for a long time for the foreseeable future. So think about it as sort of access to the kingdom sort of thing, sort of base layer of functionality. And that will be driven by really 2 things. One is that there is more work to do than there are people available to do it in our industry. So if you go and speak to our local Department of Transportation, ask them how much work they have to do, how much money they have to do it, and you'll see there's a huge mismatch between those 2 things and then multiply that across the economy and across the globe. So there is much more work to do than there are people available to do it. So not -- there's pressure for more people, not less people that just isn't the resources for those people. And secondly, as we expand into adjacent verticals, our whole strategy is to connect workflows end-to-end in the cloud. So as we expand into construction, as we build our business and operations and maintenance and similar manufacturing, similar in media entertainment, we are adding new people and seats under our ecosystem. So our addressable ecosystem is actually expanding over time. But in addition to that, as we build our AI business and deliver more value to our customers, so as we talked about at Investor Day from task automation through workflow automation to system automation, the amount of value that we can deliver to our customers will significantly increase. And I'll talk a bit about that in a second and also the marginal cost will increase. And so there will be more consumption pricing just naturally because of that to capture that value and also to reflect the marginal cost from compute of doing that. The value that we deliver really has 2 elements to it. The first one is around the machines can do stuff. And I guess that's the other important point is as you move up that value chain, the lead consumption will be a machine rather than a human being because human beings just cannot handle the complexity of the models get bigger and bigger basically, where the machines can. So machines bring 2 new types of TAM to us. The first one is that they can do stuff that humans can't so they can. To Mike's point, they can see stuff in the conceptual design phase that 2 years after construction, a human being just would not be able to conceive of and to bring that value into the process in a way that humans cannot do. So that is new value we can deliver to the customer that we capture through consumption pricing. And the second way is that machines create new monetizable time because they work 24 as a day, 365 days a year. They don't sleep, they don't go on vacation. So it is essentially creating time that doesn't exist today. So new value, new times, new TAM to us, capture with consumption.

That's interesting.

And sorry, just another thing, which you didn't ask but is important is that the consumption will be some form of metering. To be honest, we don't precisely know what it is. It could be some form of token. It could be some form of value delivery. It could be some form of API monetization. We're not quite sure. What we do know is that Flex is basically built as a sort of universal front end. So we can plug in any sort of metering system on the front end of Flex and then have at the back end, essentially deliver a consumption experience through that. So a lot of the work we've been doing over the last few years has been getting us ready for this.

Yes, absolutely. That's really interesting in terms of how to monetize time and how that might become a bigger vector. Mike, maybe from just a product perspective in architecture, what can customers leverage now in terms of generative AI from Autodesk? And to Simon's point, just around sort of the different industry clouds, do these customers eventually need to move to Forma, right, in order to eventually kind of gain the those capabilities?

So it's actually surprisingly a little known fact. We actually built an early AI tool into some of our software more than 10 years ago. We have something called Construction IQ which sort of -- yes. So the reality is there's actually been a bunch of stuff for a while. Our Innovyze water products, have a bunch of AI in them for doing flood claim analysis and all of these kind of things. So this is all over the place within the products today. Now what we are doing, what we launched at Autodesk University this year is our new assistant, Autodesk Assistant, which is now available across all -- it's becoming available across all of our products across the industries, but it's specifically in things like Forma and Revit and these kind of products, right? That is enhancing your experience across a multitude. It's the agentic sort of environment, right. We're doing a bunch of stuff. Now that is also being enhanced by the work that our team has been doing that I mentioned in the beginning. So what we also announced at Autodesk University is our neural CAD for buildings. And that's what I was referring to earlier on in these large foundation models to be building that can actually think in terms of buildings. So they're not thinking in language, they're thinking in terms of building, right? So they can compute buildings, you can interact with them and they can literally create structures and things for you, right? Now what we did is we showed that available in Forma because that was very much part of the sort of conceptual design workflow that I was talking about before, but rapid exploration, that kind of thing. But you can expect to see these features -- many of these features also beginning to appear in products like Revit, as well because, depending on where they are in the life cycle, like I said, some of these are like automation features if they relate to documentation generation or deep design development, they're going to be inside Revit. If they relate to conceptual design, there will be inside something like Forma.

Yes, absolutely. I want to move to construction, right? So we talked about architecture, right, the next piece, of course, right? And Simon, maybe it's for you. I mean, picking up off of architecture and moving to Autodesk Construction Cloud or ACC. Can you just remind us what you've said on just the growth rates of that business? I feel like it's been very healthy and certainly reflecting market share gains. Maybe that related question then would be why do you feel like ACC is gaining market share.

Because they're awesome. The -- I mean, the short answer is it's the fluctuation of the strategy and it's working, which is that we put in place a bunch of acquisitions a few years ago. We replatformed it on to build. We're doing the same thing in precon at the builders the site business -- we then extended our relationship with our large integrated design build companies at the high end of the market. We've been building the on-ramp and precon which is a bit that connects design with construction. And then we let up our channel a couple of years ago to go into the low end of the market. And what we talked about at Investor Day and at AU is when also launching a sort of much more easy-to-use version for the site business as well over time as well. And what that is then slowly doing is squeezing into the sort of middle bit of the market, which is where some of our vertical competitors operate a bit. And so what you see the result of that, and also we're rapidly iterating as well and getting better, and therefore, we have reaching site future parity and then future superiority is we are -- where we're coming head-to-head for greenfield business, we're winning more -- much more than we're losing. And then also beginning to get into some of the renewal business and some of our competitors as well. So that's quite fun, too. And then also then replicating that across the globe, adding in new applications like pay apps to build out the ecosystem and expect us to do more of that over time as well. So in terms of growth, we don't precisely split it out. And actually, as we integrate the construction business more tightly with the design business because that's our strategy. It doesn't make sense to talk separately about the design and construction business. It is a connected workflow end to end in the cloud. But as a rough proxy, the growth of the make business is the best rough proxy you can see externally. And in terms of new customer additions, obviously, we're continuing to add a nice number of new customers too and certainly better than the -- a number of our peers.

Yes. Absolutely. I certainly shows. Mike, maybe just on that point, I mean construction, meaning out in the field, right? Like it's such an intense workflow, right, just in terms of the communication, all the changes that are happening, maybe a similar question we talked about with architecture earlier, but you kind of think about the opportunities for leveraging AI in PC?

I mean, again, I can mention Construction IQ, right? There was a tool that we realized nearly 15 years ago that we could understand the patterns of behavior for subs and start to have -- we used to go at the weather report, you arrive at the construction site every morning and it tells you where you need to put your attention on the construction side, people loved it, right? That was an early tool. This is pre the sort of deep learning era we're in now. Now what we're seeing is that -- just like you said, there's so much information produced in the average construction project, but just waiting through getting an AI that can actually help you focus on what you need to focus on bring summaries, to detect changes, to predict issues that happen. This is where we focused on right now. So much building on everything that Simon said all these tools, all this platform that we've built out. This gives us the sort of data nexus, if you want to now apply AI to begin to sift through that, to give you what you need. So that -- our customers are loving that. When we show them just simple integrations of language, large language models and vision language models into that. that can go through a bunch of material and say, it looks like that picture is a leak in the basement. Something's going on, I found this other document that was published at around about the same time, you might want to look at all these things together. Just a single occurrence like that can save the company hundreds of thousands of dollars. So just being able to bring AI on top of that massive amount of data that is constantly changing, huge opportunity.

And I just want to connect 2 things that Mike said which is that -- which is essentially the AI within a vertical, a silo and the value that's what he was just talking about, but also what he said earlier with former in the conceptual design phase is essentially what AI can do is bring what I call good enough simulation much earlier into the process. So the way I think about it is it means that you start off pointing in the right direction much more, much more precisely, which means that you then avoid problems later, which cost you money and cost you time. And so if you're just in a vertical silo, you will never be able to do that. Whereas if you can connect the data with AI in the cloud using it and then use AI to make those work and those systems -- automation, that's really what we're talking about. That's also -- that's when I'm talking about creating new value, that's some of the things we're talking about. The Holy Grail here is something called constructability. When you're designing the software needs to understand what is constructible and how to make it constructible and that's our north star.

We talked so much about product and AI. I think one of the things that is helping with this, right, is actually the agency model shift, right? Because now it creates such -- so much of a closer tie-in between Autodesk and the end customer, right? Obviously, the partner is still very important, but I think there's a little bit more of a direct connection between Autodesk and the customer. And so I'd love to shift to this topic a little bit and maybe start with you, Simon, because I think we're getting into the later stages of that shift. Maybe the question is just for some of us that are kind of fine-tuning our models, how should we sort of think about that revenue benefit as we go into next year? And of course, I don't expect you to give a numerical answer exactly, but what are some of the breadcrumbs that you've given as we think about sort of that underlying growth rate of the business, which I think has been pretty healthy, plus whatever mechanical items might be on top of that.

Okay. So just to sort of press on what you said the way you because pointed to the changes that we've been making over the last few years and with apologies to some of our long-suffering shareholders who've traveled with us through these changes. But the reason why we are -- one of the reasons why we are years ahead of our competitors in AI is not just because we've been doing a bunch of work on the model is because we're doing a bunch of enabling work to enable us to do it at scale without blowing a hole in our margins, basically. And 2 of the key components of that are going more direct to customer, which is, as you said, the new transaction model but also or partly into the new transaction model and also building a sizable and automating the back end of a consumption business. And if you don't have those 2 things, in addition to a subscription mechanism, then you're not going to be able to do AI, profitably do AI at scale. And so I know it's been painful to do, but it's one of our causes for optimism the reason that's cause for optimism that very few of our competitors have done any let alone all of those things. and they're going to have to do all of them, and they're going to have to do them quickly. Otherwise, they're going to be in trouble. So that's one of our causes for optimism that we've done those hard things and have already done. As it relates to sort of just numbers, we haven't precisely said. So we've told you that by fiscal '29, the difference between our goal of 41% and stripping out the new transaction model is about 400 basis points. And you know that we've already -- by the end of this year, done 300 basis points of that headwind has come through by the end of fiscal '26. So there's 100 basis points of headwind to margin still to come. And just math because of the way the contrary revenue, as you know, flows through into sales and marketing, a big chunk of that will come through next year. So that will allow you then to estimate what the margin headwind is next year and then if you gross it up, that will enable you to do some math around what the revenue tailwind is for next year. So that's probably the easiest answer -- the way to answer the question. We haven't given any underlying guidance for next year. What we have said, just to repeat what I said at the beginning, which is that we will be mindful of what impact the economic cycle could have on the business next year, and we will be mindful of the potential disruption from our sales and marketing optimization when we set guidance next year.

Absolutely. I think that's very thoughtful. And of course, maybe bringing the topic of margins back a little bit. I mean, Mike, I think Autodesk went through an optimization exercise of its workforce earlier this year that -- I think part of that exercise was to enhance resources and kind of growing parts of the business like generative AI. So maybe the question it's a bit broader is what are the areas that you want to be investing in, in '27? As you think about that quicker product cycle?

Giving away too much.

Understand.

No, look, so we've begun the journey in investing in platform several years ago, as you guys know it. And that has been where we've been putting a lot of our money, and that's where we're going to continue putting investment right? The platform is becoming increasingly more important in this world of -- you'll be seeing these agentic workflows that have appeared in this last year. If you don't have a platform agentic workflows are meaningless because the agentic workflows literally require a platform in order to function. So for us, permitting to all of these agent-based architectures and offering agent things, we've been able to respond to this really quickly because we had a platform. We had 180 completely separate products, there's no platform behind it, it would have been tough, right? So continuing to invest in platform is going to become important. Underneath that, though, data, of course, our data systems, if you don't have proper data flows, none of the stuff works. And it's the non-sexy stuff, but you've got to get this kind of stuff. Its foundation. Our customers expect it said we've been working on it. We've made incredible progress powering a lot of the AI stuff that my teams are doing. And then, of course, the AI stuff. So I mean, what you will see next year is increased investment in AI all the different platform aspects and then specifically around the sort of agentic ecosystem as well. You see a lot of investment on the -- certainly on the technology side.

So just to talk a bit about data and why that's important and how hard it is if you have on-prem data versus cloud data and all accessibility.

Yes. So I mean, there's sort of 2 aspects to making AI work, right? One is the sort of data side and one is the actual physical infrastructure side you need to run AI. On the data side, our customers put in an enormous amount of data into the Autodesk ecosystem, which is fantastic. So we get to provide their data back to our customers to learn from their data, provide services to our customers, do all this kind of work directly and make it very, very easy for them. So that's one part of making the data equation work. So when we standardize data models that makes that flow of much, much, much more easier. On the -- but not all of our customers' data is necessarily in our cloud either. So this is where our APIs and our platform becomes incredibly important to the they can still leverage our tools without necessarily having to commit to putting everything in that the data is still flowing through our platform. So getting that entire data ecosystem right, both in the cloud from a sort of standardized architecture and then also providing these APIs is in -- so that's what we've been putting in place. So that's the sort of data side of that. On the infrastructure side, the other thing we've been doing is spending a good chunk of our time on the AI side is actually building out all the infrastructure for a large-scale AI training and then, of course, for large-scale inference. As we deliver these things, that's where you see your cost obviously climb because that's proportional to transactions and users, right? So we're in Amazon House, all of our stuff is on the Amazon architecture. I was at Amazon Reinvent just 2 weeks ago. And we are doing things on Amazon that nobody is doing right now, certainly in our space. because we are able to scale these models, we are able to train them at levels that even just 3 years ago, we couldn't imagine being able to train things sufficiently. But it's been a journey. We've been on the 3-year journey now to really get this infrastructure in place now. So now the good news is we have it in place, which is why you're going to start seeing more and more AI models come out from us, not because we're suddenly inventing more, but because we've actually got this infrastructure that enables us to just train more models as we go.

So that's the front end of the research getting -- coming down, then the productizing of it, right.

Definitely a theme in both of your answers in terms of, hey, we've done a lot of the hard work that a lot of our competitors have.

You got to be early, right?

Absolutely. Yes, absolutely. For sure. Simon, I want to go back to the model, right, and just maybe just to put a bow on the margin discussion. One of the things that, frankly, I've struggled with a little bit as I think about the model is kind of the shape of margin expansion between now and fiscal '29, right? You mentioned the target of 41%. And as you said, as Janesh said, that path should be nonlinear. But he's also said that fiscal '27 should really bear the full brunt of the new model, which, of course, implies lower margin. So I just wanted to ask how you kind of thought about it because I kind of thought about the revenue impact should sort of be incrementally less next year. And therefore, the margin impact will be incrementally less. But I just want to make sure I tie out my thinking with how --

I think the -- your understanding is what we intended to communicate, which is the impact of the new transaction model is cumulative. So by definition, the big impact, when you finish the transition from the pre-revenue cost into the sales and marketing cost, that is when the maximum impact we will be largely done with that. I mean it will continue to build. The transition will be largely done by the end of the year. So in terms of absolute number, it will be the biggest. But in terms of the increment year-on-year, that is much smaller than it was in '27 than it was in '26 versus '25. So I think you're both correct in your understanding in how we work, but I know there's been a misunderstanding. Yes, it's the increment that people care about, and that is smaller next year.

Okay. So maybe to summarize that, I mean, it seems like the dollar impact of the transition really I mean if I use the word peak, right, it should really kind of peak next year in terms of the dollar impact, but the incremental margin impact. To your point, it's been cumulative, right? So we should be thinking about -- understood. Correct.

So the peak incremental is this year and then the dollar is next year.

Understood. Understood. Very clear. Maybe I want to just to ask Simon last modeling question, then I want to wrap up with Mike. I mean just to shift to billings here, Simon. I mean we'll be finishing the invoicing transition, I think, in Q1 of '27. So it sounds like we should see still billings pretty healthy through Q1. But I think with long-term deferred, only kind of mid- to high single digits of total deferred at the end of the year. Should there be anything that really drives the difference between kind of underlying revenue and underlying billings growth going forward? Or should those 2 kind of starts to approach each other?

I mean the short answer is yes, is that once you get rid of the noise, the new transaction model and the billings transition to annual billings, yes, billings and revenue growth should start bearing much more relationship to each other. You then get the usual noise around different cohort sizes and a bit of FX in there. But substantially, thank God, billings will become much less -- well, everything would become much less noisy.

Correct. Absolutely. Well, listen, in the last minute that we've got left, Mike, I've got a fun question for you. I mean just -- as we think about that cycle time is a really interesting point that I haven't thought about, right? Which is just the amount of product development that you could be doing over the next year. When we're sitting here on the stage in a year from today, what are some of the milestones and the innovation that you want to be talking about as we look back on calendar '26?

Yes, yes. So what I want to be able to do is look across our industries and have the actual designers, the architects, the engineers understanding what AI tools actually look like in the future -- because the problem is today, most engineers and architects go and look at ChatGPT or Gemini and go, "Oh my god, you're going to make a product with a big green button that says make a building. I'm going to put it in a prompt and it's going to do that." Believe me, no architectural engineer wants that set and nobody wants that. In fact, it doesn't work because these things are way more complicated, as you can imagine than that. So -- but what's happening is nobody is yet producing the tools, but we are now beginning -- beginning to produce those tools. So if we're sitting here next year, I would love to have the sort of suddenly have this vibrant ecosystem, much like we have in the public around things like ChatGPT, but in this case, it's not ChatGPT. It's a bunch of incredible AI features they're enabling architects, engineers and designers to start creating the future world in a way that they've never been able to before. And they're leaning into these tools because they're not replacing their jobs, they're making them more efficient, they're making better products at the end of the day, and they're building on that sort of success. It's fundamentally transforming the mechanism of those industry. That's where I want to be in. It's a big goal, but that's where I want to be.

It sounds like it will be a fun year, though. Well, with that, I couldn't think of a better way to end. Mike, Simon, thanks so much for being with us here today.

Thank you. Appreciate it.