I'm Joe Verwink. I cover vertical software at Baird. I'm very happy to have Autodesk. This is software for the built world and what architects and engineers, manufacturers, operators, they use it all. Sid Oxar leads the construction strategy, Simon May Smith, Investor Relations. This is going to be a fireside chat format. If you have questions, you can email session2 at rwbaird. But let me turn it over to Simon and Sid for an intro first.

Yeah, so just sort of briefly, for those of you who don't know Autodesk, what we're trying to do is connect workflows end-to-end in the cloud with a layer of AI on top of it, something we've been working on for almost a decade and years ahead of our competitors, not just in the hard stuff, the frontier model building, but also the technology stack that sits underneath it and how we ingest and process data. We're operating, as Joe said, in AEC, in manufacturing and media and entertainment. And we're pretty excited about the future. We've also made an acquisition last week in operations. So we've had design and make and now extending into operations to complete the data across the asset lifecycle. So we're pretty excited about that, too. But I'm sure we're going to talk about it.

Why don't we talk about that? So if I can channel all the questions I've gotten on this, I think, one, strategic reaction. rationale and why something like this now, two, price paid and whether that's a fair or unfair valuation, and then three, I think a lot of investors associate Autodesk upstream with project delivery, not downstream with how these different disciplines operate after the design. Is this a totally new undertaking for you or is this more of a gap fill around things you've already been closing in on?

So I'll start with the last one because it also answers the first one, which is that the ultimate customer for our entire business is the owner, the asset owner. So right at the front end of the process, it's the owner that is trying to make something in manufacturing or is trying to build a building to generate a yield from it. And that owner then commissions a construction company, a design company, and a construction company to build it, and then somebody to operate it. And so what owners want is to understand how their asset is performing across the asset life cycle. But to do that, they need data. And today, data is stuck in silos, you know, thousands of different silos and not brought together. So in simple terms, what we're trying to do is to create a single model from right at the beginning of the process in conceptual design through to the end where you tear down the building and hopefully recycle and put up a new building. And so what we've been doing for the last, what, 15 years is building out a connected data, common data environment, as it would be called in AEC, starting in our traditional business in design, building into construction, which Sid has been responsible for and we can talk about a bit. And then the latest step is then the final stage, which is in the operations phase, which is the post-construction phase. The reason that phase is important, and the reason now is because we've built our construction business now to a sufficient size and sufficient momentum, where we're beginning to tear down the leaders in that field and take leadership in that field, that we now have bandwidth and capacity to then now focus on operations. So in terms of operations, we bought a business called MaintainX, and the reason we did is that that operates in one of the core functionality bits within the operations phase, the sort of the maintenance part. And the reason that's important is that that is a core piece of functionality across all operations assets. So whether it's a factory or a commercial building or a piece of infrastructure, every single one of them will need a piece of software to enable people to maintain it and keep it up and running. And so if you look at the sort of 40 billion TAM for operations, the biggest single chunk of it is in what's called the CMMS market. So that's the key piece of software. So that's why MaintainX, that is why we call it our cornerstone acquisition. It's the big chunk because it sits a central role. And it sits on top of a piece of software that we already have, which is the digital twin, which is the final as-built version of a building, which you then plug into sensors which allows you to monitor and over time with AI predict faults in the building and then what maintain XR does is when something goes wrong it allows you to then take action and fix it basically. So that's what it is. In terms of the multiple paid, a few things to think about. Firstly, as we've said, is we're following our construction playbook. So the construction playbook, like you said, is we spent about $1.8 billion on our construction business. We built a business that over the last 12 months has generated about $600 million of revenue, so that's three times revenue, as you can see. It's a pretty good multiple, and it's growing more than 20%. So if you look at the multiple that we've paid for MaintainX, just think about a path as we build it up, as it grows rapidly, and that multiple will come down pretty quickly. In terms of the opportunity and how we do that, there's a few things. Firstly, if you look at the construction TAM, it's about an $11 billion TAM. The operations TAM is a $40 billion term. So much bigger market opportunity for us is the first thing. And the second thing is the duration of that term, is that our design and make business is a year's business in terms of our interaction with an asset. The operations business is a decades-long business. So once you've built the building, 80% of the cost of a building is post-construction. And so managing the efficiency of that is, for the owner, is of critical importance. So that's what we're trying to do. So at the moment, we can only address, in terms of efficiency, 20% of the cost of a building it's the other 80% that we're now seeking to address with the acquisition of maintenance building on top of tandem which we bought built in organically ourselves and then the sort of final thing is data is that the maintenance business is a cloud native business mobile first business the vast majority of the traditional incumbents in this field are on-premise software custom integrations very expensive but the key thing is that getting access to the data with on-premise software is very hard. So what maintenance does is it has eight years of data which is useful and which we can apply our AI to not just in the operations phase but we can then with inference use that operations data then start making inference upstream in the conceptual design phase. So when you're doing conceptual design right at the beginning of the process if you can have something saying don't install the HVAC system because two years after construction you're going to have a problem. That is immensely valuable information for our customers, the owner. So that's what we're doing. I should probably stop there and get on to the next one.

No, that's great. And you kind of hinted it's growing 50% right now, but 50% also not unreasonable to think about next year as well. So it starts to work the multiple now.

I'm not going to give you a revenue forecast or an ARR forecast because we haven't done it, but I'm not allowed to. In terms of the opportunity, MaintainX has focused primarily on factories to start with, and is just beginning to think about a few other things where we can be quite helpful to them. The first one being AEC. As we know, we have a very, very large AEC business. And as I said, the assets, the commonality of the maintenance system is transferable across into AEC as well. So that's something we can help them with a lot. Secondly, we can help them with enterprise is that they've, because they're a small startup company, been focusing on single assets and single sites. What we can do is help them up-level those conversations to all of the assets owned by the owner across the country or across the globe. And then the third one is international, is that they are primarily a U.S. company today, and we can help them expand internationally, both with our sales teams and our e-store, but also with our channel partners, too.

I think just one thing worth adding is that while the MaintainX acquisition showed up, I think, last week, we've been looking at the operations space for over four-plus years. um it has been a natural progression because as we serve the needs of owners on the construction side the next foray for them you're targeting capital projects teams but then their facilities teams so it's a very nice adjacency for our owner base we also made an investment i believe i think was about four years ago in a company called eptura that's owned by toma bravo so while we've been investors in the company we've also been learning the space very closely and understanding what's working then what's not so in a way we've also de-risked a lot of how we think about the space

going into this acquisition so to give you an example which sid's been working on uh with the new england patriots is that we're sort of helping them build their stadium but one of the required outputs of that project is a digital twip because they're already thinking in the construction phase is how they're going to manage the asset uh once it's being built so owners are thinking about this instead of weight. Great.

Maybe let's go back to the construction piece. And you talked about that $11 billion TAM. I'm going to ask two questions. One, if you look at that TAM, Autodesk has done very well accelerated growth into the 20s. But a lot of your peers have also accelerated their growth over the last few years. So there's something happening in the category itself that is allowing for more success. Maybe we can talk about what you're seeing at kind of an aggregate or macro level, and then we'll get into how Autodesk is different therein.

Well, there's one that's notably decelerating in construction, but do you want to take the question?

Yeah, so just generally, while there are different pockets of the industry, there's some puts and takes, right? For example, right now, data centers are on fire. There's power grid upgrades that are happening as a result of the data center or the AI infrastructure that's coming up. So there's a lot of growth there. We're seeing health care growing really rapidly. We're seeing also stadiums, believe it or not, at least in the U.S., are seeing a really nice ramp as people are doing big capex upgrade cycles. That said, the industry itself is still not as digitized as you would think. It's a very big industry. And so there's just this push. If you just look at within the United States still, there's plenty of opportunity of getting companies that are still sitting on either Excel or using very low-grade ERPs to manage projects. If you take a step out of just the U.S., then you start to look at, say, for example, countries like India, which is the third largest construction market now globally. It's caught up in a very short period of time, fueled by the infrastructure boom as well. If you go and travel to India, you'll see construction really is being done using paper and Excel. It's surprising, but it is. And so now those companies that operate there are seeing that in order to do a good job, you have to use technology. And so there's just this secular trend of people using tech and our solutions to manage these projects. And then when you talk about labor shortages that are impacting the industry, schedules are compressing, projects are becoming more complex. you really need to be building right the first time. So we don't see that stopping. People have to invest in tech to become more efficient and deal with it.

Do you want to talk a bit about why having the design and construction tools together is so important?

Well, so for us, the differentiation coming out of... So that's kind of the general theme as to why not only us but our peers as well have all benefited and grown with it. So there's plenty of opportunity out there. But when you bring it back to what's different from Autodesk relative to our peer group, obviously this is one thing that we don't talk too much about. But Simon said $600 million, that's cloud. You think about how much are we generating from just the construction industry, selling them not only our cloud tools but also our modeling desktop tools. That's in excess of a billion dollars. So I say that because we've already got a very strong install base of contractors upstream using design. So that's one of our key differentiators. Historically, form of a construction wasn't as robust mature, so there was a need for contractors to go use what was best in class at the time. Given where we've reached now, the story of being on one platform starts to resonate tremendously. Obviously, when you layer in opportunities with artificial intelligence, having access to your information across the project lifecycle becomes that much more critical. So upstream, we've got a very strong foothold. We've matured our platform to be end-to-end. The third thing I'll add is flexibility when it comes to our business model. So we are not just wedded to one particular type of model. We can be user-based pricing. We can be a percentage of construction volume. We are consumption when you look at some of our enterprise customers. So that's the third piece. I said, and then the fourth piece for us is really our geo footprint. And that is enabled really by our channel partners. So we already have a very strong network across multiple countries. So that allows us to go to market at scale.

And just to sort of piggyback on the back of that, that connection between design and make as a competitive advantage, that same thing is true in operations because the final version of a building is the digital twin. And then extending that with things like MaintainX into operations, it's exactly the same strategy.

Let's talk about AI and maybe to start a bit of a thought exercise. So if you think about being a protein scientist or like a coder, AI has changed what you do forever. So let's rate that a 10 out of 10.

When you think about construction professionals or even going into design, just the work that architects do, like where would you peg them on the same 10-point scale so for us I'll talk about construction I think it's very early right now one I probably say between a one and two is kind of where it is and the reasons for that obviously adoption of technology new technology it takes a little while as we've seen when Especially when you think about where you can see a lot of the impact, I think a lot of the impact you're going to see in the field. I think people that are working out of the offices, it makes a lot of sense. Just to give you, you don't need to have some really complicated use cases to get value. There are some very basic use cases. So one thing that we have today, just imagine for a second you're a superintendent on a job site. And you're looking and making sure that everything is working in order. And you find a pipe that has a crack. So now you need to create an issue of that and let people know that, hey, there's this pipe that's cracked, so we need to fix it. Typically, when you create that issue, you have to document that issue. You take a photo of the crack, and then you document it. And imagine a big job. You end up spending a lot of time documenting issues with AI, and today we have that in our product. You can take a photo, and when you take that photo, the AI will tell you what that is using our computer vision, and it will auto-populate the description of that issue. So what would probably take about two minutes now gets compressed to maybe 15 to 20 seconds at the most. The individual just looks at it, makes sure from a quality control standpoint, it's right, and then it gets sent out. That, again, is something, it's not, you know, very complicated, but it saves a ton of time for people on the field when right now there's massive labor shortages, and you need that superintendent working on more impactful things than actually documenting stuff and spending time doing manual work. So we do believe, just given where labor shortages are going to show up, we do expect in the field you're going to see some outsized productivity gains with the use of AI. On the flip side, in the office, we're starting to see that take off more just because the user base is more attuned to using technology and some manual tasks that can be automated. They're embracing that. But the field is where I think you'll be able to get a lot of productivity gains right off the bat once this gets more mainstream.

So I wanted to ask on that because the use case you shared is super important, very valuable. I would say I actually see more kind of AI in pre-construction and then in like the scheduling aspect where the field matters the most.

So what's kind of the disconnect where people are focused on kind of the edges versus what we're talking about doing in the field is most consequential? why why the focus on pre you know why why construction yeah why is that been the initial focus it seems so first of all it's people that are in in the office are more receptive to technology and also the other piece i will say is projects are made or lost in pre-construction so if you end up scoping a bid inaccurately then the margins are going to fade once you get out there on the job side if you don't capture the right if your scopes of work don't match what are the specifications articulated by the architect you may install work that then has to have a significant component of rework so you need to remove what you put in place again it has an impact on margins so getting all of that right happens in pre-construction so that's why you're seeing a lot of companies come in or AI technologies try to make that as robust or as a risk mitigation tool. So that's what you're seeing is risk mitigation in pre-con. I think you're going to see productivity really manifest in the field.

The challenge, and maybe we can talk about this, is most folks, companies, lack data and context, and they also lack 3D engineering capabilities too, and that's something Autos has in space, but I'll wait and see if you want to talk about that but no I think you run a survey every year around like AI acceptance and where the interest is where the pain points are and I think the biggest pain point and the most recent survey was just system integration like you don't have anything connected and so the data might be out there but you have no idea how to actually use it exactly so this is why what Sid was saying in the connection design and make and then what I was saying about operations then connecting that the design to make into

operations that's why that's important what we are also starting to see increasingly within our customer maybe it's historically companies have had a whole hodgepodge of point solutions that's going to start to go away as they continue to consolidate their spend in specific platforms so that's another thing and partly is what you just raised because of data sitting in so many different silos and not systems not talking to each other one thing I've always appreciated about Autodesk is you're making these investments before anyone is asking you to do it so like cloud was 2010 give or take and then forge which became the platform services strategy a few years later and then you were kind of getting data ready for

training before gen ai was a thing i want to focus on autodesk platform services so that began as an api strategy and i kind of think that's morphing into the mcp server strategy to the point where when Claude is looking to embrace, you know, creative companies, Autodesk is part of that announcement with MCP servers you've built out, kind of the important, it gets to what we're talking about, of making it easier for customers to move data around, but the importance of the strategy and AI investments that are now happening.

Yeah, so let me talk about that subject close to my heart. So just a quick detour. And by the way, everything I'm about to say, if you look at our Q4, the last four pages of our Q4 opening commentary and the last four pages of our Q1 opening commentary, strongly encourage you to read them if you're interested in AI, what I'm about to say is in there. So not related to that question, you need data context to build a foundation model, to build a knowledge graph in which you can build a foundation model. Data is scarce in our industry because it's not available on the public web. It's locked up in a million and one different companies' systems. And so if you have access to it through the cloud, which we do, most of our competitors don't because they only have on-premise software, then you can have enough data to build foundation models. You also need context because the assets that we're building are constantly changing. A building site on day 20 is different from one on day 30 is different from one on day 40. You have to know what's leading up to a particular point and decision and what happens after it. You have to understand the sequencing of how everything's put together. So there's a bunch of context you need and data to build a knowledge graph, and those are very hard to come by in our industry. And then once you've done that, you also need 3D engineering. Just to be clear, LLMs are 2D. They're sort of words and coding. They don't reason in 3D like our models do. And 3D inference is really hard to do, and we know that because we've been trying to do it for almost a decade. Again, years ahead, as Joe said, of need, so to speak. So we're years ahead of our competitors on that. So doing all that stuff is hard. We've been doing it for almost a decade, and we're years ahead of our competitors. But then once you're doing that and you're launching foundation models, that's when the technology stack becomes important. So just to give you sort of two examples, one of which is everyone worrying about token maxing and gross margins, something we've been talking about. Gross margin pressure is something you cannot escape as you put more workflows and high compute workflows into the cloud. So what you're trying to do is to figure out how to bend the curve. But critical to that is how you ingest and process data. So that's something and the reason why Autodesk Platform Services is so important is we've done a bunch of work over the efficiency with which we can ingest data and then how we process data. So I'll give you two examples. None of this sounds very sexy, but it's critically important. One is around the data model. If you go into our customer systems and look at all the models, What you'll find is that the data is fragmented so they have an HVAC system in one file They've got this structural building in another so if you turn up and scan it You don't have a whole building to make inference on and so the reason the data model is important is it brings all of that Disparate data back together again and allows you to extract meaning from it. We've done that work It's really hard math to do that and what it means is that for any given data set We can extract more value and more meaning from it in a scalable and efficient way than anybody else At the other end of the spectrum, when you're doing inference, if you try and put 3D inference through a stack that's built for 2D, it's very cost inefficient. It uses much more capacity and costs you more money than it needs to. So what we've done is we've built our own inference stack on top of AWS, which is massively more efficient at 3D inference, cost efficient, than doing the equivalent inference on 2D stacks. and virtually all of the other stacks are built for 2D inference because nobody's trying to solve our problem. Well, very few people are trying to solve a 3D inference problem. So that's why the technology stack is so important in terms of AI. But it's also important, to Joe's point, is around how you develop your offering. So what we've been doing, and this is a sort of technical debt problem, is we've built a bunch of stuff over the last 40 years and essentially building the same functionality across the organization. And what we've been doing over the last three or four years is creating more common components so that when you update something, it propagates across the entire product suite rather than having to go in and update everything at once. What it also enables us to do with the help of AI is to start creating new value. So one of the underrated things that we were talking about last week is we have probabilistic AI models, and we're using our deterministic parametric models, which we've been using for the last 40 years, to validate our AI models. And that loop, so we create a probabilistic outcome. Top tip, don't walk into a building that's been created by a probabilistic model because it might fall down. And we validate it with a deterministic model, which we've had for 40 years. And that loop allows us then to improve our AI models in that loop. Again, sort of massively improvement. But what we've done is those parametric models sit within our traditional products, you know, so Revit and Fusion, et cetera. And what we've been able to do is to extract just the parametric model and then plug it into our AI models to make them more efficient. Doing that a year ago would have been inconceivably hard to think about doing. But with AI and new engineering techniques, we were able to do that in three weeks, extract the parametric models. from our engineering. So one of our core beliefs is that AI is about doing more with the same number of people. It's not about doing the same with fewer people. But what you have to do is to be able to conceive of hard stuff and hard problems to solve. And I think that's going to be a key challenge for most organizations. Unfortunately, one thing Autodesk loves doing is solving hard problems. That's why we started trying to solve AI 10 years ago. It's why we started to try and solve the cloud 20 years ago. We always try and solve hard problems. So it plays to our strengths.

How does that all get monetized? I've heard that the Autodesk direct sales team is trying to encourage their customers to token max because increasingly you can kind of get tied into that. What's the strategy there?

We're not encouraging to token max. And actually, and this is again another, which I will discuss with a bottle of whiskey if anybody wants to, but it's not about maximizing tokens because if you're doing that, you're going to find soon that your AI agent costs you more than a human being does. It's actually about sipping, not sucking at the tokens. and building the software that enables you to do that. Otherwise, you're going to have products that are too expensive and not doing the job that you need them to do to drive efficiency in the industries. What we are trying to do is to enable our customers to try stuff and to build stuff and try stuff, partly because that's for their benefit. But what also it does for us is it generates data exhausts, which are useful for us as they do that. So, yes, we're trying to encourage them to do it. What we're not trying to do is to consume bad calories from it. We want them actually to become more match fit as a result of it.

And that's bringing customers into kind of a new subscription, different subscription than a seed tied to a model. These are bundles of API uses that you can monetize.

So exactly that. And I think this is, I don't know whether it's consensus, but emerging consensus is that the subscription is going to be around for a very, very long time. And including a subscription will be a core level of functionality and a core level of capacity will be included in your base subscription. And then if you need additional capacity for high compute workloads and high value workloads like AI, you will essentially buy additional capacity. So very similar. So, you know, little known facts, 17% of our business is already consumption, which is a type of capacity model. There are others. But it behaves financially like a subscription because essentially people buy capacity ahead of time and then consume against that capacity on a use-it-or-lose-it basis. So consumption doesn't have to be volatile. So you can give the customer the benefit of flexibility and certainty whilst also enabling us to have predictable and routable revenue streams. Maybe with the little time we have left to talk about something more recent and you've taken control of your sales channel, you used to have a two-tier distribution model, now you're you're direct um what's been kind of the the biggest learnings from that and benefits that you originally thought you would get are the benefits coming through yeah so i so i said only you know auto superpowers is we try we do hard stuff um and so this is a good example but it's sort of a function of and you should be asking all of our competitors this is there's a bunch of stuff you have to do just to get onto the starting grid of ai you know you have to sort out your technology stack for the reasons i talked about you have to have cloud-based software you You have to have more direct integration with your customers, which we're going to talk about in a second, that's why I'm mentioning it. You have to have a bunch of different business models, so you can't just be subscription. You also have to have metered access models like consumption, et cetera, as well. All of these things are really hard to do, but Audesk has been doing them over the last five to 10 years, which means, which is why, and also you have to invest in AI and the engineering capabilities to build foundation models. They're all really hard to do. They all mess up your P&L, balance sheet, cash flows, margins while you're doing them. And we've been doing that for the last 10 years, 15 years in the case of the cloud, in a way that most of our competitors have not been doing. They have to do it, and the time they have available to do it is getting shorter at a faster rate because of AI. So the risk of getting it wrong is greater. So the most recent one we've done is our sales reorganization, But the intent is essentially to have ourselves more directly integrated with our customers, enabling that with more things like more self-service, more auto-renew, more co-term, et cetera, so that you have more automation essentially in the process. And then using that then to help our customers build on top of us to drive more new applications for them and more revenue opportunities for us as well over time. In terms of the hard, it is hard to do. So we had a significant restructuring earlier on this year. We also, at the same time, you know, ripped all of the custom stuff we'd built on top of Salesforce, put ourselves onto the base Salesforce platform, and then have added a lot of the AI functionality that Salesforce has introduced to enable sales productivity, et cetera, which we weren't able to do because of the customization we had on the platform. So all of that stuff is hard. it creates disruption, which we've talked about, but well within the expectations that we'd set out in February.

Great. With that, we're at time. There will be a breakout session, but please join me in thanking Autodesk. Thank you.