There we go. We're live.
Testing, one, two.
Yeah, we are live. All right. This has been an ongoing tradition. I don't know how many years we've done this in a row, but I hope we do this for another decade. Greg Bentley, Executive Chair of Bentley, still very, very active though. I love your script in the prepared remarks of the conference call. It's great to see that you're still so involved in the company, and it's great to see the recent results. Greg, from all of us at RBC, thanks for joining us here.
You guys are coming off a really nice quarter. I think it surprised a lot of people out there in terms of the outperformance, the raise to the full-year guide, which candidly I wasn't expecting after 3Q, and I think the enthusiasm for the business. Can you talk about what, when you step back and you sort of think about the quarter, and it's a journey, right? It's not just a one-quarter thing. Talk about the journey, talk about why Q3 was so good, and your enthusiasm as we think about calendar 2027 based on some of these drivers.
Well, Matt, thanks for having me here, and thanks to each of you for your interest. For those perhaps less familiar, Bentley Systems is the infrastructure engineering software company. We think of ourselves as the digital quartermasters for civil engineers, structural engineers, geotechnical engineers, their organizations, and those who work with them on the world's infrastructure. Half of our business is the owner-operators of that infrastructure, and the other half are the project delivery firms, the engineering firms who are their supply chain.
I might say also that the company was founded, I guess, 44 years ago now by five Bentley brothers, and we still own the economic majority of the company. So our perspective is indeed very long-term intentionally, and that's a reason, among things, that we give only annual guidance. But it is the case that throughout this year, each quarter has improved in terms of our ARR growth. So we did raise our annual outlook for 2026, and it has to do with the world's infrastructure engineers are BSY, busy as they possibly could be.
The market conditions, requirements for replacing aged infrastructure, but especially for resilience now, for self-sufficiency and so forth, across our sectors, starting with public works and utilities, and including grid especially, growing very strongly. Next, resources, including mining, are going strong. This is true across all regions in the world. We've also had a good year and quarter so far in our asset analytics business, which is newer and a bit lumpier, where we charge per asset per year rather than for individual user consumption. So those things have all come together auspiciously.
As far as the end market conditions, there's nothing about those that shouldn't prevail and pertain into 2027 as well, I think. Our asset analytics business will have more variability from quarter to quarter, but for the year as a whole, I am indeed enthusiastic about the future for some other reasons also that we'll talk about.
Yeah. So I wanted to double-click on asset analytics. I think there's a lot of sort of unique elements to that business, especially as we think about AI. Because I think the way you see it is it really closes the gap between CapEx to OpEx lifecycle within your customers. You also mentioned some of these deals can be large and lumpy. Can you talk about sort of the predictability, the visibility of these deals, and why now? Why are these doing so well right now?
Well, asset analytics is inevitable. It involves instant-on digital twins for infrastructure assets, where especially drone flights, but other modalities as well. In the case of roadway miles, we use dash cam data and then AI filters to detect conditions. Then what's unique in our approach is we apply the engineering models and simulation that the owner-operators already have, but they have never used the engineering modeling and simulation for their assets during operations and maintenance. They've used it only once during project delivery.
So with a digital twin, you want to be continuing to simulate the throughput and the resilience and the renovation and optimizing the maintenance for your fleet of assets. Asset analytics enables us to start this right off the bat using continuous surveying modalities. Our competition in asset analytics are sort of hardware-oriented approaches, and we're agnostic as to hardware. Our differentiation is being able to use the engineering logic and data to determine what's actionable as a result of what you see in capacity and conditions of asset infrastructure.
So I describe it as a keystone for us between OpEx, where the opportunity is, because that's where the value is generated for infrastructure assets. CapEx, where we have the data and the engineering and simulation has been done, it's just never crossed over to help during OpEx. I think it's going to be a large business, and our opportunity there is the engineering logic to make sense of what those existing conditions are. You're pitching utilities in the main, is where we are now, on doing this for their fleet of towers or their fleet of poles, or in the case of roadway operators, their miles.
Those are big procurements and big opportunities. So they occur. It takes a longer sales cycle, and when they drop in, it's less easy to predict. It's unlike the rest of our business, which grows granularly with additional consumption and additional seats of established software. So on the margin, it adds volatility, but our existing business is such a flywheel because of all the busy civil and structural and geotechnical engineers in the world, and it's predictable how much software they'll use and so forth.
On the margin, adding a little bit of variability. We think the business will get better as it gets bigger in terms of more predictable ARR year by year, but we'll still have quarterly variability for some, as a result of the procurement scale.
Well, I think that's what NRR ticked up this last quarter too, like you said. The base that you've built over the years in terms of the predictability of this business has been remarkable. The NRR ticked up, you guys beat, you raised. Can you talk about the bigger elements behind the NRR tick up? Especially as we get into 4Q, you've got a lot of E365 renewals, there's cap and ceilings. Is there a structural ceiling to what NRR can do with this business?
Well, tick up is the right way to describe it. For some reason, inexplicably, we round NRR to the nearest full percent, and for ARR growth, we do it to the nearest half percent. NRR trails ARR. Of course, it compares a year, the year ending to the year before that. Throughout that period, ARR has grown, as we were earlier describing, and for all of those reasons. So NRR kind of naturally tracks ARR. I do want to explain this, our E365, our mainstay enterprise consumption program. The floors and ceilings that we negotiate optionally, there doesn't have to be a floor.
In E365, we charge for daily consumption of our applications. There doesn't have to be a floor and a ceiling, but an account will say to us, "Well, we know we're going to be using more software. We're not sure how much that's going to be." Also, we have pricing escalation each year in mid-single digits, and they'll say, "We don't want to be surprised on the upside. Would you be willing to put a ceiling on how much we spend with you?" And we say, "Yes, if it's in double digits, and by the way, you ask for a symmetrical floor."
The floors and ceilings tend to be 10% or down 10% from where the current consumption is. What we charge, what we earn as revenue, what we book as ARR will depend on the actual consumption, but there's a collar that bounds it. However, if you think about it, from year to year, we do tend to report that the virtue of the floors and ceilings for us is it gives us visibility when we negotiate them with an account on what their own expectations are of how much more they'll be spending on our software. That tends to be about 10% is where we agree from year to year.
It's a little bit higher in the out years, actually, and you may optionally ask for a floor and a ceiling in beyond the next renewal year. But within any year, we're actually going to revenue and ARR would be what's actually consumed. If you think of it with a bound of 20% within the collar, there's lots of opportunity to grow faster than 10% in each account. We hope we do, and largely, that's our plan.
I want to go back to AI. Asset analytics certainly has an element of AI to it. When we think about monetization through API or MCP consumption, it feels like that's a calendar 2027 and beyond driver. I know it's still pretty small today. What are some of the guideposts that investors should look for to understand your success here and how this could influence growth?
Well, the asset analytics, which is enabled by AI, the most profound opportunity for Bentley in our accounts with AI is the agentic automation of what has otherwise been attended human consumption of our software applications, which is limited by how many people there are that can sit in front of a computer and do modeling and simulation. By programming agents to do that, the possibility opens up for optimization of designs.
Where previously a person could only explore a few iterations for a proposed water treatment plant or airport terminal or whatever, with agentic automation, you can explore a solution space with hundreds or thousands of alternatives. Then, of course, the human has to make the decisions about the trade-offs and so forth, but the result is a much better quality design that, by the way, you can quantify how much better in terms of cost and reduced risk, constructability, subsurface risk. All these things can just improve the practice of infrastructure engineering.
So, the key to enabling that is the MCP services to our existing APIs, and we have to expand our APIs throughout our portfolio. Throughout 2027, we'll get to where we have so instrumented the bulk of our ARR, and we are not yet monetizing this API consumption, and that's a deliberate decision. Well, two things. On the one hand, we have to complete the back office approach to that. We think we'll be done with that this year.
As a practical matter, when we introduce new versions of software with these APIs for MCP services, our accounts have to upgrade their existing desktop modeling and simulation applications, and they tend to do that only once a year. So, we want to make the best of this reality by encouraging as much experimentation and proving out, and in each account, experimenting with and demonstrating the increase in design quality that you can accomplish with optimization, and having that become ready to be the norm when the new software is rolled out to as fast as it can be.
But we have these various steps along the way, the milestones, as you say, most of which will take place during 2027. But our principal consideration is we want to maximize long-term adoption and institutionalize optimization on the part of our accounts for their projects, and we don't want to slow that down and deter it with concerns about monetization. We'll have visibility into the API consumption during next year, but we might well decide to wait before we throw the switch.
It's not as if we're bearing the gross margin hit of the AI consumption in the meantime, because our accounts are welcome to use their own LLMs for the inferencing layer at the agent level. The execution at machine speed of the modeling and simulation tends to be done on their existing computing resources locally and doesn't incur a cost for us. Our approach is, again, long term. We're doing fine and growing our ARR growth in all these other ways. If we can have the greatest inflection in the long term, that's our choice.
It sounds like there's still some instrumentation happening this year, but it's going to be more of a gradual. It's not like we should assume in 2027 that there's some
Everything is gradual in what we do in infrastructure engineering. It's because the projects last for years, and you can only kind of leg into new innovations even as they're available. What's important is for us in these E365 accounts to have the tool smiths in those accounts experimenting with and establishing the virtue of the new versions of software so they can ultimately influence and roll them in to be the production norm, but that takes a while.
Yeah. Resources has been going really well for you guys. I think Seequent is growing 2x the corporate average. Can you talk about, I mean, I think everybody reads in the press why resources have been doing well, but talk about the sustainability of this demand. Where are we in the curve for subsurface modeling and some of these civil projects?
Well, we've owned Seequent for five years now, and there was a period of time when it cyclically grew a little less fast, but still faster than 1.5x our company as a whole, and that was when interest rates went up and new mine exploration slowed down. So, there is some exposure to that in theory. But in resources, there's an underlying secular opportunity where what used to be 2D modeling is now becoming 3D modeling, and that opportunity was elapsed long ago above the ground, where you can see and model easily.
Below the ground, you have to sample and then infer and extrapolate it mathematically. So, that's not 3D yet as a norm. It's maybe only halfway through that adoption, and that's an underlying driver. A further underlying driver is the opportunity for 3D modeling for infrastructure projects beyond the geotechnical foundation design to understand the water permeation and environmental conditions and so forth. For instance, for data centers.
We might come back to that. But the conditions which motivate the resources demand now are not for greenfield mines, but for getting more out of the existing brownfield mines, and that's the main driver that's going on at the moment and is less subject to capital market cyclicality, I think.
You recently expanded, effectively, the Seequent portfolio with an acquisition of Eliis. Talk about what gap that filled in your business and how we should think about that in terms of aiding the segment.
Eliis is the up-and-comer in seismic interpretation. Seismic interpretation was spawned for oil and gas exploration, where the reservoirs are way below the ground and so deep that you can only sense them with big seismic waves in the ground, and it required a tractor-trailer-sized rig to do that. With the advancement of technology, it is down to a golf cart now, and you can do seismic investigations on smaller sites, such as mines.
For instance, if we take the case of copper primarily is mined in open-pit mines, but you get to the point of diminishing returns where it makes sense to go underground and go deeper, and that is where seismic comes into its own. The first incremental opportunity. By the way, Eliis is a tremendous acquisition in its own right, growing faster than our company, compatible margins, but with the cross-selling opportunity now to extend seismic interpretation into mining and civil projects. That is the first incremental opportunity.
The other is that the Eliis accounts, whom have been oil and gas companies in the past, are now diversified energy companies. As they go after geothermal, carbon sequestration, brines, and so forth, they have discovered that the mining technologies from Seequent are applicable there as well. We can go beyond the pilots that have proven that to now use the Eliis sales channel to extend into diversified energy.
You prompted the data center question, which really brings together multiple aspects of your portfolio from PLS to Seequent. I think on this last quarter, you mentioned digital twins becoming an increasing element of this. Can you talk about how we should think about the standalone growth vector for data center build-out? Obviously, there are questions post-midterm and that sort of thing, but talk about the data center business and what that means for Bentley.
Data center is a campus, and it is connected by obviously essential horizontal infrastructure networks, which are our sweet spot. We have explicit indications that major data center owners are understanding that with better visibility of their campus environment, and especially the subsurface and aspects of water and energy geothermal that can come from the subsurface,
..and to minimize the construction risk, and to have a digital twin to help them, on the one hand, in the approval process, and then in the economics. It is a future. Digital twins are the future for all infrastructure campuses, but it now looks like data centers are going to lead the way. That will be an explicit business driver for us.
Some of these projects are multiple years out. We talked about some of the cyclicality in the asset analytics business. How predictable or what sort of visibility do you have on the data center build-out?
Well, the school that the world is not going to be covered with data centers necessarily, in the work that our users do, there is such sensitivity to their data. If they're an engineering firm, it's their proprietary advantage, and they're not going to risk sharing or commingling it with anyone else. In the case of owner-operators, it's critical infrastructure data and is very sensitive that it not be exposed to hacking and so forth.
A lot of future computing, I believe, will be local and sovereign in our realm, and that might be a bit different than conventional wisdom. However, when we think about data centers, it adds a bit to demand on the grid. Even if data centers would be longer rather than stacked up immediately in front of us, which I think is the case, we would still need a lot more grid expansion benefiting our Power Line Systems business.
Yeah. We talked a little earlier about AI, and one of the questions that I get all the time is, do we have to worry about AI disruption, AI native companies? Prometheus is the one that has been in the headlines. I guess maybe a two-part question. As you sort of see the environment, kind of thoughts on sort of AI native startups, what that means, especially when you think about, you guys are very deterministic engineering grade solvers, and some of these are more general purpose models.
Well, Prometheus's spectacular scale is a reminder of how much value there can be generated from improving engineering outcomes of the sort I have been talking about that optimization makes possible. We do not know a lot about Prometheus, understandable given their early stage, but we do know the magnitude of what they have raised or what has been announced, and that suggests to me their ultimate intention is not to be in the engineering software business, which is a small business for that large of an investment, even though it will grow in magnitude.
But the advantages of combining deterministic approaches with inferencing for agentic automation and permitting this optimization, as I described, is a big opportunity. We make it available now, and we are glad to have attention called to that opportunity.
From the broader risk of I mean, you guys are in such critical infrastructure where you are not going to use a frontier lab for your customers, are not necessarily for design. But how do you think about what some of these AI natives could do? Is your space just so technically oriented, so regulatory driven that it is not really ripe for disruption from a Prometheus or somebody else like that?
Well, we would like to be a disruptor, but continuity, as I described earlier, is important. AI will help in so many ways. It is so wonderful for our production function and this opportunity for engineers especially, as they break out of the constraints of the existing established business model for transacting infrastructure engineering, which is that the engineering firms bill the owner-operators at time and materials for their work. They are charging for their inputs, not their outputs.
It is AI that will finally change that, and we showed last quarter that the engineering firms agree with that now, and everyone is going to benefit from that. Instead of engineering firms competing with one another on how well they manage their software costs, they will be side by side with us, eager to get more productive and to brand their own kind of optimization using various inferencing engines and so forth, but with our deterministic tools to assure the quality and the certifiability of those results.
That is the first wave, and there will be successive waves beyond that in ways that we cannot limit ourselves to predicting now for how AI is going to help. But I think that we have come to realize in the world that in vertical domains such as infrastructure engineering, the benefits of AI are primarily going to be incorporated and delivered by the existing vertical solution providers. If I go back to Bentley Systems' history, in our 44 years as the digital quartermasters, we have successfully incorporated innovations in computing.
We did not have to invent those. They come to us from horizontal suppliers who are satisfying consumer markets and so forth. We just have to innovatively apply them in turn to what we do, and we have been good at that, and we will continue to be good at that for the long term. It is a very exciting time to be doing that. We are a company of engineers for engineers, and AI is a tremendous opportunity to inflect, but continuity and equity in data and being able finally to reuse it with AI and to optimize, those are immediate opportunities, but not even the ultimate opportunities.
You have said that the company's basis of valuation has doubled every five years since the IPO, and we talked about a lot of exciting things, and we ran out of time to talk about some of the other things. But when you sit here today and you sort of project the opportunity for Bentley over the next five to 10 years, what excites you the most after being with the company for over 40 years? Is this the best time that Bentley is in front of it?
Well, when we think of that as a long-term trajectory, I do think we can say long term because I believe that, although we did not find it as easy to track, that our basis of valuation doubled every five years long before our IPO. The thing is, it does so steadily as well, given this flywheel of BSY busy on the part of the world's infrastructure engineers. So, looking forward, we have these waves of AI advantages that we will be able to source from many directions, these horizontal capabilities, and incorporate into our solutions.
The engineering firms and organizations would rather we do that than for them to try to piece it together on their own. So, that is a greater opportunity than ever. But something that's changing from that past pattern of continuing to innovate continuously is this one-time change to the business model of infrastructure and engineering from time billing to fixed price to start with, and then ultimately various value outcomes. We don't have to change our commercial model for consumption, because consumption is fair, and consumption will greatly increase with agentic API consumption.
But the difference between infrastructure engineering and product engineering and EDAs, mainly this business model. In these other areas, they spend a multiple of 3x - 5x more per engineer on software and AI. AI is software. So finally, we're breaking out of those constraints where there's equal enthusiasm on the part of the engineering organizations to spend more on software and AI enthusiastically to substitute for the labor, which is scarce.
Already, one out of 10 of the positions in infrastructure engineering firms can't be filled. It's not controversial now. Everyone is going to gain and share the gain together. So, as an investor, I believe we should expect, and you should expect us to continue to at least double our basis of valuation every five years.
Well, great. That's a lot of exciting things to look forward to. So, Greg, from all of us at RBC, and thanks for joining us as well. Best of luck closing out the year.
Cheers, all. Thank you.
Thanks.