Q1 2026 results. I'm Eric Boyer, Bentley's Investor Relations Officer. On the webcast today, we have Bentley Systems Executive Chair, Greg Bentley, Chief Executive Officer, Nicholas Cumins, and Chief Financial Officer, Werner Andre. This webcast includes forward-looking statements made as of May 7, 2026 regarding the future results of operations and financial position, business strategy, and plans and objectives for future operations of Bentley Systems, Incorporated. All such statements made in or contained during this webcast, other than statements of historical fact, are forward-looking statements. This webcast will be available for replay on Bentley Systems Investor Relations website at investors.bentley.com on May 7, 2026. After our presentation, we'll conclude with Q&A. With that, let me introduce the Executive Chair of Bentley Systems, Greg Bentley.
Thanks to each of you for your interest in BSY's 26 Q1. Nicholas will describe factors that contributed to commendable operating performance favorably according, as usual for BSY, with our annual full-year outlook. Werner will put this in the financial terms which continue to differentiate BSY as leading among peers, both in the quality and in the measures most meaningful to shareholders of sustained profitability and cash flows. I will supplement their on-the-ground reporting with perspectives behind our characteristically even higher prioritized endeavors to benefit our future value, in particular through advancements which make AI for us more a seminal opportunity than a terminal threat. Last quarter, I enumerated some respects in which Bentley Systems prospects are rather uniquely enhanced and accelerated by AI. Our advantages, as summarized here, make the case that leadership in Infrastructure AI is destined, given our experience and determination within Bentley Systems grasp.
Our long-established incumbency as the digital quartermasters for infrastructure engineering organizations' substantive tooling is a differentiating and immediate advantage. Quantification of that pole position will be my focus today. In particular, our modeling and simulation applications have established the de facto standards for responsible and deterministic infrastructure engineering. The stewardship for each account of their cumulative infrastructure engineering data in Bentley Infrastructure Cloud, ProjectWise, SYNCHRO, and AssetWise, positions them to leverage AI to compound value for themselves and for their own constituents at a steeper rate than ever. Beyond our existing consumption business models, AI is on the cusp of adding to our growth incrementally agentic API consumption of our modeling and simulation functionality, especially for optimization of designs and, for instance, to intrinsically improve constructability.
Our Asset Analytics initiative, spawned by AI and already exceeding $50 million in annual revenue run rate, is leading the way toward instant-on digital twins to optimize operations and maintenance, commercialized through subscriptions denominated in consumption per asset. Imbued with all these factors and shaping our distinctive AI game plan, our business is anchored by stalwart enterprise accounts. Within infrastructure engineering, these enterprises are on the leading edge of adopting, acting upon, and evolving individual proprietary AI initiatives, which we are there to support, prioritize, and enable. What will never change is that their business is our business, and their success is our success. Underscoring this affinity, 45% of our revenue comes from 220 accounts which each spend over $1 million per year with us.
Almost two-thirds of our revenue comes from 824 accounts which each spend over $250K per year with us, mostly, of course, through E365 consumption subscriptions, which include, in each case, a dedicated BSY technical success team positioned to nurture joint AI initiatives. To understand the extent to which our interests are aligned with rather than opposed to our accounts, let us drill down on project delivery firms, in particular because Engineering News-Record surveys and ranks them all, publishing individual revenue breakdowns that can help us understand their economics and our own penetration level and potential as digital quartermasters. From Engineering News-Record's two lists ranking respectively domestic and international headquartered design firms, last published together a year ago, we compile the composite ENR global top design firms ranked by their verified design billings.
For 2025, these 639 global top design firms generated $280 billion of design billings. 25% of these design billings were generated by 25 Chinese organizations. Unfortunately, because they're generally within state-owned entities, geopolitical tensions currently inhibit their accessibility for us. The top global design firms ex-China consist of 614 top firms generating design billings of $212 billion. Of these, 470 are BSY accounts, together accounting for $198 billion in design billings, 93% of the ex-China total. A considerable portion of those whom are not BSY accounts are rather pure architecture firms. With our ARR across these BSY accounts totaling $414 million, or about 28% of our overall ARR, top design firms are our largest constituency.
Per million dollars of their design billings, they spend on average about $2,000 in BSY ARR. On average, each uses about 4 among BSY's top brands, plus several other lesser brands. The top brand for these accounts, also now BSY's top brand, is ProjectWise, in use by 270 of the top firms, together generating $160 billion in design billings, representing 76% of the design billings of the ex-China top firms. In a coming quarter, we will describe joint AI initiatives with these firms to compound the reuse benefits from their Bentley Infrastructure Cloud data platform. For now, I would like to quantifiably illustrate the aligned incentives for BSY and top design firms to work together on the incipient AI transformation of their business model. The key is that infrastructure engineering software isn't for these firms overhead or administration.
It is a most necessary factor of production to enable, capture, and deliver their substantive work. Along with attendant labor and associated computing, it's a cost of revenue. To assess the economics, let's consider a representative project with $1 million of design billings. Triangulating variously, including from BSY's $2,000 average of spending per $1 million of design billings across the universe of all of our top firms' accounts projects globally, I estimate that a representative million-dollar design project would consume about $10,000 of design software. Since margins for these firms now reach about 10%, the cost other than for design software, mostly for engineers' labor, must then be $890,000. Now, what could a top firm gain by theoretically investing to the detriment of its design software vendors to somehow reduce its design software spending, say, by 20%?
For that putative inspiration, and effort, and distraction, and risk, and liability, an investment which thus couldn't afford to be much, their net profit margin would ostensibly grow from 10% to 10.2%. Back to the drawing board, consider that investing instead in AI automation and agentic API consumption and computing, even if that, say, doubled the all-in design software cost, could save at least 20% of scarce engineering time, as that is readily achievable with just agentic automation of drawings production. It seems logical to prefer reducing engineering labor because these top design firms compete with one another for a demographically shrinking talent pool of infrastructure engineers. With near-record backlogs, these resource constraints limit the new projects each firm can pursue, even though there's an abundance of infrastructure engineering work available.
Unfortunately, with the prevalent hourly billing commercial model for infrastructure engineering, improved productivity would produce more output with the same staffing, but not more design billings. One more simple change would be needed to make this fully worthwhile. Transform the design billings commercial model for such projects to fixed pricing. If I were an infrastructure owner-operator client, the only objection I can think of to fixed pricing would be a concern that as a result, corners might be cut, alternatives might not be pursued, and thus engineering quality could be compromised. With API consumption enabling demonstrable agentic optimization, that concern can be overcome technically. Each human engineer, now augmented by busy agents and automation, remains daily in the loop, but delivering 20% more design billings on additional projects in the same time, yielding very worthwhile AI-enhanced economics.
The top design firm enterprise, by investing in proprietary AI agents to automate and leverage and to API consume trusted modeling and simulation functionality, could now generate IP-level margins of over 24% on the same engineering inputs. As the owner client gets better assured and more timely quality of designs at no higher cost, everyone wins, including the fully employed and probably gainfully incented engineer, not to mention the software and computing providers. While respecting confidentiality of our enterprise accounts' individual AI plans and strategies, Nicholas will provide a brief update on our Infrastructure AI Program with accounts for such joint initiatives, among his other content. Over to you, Nicholas. Thank you.
Thank you, Greg. We began 2026 on a strong footing. Our Q1 performance demonstrates our ability to consistently execute against a backdrop of sustained global demand for better-performing and more resilient infrastructure. Before going further, my thoughts are with our colleagues and users affected by the conflict in the Middle East. I am incredibly grateful for our team in the region. Their commitment to our users has been unwavering, and their dedication is inspiring. Following up on Greg's remarks, I will start with an update on AI. We continue to execute on our AI initiatives across the portfolio, but I will focus on 2 recent highlights. First, on Bentley Asset Analytics. To take this business to the next level of scale, I am delighted to welcome Bryan Friehauf as our new General Manager. Bryan joins us from GE Vernova and was previously at Hitachi.
He brings a wealth of experience in scaling enterprise software businesses in the operation and maintenance phase of the infrastructure life cycle. Second, on Bentley Open Applications. We have engaged with leading infrastructure organizations, both engineering services firms and owner-operators, as part of the Infrastructure AI initiative announced at the end of 2025. The feedback has been clear and consistent. There is strong demand for us to instrument our applications, so they can power users' own AI-driven workflows. Based on this strong feedback, we have prioritized the development of both APIs and MCP servers. In fact, we just released an MCP server for STAAD, our flagship product for structural analysis. To give you a sense of the potential, this allows an AI agent like Claude to interact directly with STAAD to optimize the structural design at machine speed.
The ability to iterate on complex design trade-offs so quickly is transformative. Our next steps are to instrument other key applications and to validate the commercial model for this powerful new usage pattern. We look forward to sharing our progress. Turning to our business highlights. Our year-over-year ARR growth for Q1 was 11.5%, in line with our expectations. Our net revenue retention rate remains strong at 109%, consistent with previous quarters and highlighting the stability and growth within our existing accounts. Our Enterprise 365 commercial program continues to drive steady growth, both in terms of conversions as well as flow uplift at renewals, noting that Q1 is our smallest quarter for renewals. New logos contributed again 300 basis points of ARR growth, primarily within the SMB segment.
Through Virtuoso, our flagship commercial program for SMB accounts, we again added over 600 new logos in Q1. At the same time, our growth model is evolving with an increasing contribution from cross-selling and upselling to existing Virtuoso accounts. While our renewal rate remains high, the sheer scale of the Virtuoso base creates a natural churn dollar amount to overcome each period. Our combination of new logos and existing account expansion allows us to continue to deliver strong net growth. Turning to our performance by infrastructure sector, resources was the fastest-growing sector in total and across each geographic region. We expect another strong year for Seequent, bolstered by improving mining fundamentals. I will come back to how we plan to expand our addressable market even further into critical resources in a few minutes. Our largest sector, public works utilities, delivered a solid quarter, driven by robust global infrastructure investments.
PowerLine Systems continues to benefit from increasing demand for grid resiliency and new power generation, as well as international expansion. The adoption of Seequent applications for the civil infrastructure also supported growth in that sector. Growth in the industrial sector continued to be solid, while commercial facilities remained relatively flat. Turning to our tone of business by geographic region. In the Americas, our largest region, the U.S. delivered solid growth, supported by stable public funding at both the federal and state levels for transportation, grid, and water projects. Private investment was also robust, particularly in resources and AI-related data centers and power generation. Latin America delivered a standout quarter with strong performance from Seequent and Mining and from our increased focus on transportation in the region. EMEA was our fastest-growing region in the quarter.
This strong performance was achieved despite the conflict in the Middle East, which saw some project delays and a shift in consumption to other regions. This was more than offset by strengths elsewhere. In the U.K., growth accelerated as major projects moved into the delivery phase, just as we anticipated last quarter. We also saw robust growth in Africa, driven by increased spending in mining. Asia-Pacific delivered solid growth, with India once again leading the way and Australia showing improved momentum. While we continue to navigate persistent headwinds in China, which represents approximately 2% of ARR, the strength across the rest of the region more than compensates. I would like to take a deeper dive into our resources sector. This is a part of our business that has become increasingly significant, and we believe it's important for you to appreciate its journey and its forward-looking potential.
When we acquired Seequent almost 5 years ago, our primary objective was strategic: to integrate their best-in-class software for understanding the subsurface into the world of infrastructure. We knew that a misunderstanding of ground conditions is a primary cause of delays and risk in major infrastructure projects. As a byproduct of that strategic move, we also acquired a sizable and thriving business in the resources sector. I am pleased to report that the initial strategy has proven effective. Since the acquisition of Seequent, we have grown our subsurface ARR in civil infrastructure by a factor of 4, in part due to successful cross-selling into the existing Bentley accounts. The potential for further growth is significant as engineering services firms adopt a ground-informed design approach, bringing detailed ground investigation in as a foundational step before design begins, much like they already do for above ground surveying.
At the same time, Seequent has continued its impressive growth in its core mining market. Seequent's growth rate in 2025 accelerated as the geopolitical climate and race to AI increased the focus on critical minerals. We expect these trends to continue in 2026, contributing to another standout year for Seequent. It is important to note that Seequent has delivered strong growth even during the mining exploration slowdown starting early 2023, when production mining companies used our solutions to mine existing deposits more efficiently. The potential of Seequent in resources extends beyond traditional mining. Seequent's technology is pivotal for other critical resources that are essential to the global economy and to society. Take, for instance, new sources of energy. Seequent software is already instrumental in the operations of more than 60% of the world's high-temperature geothermal electricity generation.
It's being applied to new enhanced geothermal systems by companies such as Fervo Energy, a winner at our 2025 Year in Infrastructure Awards. Their Cape Station project in Utah, for example, demonstrated how new drilling techniques and digital technologies are making geothermal power increasingly accessible and economically viable. Clean, renewable, and consistent baseload energy is more critical now than ever as AI and data centers demand more power. Seequent's impact extends to the most vital resource of all, water. Groundwater supplies about 50% of global domestic water and over 40% of irrigation water. Most of these resources are under stress due to overuse. Seequent software is used around the world by engineering consultancies to manage these resources, mapping aquifers from California to India, designing managed aquifer recharge facilities, and constructing a digital framework for groundwater management models.
Nearly 5 years since the acquisition of Seequent, resources have become our second-largest sector, accounting for more than 20% of our sector attributable ARR, and it continues to be our fastest-growing sector. In summary, it was a strong start to the year. We are executing well in a robust market, and we're excited about expanding our reach further within the resources sector. Before heading off to Werner, a quick update on our event strategy. We are decoupling our YI awards from our user conference to create two distinct world-class events. Our Year in Infrastructure event will now be exclusively focused on our global awards competition. We are keeping its intimate and celebratory format. This year it will be held in Singapore from October 6th to 7th. Separately, we're launching a new large-scale user conference dedicated to product learning, best practices, and community networking.
The very first will take place in Toronto in April of 2027. We believe this new format will allow both events to thrive. Now, for a detailed review of our financial results, over to you, Werner.
Thank you, Nicolas. We are pleased with our strong start to the year, which continues the momentum we built through 2025 and positions us well within our financial outlook for 2026. Total revenues for the first quarter were $424 million, growing 14.5% year-over-year and 11.9% in constant currency. Our growth continues to be driven by our mainstay subscription revenues, which represent 93% of total revenues and grew 14.7% for the quarter or 12.2% in constant currency. Our E365 and SMB initiatives continue to be solid contributors. In our smaller and less predictable revenue streams, perpetual license revenues decreased 18% in constant currency.
Perpetual license sales remain a very small part of our business and are, as always, the less controllable and less predictable component of our revenue mix. Service revenues increased 25.8% in constant currency, driven by long-awaited re-acceleration in Maximo-related services from our Cohesive business. Last 12 months recurring revenues now stand at $1.44 billion, increased by 13.3% year-over-year or 11.5% in constant currency and represent 93% of total revenues. Our last 12 months constant currency account retention rate remained consistent at 99%. Our constant currency net revenue retention rate remained at 109%, consistent with recent quarters. The combination of our high retention rates and new business momentum gives us confidence in the continued durability of our recurring revenue growth. Now turning to ARR.
We ended the first quarter with ARR of $1.495 billion at quarter-end spot rates. On a constant currency basis, our year-over-year ARR growth rate was 11.5%. Our sequential quarterly growth rate was 2.5%, all organic and in line with our expectations for the quarter. We continue to expect our quarter-over-quarter ARR growth seasonality to be similar to 2025 and thus organic year-over-year ARR growth rates to be relatively stable during the year. Our GAAP operating income was $126 million for the first quarter. As I've discussed previously, our GAAP results can be impacted by deferred compensation plan revaluations and acquisition-related expenses. Moving to our primary profitability measure, adjusted operating income less operating stock-based compensation or AOI less operating SBC.
As we announced with our 2026 outlook, this is the first quarter we are applying this refined metric. This change aligns the treatment of cash-settled and equity-settled acquisition-related stay bonuses, removing M&A-related volatility from this key operational metric. AOI less operating SBC was $141 million for the quarter with a margin of 33.2%. This quarterly margin performance was in line with our expectations, positioning us well to deliver on our annual margin improvement. As a reminder, we plan to invest earlier this year, resulting in operating expenses being more weighted towards the first half compared to 2025. Our free cash flow for the quarter was $188 million. This result was in line with our expectations and reflects two key factors we signaled on our last earnings call.
First, our 2025 free cash flow benefited from exceptionally strong collections at year-end. This created a timing benefit in 2025, which, as anticipated, resulted in a tougher year-over-year comparison for the first quarter. Second, the plan for operating expenses to be relatively more weighted towards the first half this year is reflected in comparison to 2025 for both profitability and cash flows. Looking beyond quarterly timing fluctuations, on a last 12 months basis, free cash flow of $492 million was up 13%, and we remain on track to meet our full year free cash flow outlook of $500 million-$570 million. We continue to execute a disciplined and balanced approach to capital allocation.
During the quarter, we repaid at maturity the outstanding balance of $678 million of our 2026 convertible notes, utilizing borrowings under our credit facility and cash on hand. The retirement reduced our fully diluted share count by approximately 10.66 million or 3%. In total, our net debt decreased by $134 million in the quarter. We also returned capital to shareholders, deploying $54 million for share repurchases and $21 million for dividends. Our balance sheet provides significant strategic flexibility. At quarter end, capacity under our credit facility was $756 million, and we reduced our net debt leverage during the quarter from 2.1 to 1.9 times adjusted EBITDA.
Subsequent to quarter end, we closed on a new $550 million term loan A under the accordion feature of our credit facility. This transaction was completed at attractive terms and used to repay outstanding borrowings under our revolver and lowering our interest cost. With the term loan in place, total borrowing capacity under our credit facility increased to $1,850 million. This provides ample capacity to support our strategic priorities, positioning us well ahead of the 2027 notes maturity, while also funding potential programmatic acquisitions, ongoing share repurchases, and dividends. We continue to actively manage our interest rate exposure. Our safeguards include the low fixed coupon on our remaining convertible notes and our $200 million interest rate swap expiring in 2030.
Finally, we remain comfortable with our 2026 financial outlook range that we provided just over two months ago on our Q4 call. With regards to foreign exchange rates, for the first quarter, the U.S. dollar has strengthened slightly relative to the exchange rates assumed in our 2026 annual financial outlook, resulting in approximately $2 million less revenues from currency changes. If end of April exchange rates would prevail throughout the remainder of the year, our Q2 to Q4 GAAP revenues would be negatively impacted by approximately $3 million, relative to the exchange rates assumed in our 2026 financial outlook. With that, over to Eric Boyer for Q&A. Thank you.
Thanks, Werner. Before we begin, just as a reminder, please limit yourselves to one question so we can get to everybody today. First question comes from Joe Vruwink from Robert W. Baird.
Great. Thanks for taking my questions. You know, I guess, Greg, the example you shared is interesting. I think the prospect of supporting $200,000 more in revenue by spending $10,000 on software is something all of your customers would gladly entertain. What do you think about in terms of Bentley's product efforts in terms of bringing that proposition closer to reality? What still needs to be done, and what sort of deliverables do you need to demonstrate to your customers so that they believe and give you the buy-in?
Well, I think the path to that foreseeably is the API consumption. The notion that agents spun up by the engineers can do more iterations in the same time. It not only will deliver a better quality of design, but the engineering firms will be able to substantiate that to the owner-operators to accelerate this transformation to fixed pricing. Well, what's really great about that prospect for us is the more that engineering firms and owner-operators are excited about and ready to act on this transformation to a new commercial model, that's all good for our prospects. It's where everyone wins, as I say. It can't be doubted that this will happen with the AI inflection as it is happening in every other service industry. Engineering is one of those.
It's been slow until now, and this notion of optimization that literally is in front of us at the moment is gonna speed it up, I think.
Thank you.
Our next question comes from Matt Hedberg from RBC.
Hey, great. Thanks for taking my questions, guys. Good start to the year. You know, when we look at your business, you know, you guys all highlighted a number of, I think, really interesting company specific catalyst that seem to be moving in your favor this year. You know, to us, when we sit back and look and, you know, it seems to present an opportunity that could push constant currency ARR towards that higher end of the range this year and perhaps accelerate versus last year. I guess, you know, when, you know, when I sit back and think about all these opportunities, if you were to highlight the one or two things that you're like, "These are the most important things that could deliver those results," you know, what would they be? 'Cause it seems like there's a lot of opportunity here for you guys.
Well, I think you're characterizing it exactly the right way, Matt. There's a lot going on, and that is very positive for the company. We benefit from very strong tailwinds here in both infrastructure and in resources, as I highlighted during the prepared remarks. For us to get to the, let's say, higher part of the range, we would need both resources and mining in particular to continue to go strong throughout the year. Right now everything is signaling that this is trending well. We need Bentley Asset Analytics to continue to grow strongly throughout the year. We need, of course, you know, the core business infrastructure to continue to go well. Then, you know, probably, you know, another programmatic acquisition. Yeah. If you have all of that lining up, then yeah, we're talking about the higher part of the range.
Thanks a lot, guys. Best of luck.
Next question comes from Jason Celino from KeyBanc.
Great. You know, you know, this actually goes back, you know, to Joe's question. You know, there's growing debate among the investment community on whether to charge or not charge for access to data. It seems like Bentley is 1 of the only software platforms trying to monetize this way through API consumption. Frankly might be the preferred way investors may wanna see it. Do you foresee any risks from a customer perspective on charging this way since other, you know, horizontal software companies have decided to be more open and not necessarily charge for it? You know, Nicolas, the stat MCP server that you talked about, are there other MCP servers that you might need to stand up? Thank you.
Definitely. Right. Maybe a point of clarification first. When we're talking about API consumption, we're talking about a indirect usage to our engineering applications. As opposed to having a user directly interacting with the applications, we have an AI agent that is directly interacting with the application. There's still a user behind the AI agent, but it's the AI agent which is now being able to do with the application, you know, a level of optimization which was simply not possible for an engineer on his own. Okay. We have Bentley Infrastructure Cloud, where the data actually resides. Yeah. Here, indeed, we're not monetizing the access to that data because that data belongs to the infrastructure organization that are entrusting us with that data as part of the platform, right? Those are two different things.
The ability, you know, to leverage AI to optimize designs or even generate, you know, parts of the design is just a fantastic value proposition. Clearly our user base pricing right now, whether we're talking about attended consumption as part of E365 or annual subscription with Virtuoso, is not gonna capture the full value that is gonna be created, right? We're discussing with the representative accounts who are involved in our co-innovation initiative called Infrastructure AI. We're talking about a different commercial approach where it would be indeed a API consumption based pricing, maybe token-based, because that seems to be, you know, the common denominator now across different AI tools. Yeah. That's really for the engineering applications. For Bentley Infrastructure Cloud, at least in the foreseeable future, the pricing is indeed user-based. Either user-based consumption or attended consumption or annual subscriptions.
May I make clear that today we only monetize attended consumption. Our strategy is to introduce and increase the API consumption and give users and accounts a chance to explore the potential of that and learn in the process what it costs us, what it could cost them, and what the benefits and values are so that we can arrive at an appropriate monetization approach to that which we are open-minded about for now.
Thank you.
Thanks, Jason. Next question comes from Siti Panigrahi from Mizuho.
Great. Thank you. Just following up that AI part, Greg, you laid out really a compelling case for AI within your customer base. Wondering what sort of feedback have you got from your customer and how should we think about the TAM expansion for Bentley or even, excuse me, the ARPU uplift potential as AI drives even deeper multi-product adoption? Maybe it's not near term, but how should we think about ARPU and TAM expansion? In the same context, Werner, how are you looking about the, you know, margin and even there is incremental engineering and computing cost implication as in Bentley start focusing on building AI products. How should we think about the investment intensity over the next 12 to 18 months?
I'm gonna ask Nicholas to speak about the account reaction. I may just say a particular multi-product scenario that has me excited is during design to be able to have an agent using SYNCHRO to explore the constructability of what's being designed while it's being designed. This doesn't imply that the design firms that we talked about will be doing the construction necessarily. Someone will be doing the construction subsequent to their involvement. They'll be able to say to the owner operator, "We're going to be able to give you a design where we already know the economics of construction. We've simulated that in the course of our design." It's just an example of something not even conceivable now that AI will introduce through API consumption. Nicholas, perhaps you can speak to the reaction of accounts to the prospects for this instrumentation.
I'll do that, and then also get to the TAM question. The reaction of our accounts is really validating the opportunity for this indirect usage of our applications. Now, we do see a difference here between the very large infrastructure organization, in particular, the very large engineering services firms and all the others. Yeah. The very large ones-- By the way, exactly like it was shared with the AEC Advisors CEO survey of 2025. The very large ones are the ones who are really investing in their own AI-driven workflows. You know, they are the ones who are exploring with us how exactly they could start to use our applications indirectly through AI, right?
All of the infrastructure organizations we talk to, all of them, whether they're big or small, are welcoming that we deliver our own AI capabilities as part of our products, right? Back to the very large one. These are still very early days, yeah. As we commented actually last quarter, we're much more in the mode of exploring and validating and then confirming, for example, that yes, an MCP server for STAAD makes a lot of sense. Yeah. An MCP server for another application. We're very focused on adoption as well. The monetization will come next, right? Where we are doing monetization right now with AI is really through a different offering, which is Bentley Asset Analytics, which we can talk again in a moment.
Now back to your question about the TAM. I think it's a great question because the TAM that we shared, at the moment of, our IPO when we became public, was all based on the number of engineers. Effectively, if we're talking about indirect usage of our applications through APIs, we are somewhat removing that natural cap of how many engineers are out there. Yeah. How much can they consume within a given day, in terms of applications for Bentley. We're very excited about that as a, as a long, let's say, longer term growth opportunity for us to actually expand our TAM by having indirect usage of application through AI.
Werner?
Yeah, maybe on the margin. I think as Nicolas said, it's early days. It's completely immaterial as of now. I could see that gross margins will be impacted, but it's likely, and this has to be considered then in the monetization of the product.
Thank you. Thanks. The next question comes from Dylan Becker from William Blair.
Hey, everybody, appreciate it. maybe Greg or Nicholas here. I think the shift to outcomes obviously makes sense and kind of your ability to facilitate that efficiency gain, I think is abundantly clear. I was maybe wondering on the importance of kind of your services org and the direct customer relationships you have from a go-to-market perspective in how you can kind of help those customers bridge that gap from a change management perspective and kind of help maybe accelerate that push in the economic delivery model, if that makes sense.
One thing that I'll remark upon is that our dedicated teams in E365 over the past year or 18 months have discovered a new productive way to help our accounts. It's by helping them pursue new business. It's helping join their pursuit teams to bring new ideas to owner-operators, and I hope that will include as we implement the optimization approaches, how that gets sold and communicated, among things, to enable fixed price, which is the beginning of outcome-based contracting.
Yeah, at the end of the day, when it comes to the commercial model of the engineering services firms we serve, this is very much to be decided between them and their clients, you know. Depending on the industry, it will vary quite a lot, you know. Some of them have already embraced fixed-based pricing, others are still in time and material. What we can do, besides, of course, you know, demonstrating the power of AI and how it changes the whole value proposition. Besides that, we have a lot of high-level advocacy efforts that we are doing with governments, with the financial sector as well, with the clients.
Some of them are through our infrastructure policy advancement team. Some of it is also through our services arm called Cohesive, which is, you know, providing consulting to owner-operators, and, you know, giving insights about what else could be done. You know? What's the art of the possible and potentially what does it mean for their commercial model?
Perfect. Thank you.
When Nicolas describes some have begun fixed pricing, it's primarily those that do private sector work, where private sector owner-operators are quicker to adapt and evolve than government-funded public works and utilities.
Thanks. The next question comes from Jay Vleeschhouwer from Griffin Securities.
Thank you. Good morning. Nicolas, a quarter ago on the call, you described Bentley Infrastructure Cloud as your data foundation for AI, and I assume that's more broadly so aside from AI, that in terms of the importance of Infrastructure Cloud. With that in mind, could you talk about perhaps some of the adoption metrics for Infrastructure Cloud? Do you think of it as a kind of prerequisite for or leading indicator for other business, AI or not? Then if I may, a point of clarification, given all the API discussion earlier. In a complex federated system of that kind, how do you assure, when customers adopted good or better throughput or performance in a complex API system?
To the first question. When we're referring to Bentley Infrastructure Cloud as a data foundation for AI, we're really talking about the AI efforts of the infrastructure organizations we serve. When they upload files, when they connect data systems into Bentley Infrastructure Cloud, and all of the data from those files or the systems is mapped to a schema, that data becomes suddenly queryable, it can be analyzed, but it can also be leveraged by AI. Right? Now, we made it very clear that we are not using that data, which is in Bentley Infrastructure Cloud, to train our own AI. That definitely sets us apart from other software providers out there. Now, data ownership is a very, very sensitive topic across all industries, for sure in infrastructure as well. You know?
We do not think it's right to be leveraging the intellectual property of some companies to train AI that will benefit other companies. Right? We just don't think it's right, we're not gonna do it. The only time we use data within Bentley Infrastructure Cloud to train AI is when we're explicitly asked by infrastructure organization to do so, and then it really is their data. By the way, we give full transparency of what data has been used in order to train AI. That, you know, sets us apart to the extent that it really becomes a reason why infrastructure organizations also choose Bentley Infrastructure Cloud as opposed to other alternatives, because they really have trust that we're gonna be good custodians of that data on their behalf.
When it comes to API throughput, indeed a point of clarification. For example, with the demo I was showing with STAAD, we are in a sense, in the first step of the instrumentation where these applications are still running on the desktop. In fact, you could see it with the video. It was still running on the desktop. They're not yet platform services. That will be the next step. The first thing we're doing is making sure that the APIs exist, that there are MCP servers available, so it's easy for AI agents who interact with those applications. All of that interaction is actually happening on the local computing station, on the personal computer. Yeah. At some point, of course, you know, that's what we're discussing also with accounts.
You know, we're talking about the longer term here. You know, some of those workloads will be moved to the cloud to become true, you know, cloud services. We'll just make sure they are as performant as they should be, like we do for all of the cloud services that we offer.
Thanks, Jay. The next question comes from Alexei Gogolev from JPMorgan.
Hello, everyone. Great to see you all. You've mentioned that it is still early days for applying AI to mission-critical engineering, and you are leading the exploration to building and drive the adoption of highest value AI-powered workflows. Could you maybe speak about some of the initiatives on that end? What efforts you're taking?
We have, you know, if you're referring also to one of the earlier comments around, we're making progress, you know, with our initiatives across the portfolio. We have both efforts to deliver our own AI capabilities as part of our product, and then efforts to instrument our applications, in particular the engineering applications, so that they can support AI-driven workflows that are being created by the infrastructure organizations that we serve, right? Those are two. For the latter one, there is a co-innovation initiative that we've launched at our annual conference last year, Year in Infrastructure, so towards the end of 2025, where we invited infrastructure organization to join, talk to us, and explore with us what are the use cases they're going they would like us to be able to support in our applications.
It was no surprise that the actually first use case as identified was for STAAD, because we had already seen STAAD, which is our flagship product for structural analysis, being used this way, because it had an API in some of the submissions for the Going Digital Awards at our annual conference over multiple years, by the way. It was clearly the first one. No surprise that there was a very strong demand for us to create an MCP server based on that API in order to support the interaction of AI agents with STAAD. Yeah. That's the way we're interacting with them in that context, which is understanding what are the use cases with the biggest potential, validating with them, and then creating those MCP servers.
We have multiple efforts across our pretty wide portfolio of engineering applications to create the MCP servers directly if the APIs already exist. If not, start to create the APIs themselves. Okay. Now for the AI capabilities that we offer, we have AI capabilities as part of the engineering applications that will automate, you know, parts of the design workflow. You know, typically we go for the, let's say, mundane task, which are extremely time-consuming, like drawings production that we talked about in multiple quarters ago. Yeah. We have AI capabilities built within Bentley Infrastructure Cloud, for example, in order to facilitate the search of engineering data using natural language. We have AI capabilities being part of Bentley Infrastructure Cloud in order to help, you know, navigate very complex construction models as part of SYNCHRO.
We obviously have AI pretty much for all of our applications for Bentley Asset Analytics, whether we've developed them or we acquire them. We even have AI in Seequent, which I spent time highlighting today as part of my prepared remarks, right? There are efforts related to AI really across the portfolio. It's a wide range of initiatives.
Thank you. Thanks, Nicholas.
Thanks. The next question comes from Daniel Jester from BMO.
Hey, good morning. This is Will Hancock on for Dan Jester. Thanks so much for taking our question. You hit on data ownership a few minutes ago, and I just wanted to double-click there. Are you guys seeing any shifts in customer behavior, you know, whether that be increased willingness to use data to train AI models? Or conversely, more cautions around permissions? Are you guys seeing any variance across end market or customer size? Thank you.
What hasn't changed is the fact that it's let's say the largest infrastructure organization, especially the largest engineering services firms, that are the most advanced in their own AI efforts, and therefore looking into their own data on how, you know, they can train their own AI models. That hasn't changed. What we have seen in the last few months is a very, you know, suddenly, infrastructure organization realizing that not all software providers have a principled approach to data. Yeah. Then, you know, therefore pushing for a lot of clarification on terms and conditions and access rights, et cetera, for the data.
That's what really sets us apart, which is already back in 2023, we took this very principled approach. You know, our commitment to data stewardship that made it very clear that our users' data is their data always. We don't use it to train our AI unless explicitly directed to do so by the infrastructure organizations we serve. It is indeed a very sensitive topic, and I expect it, as basically infrastructure organization become more and more educated about, on one end, the potential of AI, and second, the fact that there is a difference here across software providers. I do expect this topic to become even more sensitive going forward.
I'll add that while the immediate attention is paid to training AI models, I think the ultimate value of this data will be its reuse in future designs. That's never been the norm because in attended consumption, an engineer will always prefer to start with a blank screen and originate a new approach. Each firm has a very valuable archive of project data in ProjectWise Bentley Infrastructure Cloud, AI will be good at finding and suggesting and modularizing and parameterizing. Ultimately, when that reuse can be informed by the operating and maintenance performance of those designs, which the engineering firms will increasingly be in the business of improving and optimizing, that will be a virtuous cycle that will yet reinforce the valuable proprietary advantage.
When I say that engineering firms can earn IP returns, I mean returns on that valuable data and knowledge, their accumulated compounding advantage with Bentley Infrastructure Cloud.
Maybe I'll put one final point, because it is such an important topic. Because some of you may ask, "But wait a minute, you know, are the other software providers who are shamelessly leveraging the data that is stored in their platform to train their AI, do they have an advantage?" I will clearly say no, and for two reasons. Number one is because of their stance, infrastructure organizations have a higher tendency of choosing us, yeah, and our platform, and because they can trust us. Second is for all of the AI capabilities that we're developing that I mentioned before as part of our initiatives across the portfolio, right? We don't do it in isolation.
We do it always with representative accounts, and we've never had an issue of not having enough data to train our AI models, either by using our own synthetic data or getting data from those representative accounts involved that they deem they think is not, you know, sensitive, not differentiating, and they're happy to contribute, right. We clearly see this as a net positive, for, and a net advantage for us.
Great. Thank you.
Yeah, thanks. The next question comes from Taylor McGinnis from UBS.
Yeah. Hi, can you guys hear me?
Yes.
Yep.
Okay, perfect. Thanks so much for the question. If I look at your guys' constant currency net new ARR growth historically, I think it's been around $26 million-$27 million. In this quarter, it was $37 million, up 36% and really strong. When you just think about the drivers of what drove that strength in the quarter, could you unpack a little bit of that? Was there anything one time or different seasonally this year compared to what you've seen in the past that might explain some of that? How do we think about that in the context of what you're thinking about demand trends going forward?
Maybe I take that. Q1 saw a continued strong momentum as we exited 2025. The quarter-over-quarter growth puts us somewhat ahead of what we saw in Q1 2025. It was, I would say, predominantly led by the continued momentum of our Seequent business and mining just doing still very well. As we said, that puts us in a pretty good position for our full year outlook range. It's still Q1 is 20% of our ARR opportunity based on contract resets and it's gonna see that the rest of the year plays out well as well.
The final question comes from Joshua Tilton from Wolfe Research.
Thank you guys for sneaking me in under the bell. I'll give you guys a break from the AI questions. Maybe just 2 points of clarification on my end. First, you talked about Seequent being really strong last year and expectations for it to be strong again this year. Do you expect it to accelerate again in 2026? My second part to that question is, on the other side of that, you kinda cautioned us, or maybe I'm reading into it a little bit, but you cautioned us around Virtuoso getting big, and a churn dollar amount that you have to overcome each quarter. Is there anything we should, you know, think about or be paying attention to or anything unusual about that churn that you're trying to signal to us, this quarter?
Okay. I'll try to be brief since we are at the end of the scheduled time. On Seequent, no, we don't expect further acceleration in a sense that as part of our plans for 2026, we assume the same level, of course, that we've seen towards the end of 2025, right? Clearly, in 2025, there was an acceleration, and for 2026, we just assume this is gonna continue, and Q1 basically proves us correct. Yeah. On Virtuoso now, we just wanted to explain that in addition to solid growth coming from new logos, we now also have growth coming from existing accounts. In previous earnings calls, we were only talking about new logos. Now, clearly, we've developed a new muscle here, which by the way, also helps for retention.
There's a clear correlation, which is if we see accounts using more than one product, then they will have a higher propensity of staying with us. Yeah. The overall retention rate remains stable at a high double-digit, very similar to what we've shared in previous quarters. It's just mathematics. It's a much higher base, of course. Indeed there is a churn amount in terms of U.S. dollars that needs to be compensated for. That's it. Yeah. It was just explaining all the different puts and takes and explains why even though we have this new muscle, you can see the growth being still very consistent with Virtuoso.
Maybe I'll just say in closing your first question about acceleration, there hasn't been questions particularly about geopolitical events in the world. The notion that each country needs to be more self-sufficient in its defense and its resources and so forth, represents a commitment to investment and infrastructure. It's not limited to resources. You need the resources for infrastructure. You need the infrastructure for resources, and I expect that to be not a short-term phenomenon to our benefit. Thank you.
Thanks. That concludes our call today. We thank you for your interest and time in Bentley Systems. We'll talk to you again next quarter.
Thank you.