Modelon AB (publ) ($MODEL)

Hi, and welcome to this Live Q following Modelon and their Q1 report. And with us today to present the report, we have the CEO, Jan Haglund; and the CFO, Jonas Eborn; and also the Chief Strategy Officer, Johan Andreasson. And here we will be giving you a demonstration about the new AI assistant. But before I hand it over to the company, I would like to remind you to ask questions throughout the presentation. [Operator Instructions] And without further ado, I would hand over to the company.

Thank you very much Jessica, and thank you everyone, for listening to this report of Modelon's first quarter 2026. And as Jessica mentioned, I will be joined today by colleagues, 2 colleagues, Jonas Eborn, and Johan Andreasson. Johan is our Chief Strategy Officer and also very much our driver for our AI strategy, where we have made significant progress during the last 2 quarters. And I thought it was time that we show something concrete and live in today's call. But before that, let me summarize the quarter. The first quarter of 2026 came in as follows: net revenue came in at SEK 15.3 million, which is 25% down from last year. A lot of that is expected since it's due to currency effects. This is not currency adjusted. A lot is also due to lower services revenues, which are part of our strategy to focus on product sales. But some of it is also related to lower software revenue. And as you see, the annual recurring revenue, which came in at SEK 47.2 million is down 5% quarter-over-quarter. This continues a negative trend that we've had now for some time on decreasing ARR. This quarter, it's the effect of both what we saw in the fourth quarter when we lost some U.S. federal accounts due to budget changes on the federal side in North America. This quarter, we have, in addition to that, also lost a number of smaller accounts, primarily seat accounts working with our legacy products. So you can see this as an indirect effect of the transformation that Modelon is ongoing, a transformation towards product sales, but also a transformation towards growing markets in thermal fluids and in particular, in data center cooling, which is a topic that I will come back to in this call. Then operating expenses in the first quarter came in at SEK 23.2 million, significantly lower than the corresponding quarter last year, 22% lower, which then contributed to an improvement in adjusted EBIT by SEK 1.1 million to minus SEK 7.2 million. We continue to win business and new customers, both in North America, in Europe and in Asia. In North America, we can note Babcock Power. It's a large U.S. provider of energy solutions that have been using Modelon Impact for some time and that added software licenses for their simulation of energy systems. We also are happy to see a new European customer, the European Slovenian consultancy company, Resalta, that work in the energy space. They will work with Modelon Impact in order to sell their services to do simulation and to offer value to their customers. And we've had a very interesting quarter in Japan and Northeast Asia, where we welcome 2 new important major customers. They're still small in terms of business volume for us, but they are largest customers, and we see large potential in them. The first one is Kansai Electric Power Company that provide energy for the Southern and Western parts of Japan. They have selected us both for software licenses and services. And then we have a major global Japanese provider of home appliances, a brand that all of us know and use that contracted software licenses and expert services for system simulations of some of their new product models. We're really happy to see this growth of new customers coming in. Now from the result and the decrease in annual recurring revenue, it's also obvious then that the new revenue did not compensate for the loss of software revenue in the first quarter. As I mentioned, we did lose a number of small accounts with legacy products, and that added to some of the lost ARR that we saw during last year from the United States. So this calls, of course, for action. And we are extremely focused on this to turn around ARR. And I remain optimistic that we have a very good chance of seeing growth during the year. As I mentioned the last quarter, we have since Q1 2026, a new sales and customer success organization. And what we now see is that organization, thanks to new leadership and thanks to new people are generating a growing pipeline. So that's very promising to see. What we also see now is that we are getting out with differentiating features that are catching a lot of customer interest, in particular, around artificial intelligence, where I feel that we have -- we're very well positioned, and we have a technology base that I will come back to that allows us to go very fast in this space. So I see that new AI features that we will mention later on in this call will give us an opportunity both to upsell towards existing accounts and to differentiate in competition with other larger companies to win new deals and new market share. And the third thing that makes me optimistic, especially for the second half of this year is what we do in our library space. We're adding new libraries, which is one of the core expertise and has been for many years of Modelon. And in particular, we're adding now library content in the data center space, which allows us then to address the large and very fast-growing data center market and in particular, for data center cooling. So all in all, this is our plan. It's a summary of our plan to turn around ARR during the year. But we are, of course, disappointed of the negative development during the last couple of quarters. So talking about data center then. Modelon has expertise, and we're developing products in AI data center cooling. We've done that for some time. And in fact, many of the products that we have had are a very nice fit to the new technology that is being used to cool AI data centers. Now this comes in very handy because data centers that used to be cooled only with air, basically fans, they now need to have new systems and new technology, in particular, based on liquid or the combination of gas and liquid, which often is referred to as 2-phase cooling. And that's something that we know very well. It's quite complicated physics, but we have not only expertise, but we actually have products and models that are available off the shelf to model these kinds of new systems. And what customers are telling us now is that they're using system simulation in order to create stable systems. A data center system consists of many different components. And you may think that you could dimension them one by one. But more and more, we see that it's necessary to look at the whole system in order to make the whole data center cooling work as it should. In fact, you could end up in unstable situations where the system becomes unpredictable if you don't do this properly. That's how system simulation from Modelon fits in, and that's, in fact, what these major companies are using us for. And then once you've built that simulation model, you can use it in the operations phase. You can use it to calibrate your live operations. You can use it to optimize parameters, for example, to reduce cost. reduce energy cost or reduce resources, for example, the use of water, which is critical in many geographies. So all in all, we see a growing market here. We have a lot of ongoing discussions, leads, and I remain optimistic about our development in this space. I mentioned AI. And during the last quarter, we have had a lot of development and actually concrete deliveries in the AI space. You can read more about this on our web page where there are several blogs and other information. But if I just summarize briefly here, what we actually released as late as this week is a new AI assistant integrated into Modelon Impact, which is based on state-of-the-art generative AI. It's not any particular AI. Well, we've selected one, but we're not tied to any particular AI service. We can, in fact, choose whatever is the leading AI services that are available right now. What makes it special is that this is not just a general chatbot, but it's actually trained on Modelon's libraries and simulation platform. That means that a lot of knowledge that was before either in the head of our experts or customers' experts or perhaps in manuals or online tools now is readily available, integrated into the system where you in [ clear text ] both can ask and get answers, not only about general questions, but in fact, where you are right now in the tool to get tips, guidance and troubleshooting directions. And we will look at that in a demo in a while. So I see a lot of value here in supporting users onboarding and troubleshooting. And I've been out talking myself to customers who have really confirmed that they see a lot of value here, a lot of potential for time saving and also a lot of potential to broaden the use of the tool, not just to expert, but actually to new people coming in. And we have this available now. It's available to Modelon's customers, to Modelon Impact on the cloud, and we have been trying this also with pilot customers for some time already. But it doesn't stop with that because while the AI assistant is perhaps more for help onboarding and troublesooting, we all know that AI can do so much more. And often, this is referred to agentic AI, where you basically hand over the problem on a different level to an AI agent. And we have done a lot of work in this space, too. In fact, AI agents are already integrated and available for advanced Modelon Impact users. The reason why we can make this so quickly available is that our fundamental technology, which is based on Modelica, an open language and an open standard, it's basically text. And the combination of a text-based standard and large language models is really the ideal fit. So that's why we have been able to quite quickly generate value, also thanks to our cloud-based platform, which readily integrates with all cloud-based AI services on the market. We have applied this to different application domains. There's one long study here that Johan will talk about in a while, which looks into vehicle dynamics, so analyzing how cars move are designed and how to stabilize that. It's a complicated area where we have shown then that an AI agent in combination with physics-based simulation in Modelon Impact can generate results in minutes or hours that before took days, weeks or even months. It's quite fascinating to see. Now this doesn't mean that our customers can delegate their whole work to AI agents. Like all AI, there are limitations and there needs to be human control points, there needs to be traceability. So there are a number of questions that need to be asked in order to use agentic AI in a clever way. And that's also something that we are right now guiding our customers with. And you can read more, as I mentioned, on our blog page. But a lot of progress in AI, and you should stay tuned for more coming from Modelon. But now I've talked a lot about AI, but I think it would be time that we actually look concretely into what we have. So I'll invite Johan here next to me, Johan Andreasson, our Chief Strategy Officer. And Johan, I know that you could you prepared a little bit of how we could demonstrate what we have today in the product.

Yes. So let's start with the AI assistant then and see how it would look concretely when you need help to debug something that you built.

So what is this system here?

Yes. So we're actually looking at a heat pump, and this is one of these so-called 2-phase systems that are quite complicated and so what you can see here on the screen is the system, and you can also see at the very top here that there is an issue with this model. It doesn't simulate as expected. So...

That looks really scary. Compilation failed itself.

Yes, yes. And of course, if you open and look for the log, you can get much more details. But now there is also a much easier way to handle this, and that's just to ask the AI assistant directly what we can do about this problem. So let's try it.

Let's have a go. So this is running live on the tool right now.

Yes. So if we say what is wrong?

What is wrong? Okay.

What is wrong with my model?

What is wrong with my model? Yes, that sounds like a question.

And we see what it's saying. So as you say, it's running live now, and it has full context awareness. So it knows what model we're looking at. It knows the latest simulations and everything. So that's what it's looking for now.

So you didn't tell it anything before. It's just looking at the problem right now and then answering what it sees.

Yes, exactly. And as you can see here, it's starting with giving a recommendation here. It says that there is a missing, something missing. You have really good eyes, you can see that I actually missed to connect here, right? So I don't have to worry about the log. I can get the recommended action directly from the AI assistant.

Yes. Okay. So does this happen often to our customers, I think?

Yes. I mean it's a common problem. The more complex systems you build, the easier it is to forget something, right? And so this is a significant help for both experts and not so expert users, I would say.

Very good. Very good. But then I know you have something else you wanted to show here today.

Yes. You mentioned Agentic AI, right? So we're all lazy by nature. So let's see what we can have the AI do for us more. So we have another example here that I will show, and I'll bring it up here.

Yes. Okay. This looks like a car now.

Correct. So what we see here is a compact car. And what we wanted to do here is to see can we improve what you call the lateral performance of this car? So I should mention, first of all, that this model that you can see here with the car and the boundary conditions, it's not something that I built. It's actually built by the agent.

Built by the agent. Yes. And then you ask a very high-level question here.

Yes. So I asked it, please improve the lateral performance of this particular vehicle then. And as you say, it's a very high-level question. So what happens now is that it's actually suggesting an interpretation of what that could mean, right? Because there could be different interpretations of that. And it's actually saying, so lateral performance, it means how fast can you go in the curve, what's the grip you can reach and also how easy is the car to control at the very limit, right? So I say, yes, of course, great idea. Let's see what happens.

That's what we call the Moose test in Swedish, right?

Well, this is slightly different. Moose test is another version of this, which focus more on stability and less on the ability to go fast. But -- so what happens here is you see we get a lot of numerical values. It's providing me a lot of detailed information directly in the prompt, which is, of course, great. But it's much easier to look at it graphically, right? So it's also providing this report for me here, where you can see we have a summary that is then focusing on a visualization of -- you see all these different dots or the different cases, it's right? And you can see it's interactive. It started with this value down here. And effectively, what's good in this graph is to be top right. You want as much as possible acceleration, means you can go faster in the curve, and you also want it to behave nicely if you go over the limit, right? You don't want it to spin or anything like that. So top right is best. And you can see that based on all the cases it run, it also identifies here automatically what are -- what is the boundary or the pareto front, as we say, for this. And based on that, it also suggests a winning candidate.

Okay. Wow. So all this was generated from your high-level question then.

Yes, exactly. So of course, under the hood, we are developing, as you mentioned, skills to make this more and more powerful. So the strength of this Agentic AI is growing every day.

Yes. Very good. Very good. Fantastic. So what about the quality of this work? I happen to know that you have a PhD in vehicle dynamics. So you, if any, can judge the -- is it right here in the recommendations?

Yes. It's correct in the recommendations based on the -- what we agreed when we started the study. So in that sense, it's absolutely right. And the other thing that is super important to remember here is that it's not fabricating anything, right? It's running Modelon Impact simulations that are based on validated models. So it's not inventing any physics. It's just operating the tool like a vehicle dynamics engineer would do.

Wow. Fantastic. There's a lot more to say about this, but I think this gives a very strong indication of what Agentic AI can do with Modelon Impact and physics-based simulations.

And may I just add one more thing that is important also in the spirit of openness and everything. And that is that you could choose, right, what type of Agentic provider that you're interested in, it's effectively open for most of them.

So it's not really important what AI agent you use. What's important is that you have a physics-based back end.

Yes. So what we're offering here is that you have a great platform and you can choose your agent yourself.

Very good. Thank you very much, Johan. Maybe you can help me to come back to the PowerPoint here. There we go. Thank you. So with that demonstration, I'll actually hand it over to another colleague here, Jonas Eborn, our CFO, who will give us a financial update for Q1. Yes.

Thank you, Jan. And I'm starting with the annual recurring revenue as usual. In the first quarter, the ARR amounted to SEK 47.2 million, which is a decrease since last year of 12%. The ARR is reported at constant currency. So this is adjusted for all currency effects, which were quite significant in the period also. Quarter-over-quarter, the decrease is 5%, and this is due partly to Modelon Impact churn from, for example, federal accounts last period, but also the legacy accounts, multi-platform products that we sell, roughly about the same in quarter-over-quarter decrease in both of those. If we look at the revenue, our revenue in the quarter came in at SEK 15.3 million, which is 25% lower than last year. This is not currency adjusted. So around half of the decrease is due to FX effects. The software revenue is SEK 12 million, which is primarily recurring revenue, SEK 11.8 million. Services, we are at SEK 3.3 million, which is a lower or more -- a bigger decrease than in software revenue, a minus 38%. The development costs are decreasing as well. And we've seen that since the restructuring that was done in the second quarter last year. We're now leveling out at the level that you see here, SEK 7.5 million development costs in Q1, and we're expecting roughly this level of cost going forward in 2026. If we look at the expenses and EBIT, the OpEx came in at SEK 23.7 million, which is lower primarily due to lower personnel costs. The personnel costs were SEK 16.3 million in the quarter, around SEK 5 million lower than the previous period. And all in all, lower revenue and also lower costs. This resulted in an adjusted EBIT about SEK 1 million better than the last period. minus SEK 7.2 million was the EBIT -- adjusted EBIT in the period. Cash flow was minus SEK 2 million. This is not as good as last year. Last year was boosted by a couple of multiyear deals. So we got significant large payments, giving a positive cash flow last year. We're looking at how we can improve on this, of course, we expect the full year cash flow for 2026 to improve quite significantly with the lower cost base that we have currently. At the end of the period, our liquidity was SEK 40 million roughly, SEK 39.8 million. And then we can go back to a summary from Jan.

Very good. Thank you, Jonas. So just wrapping things up before the Q&A. What we've seen in the first quarter is a weak ARR development. The new sales that we had, a number of new and quite prominent customers did not yet compensate for losses of small accounts with legacy products. But we have a plan in place. We have a new sales organization in place. We have AI as a differentiator, and we're addressing growing markets. So I remain optimistic that we can return to growth during this year. Where we see improvement, continued improvement is in the cost structure. We're still doing the transformation of the company that we initiated second half of 2024, and we have consistently then been able to show improvements in EBIT, thanks to a more efficient cost structure. And we saw that also in this quarter despite lower revenues. We're proud to have large companies, both operators and equipment suppliers among our customers for data centers. And we're collaborating with these large customers in a way that we provide value, but they also can provide their own intellectual property into cooling system simulations, which is a very critical area to build the world's AI infrastructure. And talking about AI, we can, of course, draw a lot of benefits from AI, both in our own development, but with customers, we have started now during the quarter to put out new features that will significantly increase the productivity and take away pain, for example, pain of troubleshooting in system simulations, which sometimes can be a bit complex. And we have shown then that you can work with AI on many different levels. You can ask simple questions, you can ask for troubleshooting help, but you can also ask high-level engineering questions where you probably can't delegate all of your work, but a lot can actually be speeded up. And especially if you have domain knowledge like my colleague, Johan Andreasson here showed, you can actually generate results that before took several weeks or months in just hours or even minutes. So a lot more to happen in this space. And I'm very optimistic that this will be a differentiator in selling to both existing customers and new opportunities. So with that, I thank you all for listening, and I'll hand it back to you, Jessica.

Thank you very much, and thank you for a very interesting demonstration. So let's move on to the Q&A session. And we have a lot of questions from the audience, but let's start with some of my questions. And the first one is regarding the drop in ARR during the quarter that was mostly attributed to small one-seat customers. But are you seeing stable retention among your larger enterprise customers with more than one seat?

Yes. Thank you for the question, Jessica. Yes, I've been out visiting customers myself, and we have a sales force that are talking to our customers every day. And I think with several customers, we have a much broader position where customers are using us in various ways as even the system of record for designing their systems in various applications and various industries. But primarily, I'd say, in different kind of HVAC systems. And that could be heating or cooling systems in different domains that could, of course, be for data centers, but it could also be for aerospace or for automotives or even for residential applications. So yes, we see a lot and hear a lot of positive feedback from those customers. I think where we have seen the change is primarily then in customers using our legacy products, products that we perhaps deliberately have not been putting so much attention on since we have been focusing our investments towards growing markets and growing geographies.

And you sounded quite optimistic regarding a return to growth in the second half of the year. Could you just summarize the driver behind your optimistic outlook?

I think the key thing is that our new sales organization is generating a growing pipeline. Then a growing pipeline still means that you need to convert into deals. And with some of the larger opportunities that we're looking into, these are large customers that take their time in evaluations or purchasing processes. So we do have long sales cycles. They can be between 6 and 9 months. But I think a growing pipeline is something that I think gives us a lot of sort of optimism and I think also reasons for optimism for converting into deals and revenue growth. I think on top of that, what Johan Andreasson showed here in a while that AI is really on everyone's lips right now. AI is not just yet another feature. It's actually a radical shift in how customers will work with the whole engineering flow, but with system simulation in particular. And I feel and hear from customers that we're well positioned in this space. I also know that we have more ongoing. So that also tells me that we have upselling opportunities, both with existing customers and an edge when we are competing with other customers for winning new deals.

And regarding the pipeline, could you put some more color on it regarding what types of companies and what industries and geographies?

Our key geographies remain the same as they've always been. We see the advanced markets in North America, in Europe and in Northeast Asia. Those are the key markets, and we're well positioned there. We have strong sales and sales support teams in all those 3 markets. So -- and in particular then, I mean, we address different kind of industries, but there are some common trades. And the common traits, you would say, are very much in thermal fluid systems. That's a critical word. But what it really means is different kind of heating systems based on liquids or gases that you will find in anything from a heat pump to the cooling system of an airplane, the cooling system of a car and very much the cooling system of a data center. And especially the latter then, we can read about it in the newspapers every day that data centers is a growing area with a lot of money going in and with a lot of new technology driven by GPU technology. NVIDIA is one of the main suppliers there. But that requires then a whole ecosystem of cooling providers and new technology. That's where we come to play. And that's why system simulation is so important to really put all those components together from different suppliers.

And some questions from the audience just regarding the data center opportunity. How soon do you expect this early-stage engagement to significantly impact revenue?

If you go back, we have already had significant impact on revenue. You can look back into a press release we did in December 2024, where we were happy to announce a major hyperscaler among our customers. So we're -- I think it's going to be a step-by-step journey as anything. And I expect that with the leads I see and the growing pipeline, step-by-step, starting from the second half of this year, I have good hopes that we will be able to announce more data center customers, either on the equipment provider side or on the operator side.

And some more questions from the audience, and it's regarding the AI assistant here. What does your concrete pricing and packaging strategy look like to successfully convert this customer interest into Agentic AI into actually increase in ARR during the remainder of 2026?

Yes. Thank you for the question. It's really an important one. So I think AI will -- and AI features will contribute to our ARR in 2 ways. One is just that we become more competitive. We'll be able to sell more software licenses because our tool becomes more attractive. And that's actually how we're starting out right now. But we will, of course, be looking into more advanced monetization models. I mean the prevailing way to monetize AI today, if you look into other companies, is token-based, so consumption-based models. And that we have not launched yet. But as this grows and the usage grows, we will leverage those opportunities in a way that is transparent for customers and generating new revenue opportunities for us.

And another question from the same investor. Given the strong global trend surrounding AI infrastructure, how large share of your future sales pipeline currently consists of this data center solution? And do you expect this segment to compensate for the loss of small accounts within the legacy product?

So I cannot give an exact number of how large part of our funnel or future sales that will come from data centers. That would be a too detailed breakdown at this time. But what I can say, you can just look at the market data in market data for thermal fluid systems and cooling systems, it's pretty clear that the fastest-growing market right now and also already pretty big is the data center market. That's why we see most of the development. And you can look into the corresponding earnings calls from, I think, any equipment provider right now, they will have data center on their radar. So we are addressing both them as equipment providers because they need to simulate their components that could be chillers or fans or heat exchangers and the end customer, which very often is a data center operator that puts everything together and needs to dimension a certain data center at a certain power efficiency in a certain geography. So I mean, we see growth opportunities with all of those.

And moving back to the AI Assistant. You noted that it helps non-experts getting started. Do you expect this to shorten your typical sales and implementation cycles?

I think it will help both nonexperts and experts. I mean, quite frankly, even some of our experts can run into things where they need to spend unnecessary time, for example, for troubleshooting. So I think it will help everyone. But as you point out, it's probably primarily the newer users, either if you're a newcomer, beginner or that you have spent only some time in the tool, you can quicker get to results. So yes, I think it will shorten the onboarding time. It will facilitate the onboarding. It may even reduce the need for support from us. And I expect that it will also help expanding the use. So rather than staying with 1 or 2 experts, I think this creates an opportunity to quicker go to more engineers at the customer who can use and get value out of our tool and products.

And my last question, what should investors expect from Modelon for the rest of 2026?

I think I gave an indication of what to be expected during this year. We are continuously innovating both in the platform and in our libraries and also in our business models. So I think investors should expect that we continue to have focus on AI, how AI combined with physics-based modeling through Modelon Impact will help customers and will continue to generate value and revenue opportunities. I think investors should continue to see how we deliver value and come out with news in the data center space, both from a customer point of view and from a technology point of view. And last, I think the question on monetization is very relevant also. I think investors will -- should expect that we think through and maybe add monetization opportunities out of things like scaling, AI and other things that differentiate us, thanks to being a cloud-native company.

Thank you very much, Jonas, and Jan.

Thank you so much, Jessica, and thanks, everyone, for listening.

Thank you.