Schrödinger, Inc. (SDGR) Earnings Call Transcript & Summary

Welcome, everyone, to our annual health care forum. We're kicking off things here with Schrodinger. Happy to have both Ramy Farid, the CEO; and Richie Jain, the CFO. So welcome both of you, and thanks for joining us today on our virtual fireside chat.

I guess, Ramy and Richie, maybe walk through for the investors that are new to the story, because a lot has even changed in the last 6 to 12 months about the Schrodinger platform and how you've made some changes.

Of course. Yes, so I'll start us off. At the highest level, what our goal is, our mission is to develop a computational platform that allows researchers both in life science companies and in material science companies to design better molecules more rapidly, more efficiently. And that requires developing a platform that can replace experiment because the traditional way of doing drug discovery is by trial and error. You make a molecule in the lab, you assay it, check its properties. And if it doesn't have the properties you're looking for, which, of course, will always be the case when you start off a project, you start to optimize it. So you try to make a change to the molecule. And obviously, that's very time consuming and prone to huge failure rates, as we all know. So the whole goal of computationally driven drug discovery materials design is to do all of that on a computer and do it accurately, in other words, replicate the experiment and do it on a massive scale so that you can test huge numbers of molecules and find that perfect molecule that somehow magically balances all the properties that are required to be a drug or a particular material. So what we have actually successfully done, and we would argue, and I think with a lot of evidence to back it up, that we are the first and only company so far that has developed a platform that can replicate experiment reliably on novel molecules, completely novel molecules, new chemical entities on a scale that is large enough to actually be able to find those very special molecules. And here's the key. It's validated. We have for the last 15 years been actually using the platform ourselves to advance a number of drug discovery programs with companies that we either have co-founded or with pharma companies on our own behalf, and the success rate and the track record is extraordinary. We've got 15, 16 programs in the clinic. All the biotech companies that we have co-founded had really highly successful exits. Pharma collaborations are going well. Our own programs are progressing and going very well. And on top of all that, we've been licensing that software to the whole entire industry. Every pharma company is using our platform, some at different scales, some at very large scale, some at a little bit lower scale. But everybody is using it, we'll get to that in a second. And essentially, our customer retention rate is 100%. What does that mean? That's another validation. I mean you don't keep renewing an annual license contract if the platform isn't having a profound impact on the projects. Now here's the thing. The most exciting sort of technology that we've developed is relatively new actually. And so we're still in the early days of sort of scaling up and achieving the true TAM of this business, which is far larger than where we are right now at around $200 million per year. We have a handful, let's say, roughly of pharma companies that are using the technology at scale. That's very exciting. But there are other companies that are sort of still ramping up, and we see that as a tremendous opportunity for growth in the coming years for every pharma to be using the software at scale. And then the last thing I'll say is we're very excited about -- we have a big investment in the platform. We're a science company, an innovation company. We're leading the field. And this year, it turns out that we have a few very exciting new products that we've released, probably one of the most exciting is predictive tox. It's one of the biggest challenges in drug discovery, is predicting toxicity associated with binding to off-targets, and we've released that this year. And so we continue to make scientific breakthroughs, lead the field with new products. We expect that will continue to also lead to growth in the coming years from adoption of these new technologies as we continue to reduce the time it takes to get to a development candidate in drug discovery or a material and increase the probability of success with much higher quality molecules.

Great. And I'll just add a couple of the recent changes that we've made reflecting the strategy that Ramy just outlined. We have made some changes to kind of simplify and clarify the business structure. We had been executing a few programs in the clinic on our own. We are seeking now to partner those programs. Ramy referenced 16 programs in the clinic that are advancing. Those are advancing in the hands of our partners, where we have downstream milestones and royalties on those programs. We put out a 3-year goal of achieving adjusted EBITDA profitability, which is achieved by growing both the software and drug discovery businesses and maintaining expense discipline. And then we also announced a change to focus and emphasize hosted contracts in the software business as opposed to on-premise contracts. We think that will also take about 3 years to transition over. Today, it's about 25% hosted. We expect to get to 75% hosted. Given the way revenue recognition works, this is a very common transition for companies to make towards hosted and SaaS solutions. But in the near term, it will have the impact of reducing revenue, especially in 2026. All the while the business is still growing. So we've changed and emphasized ACV this year as a business operating metric to track the business growth while the revenue catches up to ACV over the course of the transition.

Great. And maybe as a follow-up there for both of you. Why pivot towards this strategy now? Maybe I'll just broadly leave it there.

Yes. Maybe I'll start. You're talking about the transition, of course, from on-prem to hosted. Is that right?

Exactly. And moving away from the internal pipeline, away from the clinic.

Both things.

Both strategy. Why now? And why do this?

Yes, it's an excellent question. What we recognized is that the reason why we believe that we have developed this platform and why it's working as well as it is, why it's as validated as it is, is the result of this unique synergy between our software business, licensing the software to thousands of users, and using the software ourselves in collaboration with other companies or on our own behalf in the discovery phase. What the result of that is, you can imagine how straightforward this is actually, that by using the software ourselves, we're learning what works, what doesn't work and we can improve it in real time. And of course, the validation is quite nice. All this validation that comes from creating development candidates in these high-quality molecules is very helpful in convincing a whole industry to change the way they do drug discovery. That's what we've done. Drug discovery used to be done by trial and error, now we're using computation. And so that's what we wanted to focus on, those synergies. And that doesn't require running clinical programs. So synergies are sort of lost. Now you start getting into other factors, other things at play other than just designing molecules, which is where our core competency is, where our competitive advantage is. And I'll tell you something else, too. I mean, to be very honest, when we started the clinical programs, since then the world has changed dramatically, what is occurring in China and the way they're able to do clinical trials much more cheaply and more quickly, the cost in the U.S., the Project Optimus, oncology itself. A lot of things have changed that have resulted in the cost associated with running clinical trials being quite a bit higher than we had expected. So again, focusing our capital on where it makes sense, on these synergies, on the software business, on advancing the platform and on the discovery portion of our therapeutics group is a big reason why we made this change. I'll hand it over to Richie to elaborate on that. And also, it would be good to talk about the hosted, I think, transition too. That's important.

Yes, I'll cover the hosted piece, which is we have been moving towards hosted contracts gradually over the past few years and, through that process, have surmounted a number of key hurdles. We have transitioned some of our largest customers from on-premise deployments to hosted as well as doing initial deployments hosted. But with our largest customers, we've passed all the tests from vendor audits, supplier audits, quality requirements. We've passed the bar with our most difficult customers. And customers are increasingly moving towards cloud-based solutions, especially in the Western world. So it satisfies their objectives. It also -- as we were looking through the business, a number of our deals were moving towards hosted solutions over the last couple of years at the customer's direction. And so we were kind of reaching a tipping point where it made sense to move all the way over. From a customer-facing point of view, from a support point of view, we can support better, we can deploy faster. We can get the customer up and running in a shorter amount of time from when we get the order to when we can deploy. From an investor point of view, obviously, our profile over the last few years has been lumpy just given the on-prem accounting recognition rules. And so we think this will provide a better picture for investors to measure our business and have a more smoother, predictable profile. And finally, from a renewal perspective and an ongoing support perspective, we are just better positioned to understand the value customers are getting from our services, from our software and ensure that they're deploying it properly across their organizations, that every site, every location is using it uniformly such that when we approach a renewal, we are better positioned to understand their needs and our needs and capitalize on that opportunity.

Great. That sums it up perfectly. Okay. Moving on to the end markets here. How is the funding environment? What are you seeing budget allocations looking like? Maybe we'll talk about that broadly and then we'll go dive into maybe across the different tiers, the customer tiers that you guys are organized around.

Yes. I think like has been widely reported, we're obviously, like these other reports, optimistic about what this year looks like, especially compared to last year. So where you can see that we have guided to 10% to 15% ACV growth relative to lower growth last year, which was the result of the budget pressures both in pharma, which is sort of scary, but again, I think that's getting better, and obviously, in biotech. So we're confident with our ability to achieve that sort of growth, following a very difficult year, obviously, for the whole industry.

And just to add to that, Scott, our growth outlook for this year reflects not just one budget category. We have released number of new products this year. The [ leads ] are releasing a number of new products this year that will touch additional budgets. So that's a part of our growth story for this year.

Yes. Let's dive into that maybe more. So I think a big part of your next growth algorithm is unlocking more budgets that you haven't had exposure to in your life science end markets and your customers. Predictive tox is one of them. Let's talk about that strategy and where you think you can take that over the next several years.

Yes. So there are a number of things that are pretty exciting about that. The obvious thing, of course, is tapping into new budgets. It clearly increases the TAM for the business. But here's one of the things that we're most excited about with regard to predictive tox. Because of the nature of current computational methods, so the current methods require huge amounts of training because they're machine learning based solely. What does that mean? That means that they get done later in projects, way later. And so what does that mean? That means that essentially, so you're doing a discovery project, working on a molecule, you're working on it for a few years, you've spent tens of millions, $20 million, $30 million, $40 million, 3, 4, 5 years. And then you'd start testing it first maybe using these machine learning models, which now start to work kind of because you've generated a huge amount of data, which is, of course, what's required when you're using machine learning-based methods. And if it lights up and is starting to show toxicity, that's it. Project is done. It's very difficult to -- what are you going to do about it? I mean you have to go back to the drawing board and start redesigning the molecule even though you've kind of started to hone in all the properties. And here's what happens. You start trying to improve the toxicity profile and, of course, you start messing up everything else. It starts not being potent. It starts not being soluble. It's not permeable, whatever. It's a multi-parameter optimization problem. The nature of what we've built is, first of all, it's highly accurate. But remember, I said at the beginning, it can be used on completely novel molecules because it's physics-based. It's not machine learning based solely. So what that means is that you can use it early in projects, way earlier. And of course, that has a huge impact on its TAM. So not only are we tapping into new budgets, but we're creating a new sector in some sense, a new budget that is those same people that are running the toxicity, but moving it way up in the process, which requires, of course, way more usage. And here's the other thing. With regard to these other solutions, they just tell you, yes, no. You're toxic or not. They don't tell you why. You can't do anything about it. The methods we've developed not only work on novel molecules, you don't need to train because they're physics-based, but you get a picture literally of the molecule, the structure of the molecule bound to the target that's causing the toxicity. And that means you can start to dial it out and, again, use it early on in the process as part of the multi-parameter optimization workflow.

Two follow-ups there. One, are these new products that you're developing in response to customers' needs, are coming directly from them? And secondly, can you build these all yourselves? Or is there a way that you need to have some bolt-on acquisitions to advance this growth initiative of developing more solutions to expand the budget?

Yes, it's a great question. It ties back to what I said earlier. Every company almost in every field struggles to innovate through asking customers what they want. There's the famous quote from Ford. He said, if he had asked people what they wanted back before cars existed, they would have just said, faster horses. And you've heard quoted like that from so many innovative companies. In order to really understand what it is that's going to change things fundamentally and really innovate, you have to have a deep understanding of the problem. And in our field, where are you going to get that? By doing it ourselves. And that's why we have a therapeutics group and why we built it. So the impetus for developing this technology came from our own projects and our own collaborations. We kept finding that we were running into this toxicity problem. Of course, it's off-target problem. In latent projects, you're hitting hERG and you discover that kind of late. Now hERG is an example where people are testing that a little bit earlier, but that's an example. And there's so many targets like that. So then what we did is we went and we had meetings with senior people at these companies and interacted with them and said, hey, what do you think about this idea? And of course, got really great feedback. But the original idea has to come from within the company that's actually innovating to really make groundbreaking sort of scientific advances.

And then on the second part of the question...

To address the second part, yes, Scott, I think we have positioned ourselves from a balance sheet point of view to have capital. We have a capital position to fund the business for the next few years to get to the point of profitability. As it relates to M&A, within predictive tox, I think what we're developing is truly unique and has the potential to transform the workflow. But just taking one step back and looking at where we sit in the entire workflow, if there are capabilities that are complementary to us, that are near where we sit in the workflow, we will, of course, consider M&A. I don't think we're going to do clinical trial optimization. That's very far downstream from where we are. But if we can find additional capabilities close to us that are complementary with the platform and are complementary with our customers, we'll take a look at that.

Great. Maybe let's talk about AI. On your recent earnings call, you talked about embedding agentic AI on your platform. Maybe just talk about the benefits of this and how many applications we can see with agentic AI as we head throughout this year and beyond.

Of course. Yes, it's very much on top of a lot of people's minds. So let's first make sure we understand really what agentic AI means, and then we'll tell you why it is something that we think is very important and what we're doing there. So at the moment, you can imagine this being the case, especially when a technology is just sort of coming online and, again, a lot of companies and a lot of different spaces that are innovators run into this issue, which is there aren't a lot of experts that know how to use it and truly use it, use it correctly at scale. Of course, the scientists at Schrodinger can. And we put a huge effort into making the software easier to use, building workflows, writing lots of online courses and so on to train the next generation of computational chemists. But another solution to that problem is agentifying technology, automating it, making it so that -- not replacing humans. And I think a lot of people understand this. You're not going to replace humans, but you're going to make them more efficient and augment their capabilities so they're not doing sort of the menial sort of things and allowing them to scale, make one human significantly more efficient, being able to support more programs and take advantage of this extraordinary technology. So that's the goal and that's what we're working on. Now we should be clear. This is not easy. The agentification of many technologies is taking longer than people thought because it turns out humans are pretty good at driving cars, for example. And look how much longer that's taking. And that, believe me, is way easier than designing a drug, way easier, way less of a complex problem. But still, nevertheless, all of these are complex problems. So if we can start to encode the knowledge that humans have, the most expert humans into the technology and make humans more efficient, obviously, that will have a really huge impact on the whole field and, of course, on our business as well because, of course, it increases the demand for the technology in a really serious way. So we're very pleased to be working with a number of the sort of large companies that are building these LLMs. Anthropic is one of them, as we mentioned in our earnings call. Great discussions with them. We think this is something that requires a partnership like that, where we supply the expertise. But of course, the sort of foundation that these companies have built is quite difficult to replicate, obviously, fully internally.

Just to add two points, Scott. We've set up the business to capture this additional demand on the software. So what I mean by that is we have throughput-based licensing for the majority of our products. So as there are additional workflows being called by our user base or an expanded user base, we capture that all by selling on licenses and tokens as opposed to seats. And then the other piece around agentification is that we have had historically a relatively small user base that have the experience and the capabilities to run our tools at scale. But with agentification, we expect that we can, over time, increase that user base by converting over chemists trained on traditional methods over to computational-based methods.

I think those are two really important points. One, you're not a seat-based. You're utilization-based, throughput-based. And two, that these agentic will be able to -- part of, I think, the barriers to entry was not having people be able to use the software in the most effective way, and having these agents should certainly improve that. So those are really important points.

Yes.

I guess let's walk down the financial model a little bit more just to clear any misunderstanding. So 10% to 15% ACV will translate into future revenue streams next year. This year is the sort of transition year. We're expecting profitability by '28. Let me walk through that, right? So currently, in the legacy model, we had a lot of multiyear contracts that renewed in the fourth quarter that were not hosted. And we're moving towards a more stable quarterly distribution of revenue streams as we head out over the next several years. And then margins should take a stair step function lift next year and beyond as we get towards profitability. This is a question for Richie. Maybe walk us through what the implications are for your guidance in case there's any misunderstanding about the financials of this process of getting to a more stable quarterly revenue project stream.

Yes, thanks. So at the end of 2025, ACV and revenue were in lockstep. Software revenue was $200 million, software ACV, $198 million. And to reemphasize, ACV and revenue will always equal each other over the course of time. So if we sign a $1 million contract ACV, the revenue will equal $1 million over the duration of that contract. So in this year where we are in earnest starting the transition, and what I mean by that is the majority of our contracts are 1 year or less, so at the renewal, over the course of 2026, we expect to convert the majority of these contracts over from on-prem to hosted to the end goal of 75%. And why not 100%? There are always going to be some customers where conversion is just not an option. It's based on the geography, based on what the end market is. But for the most part, Western world pharma biotech customers, we expect we will be able to convert. Actually, by the time we had our Q1 call, we had already converted over one of our pharma customers from on-prem to hosted. Over the course of the past few weeks, actually, we've tackled one of our multiple year on-prem deals and converted that over to hosted actually before the renewal date. So we're tackling these throughout the year. But given the majority of the business is booked in Q4 just given the budgetary cycles of pharma companies and biotech companies, in that quarter when you switch from on-prem to hosted, you're going to go from on-prem, almost 80%, 90% revenue recognition in the quarter of the booking, to ratable. Let's just pretend a deal was booked November 15. You're going to be picking up 1.5 months of revenue recognition this year and creating a large deferred revenue balance that will be recognized in 2027. So that is what has driven our focus on ACV for the year in a year in which we expect revenue to decline because of the phenomenon I just walked through. The key point for investors to really focus on is that everything I just said is cash flow neutral. So our cash flow from operations that you would see at the end of 2026 will not change from all of this accelerated transition to hosted. In that line item, it will be exactly the same irrespective of how we've approached the year. And as you think about that over the course of the 3 years, we maintained the 10% to 15% growth trajectory for the 3-year period and expect to get to 75% hosted revenue. The reason we think those are important benchmarks are that revenue will start to converge with ACV. You'll start to see the two track each other. But as we continue to grow the business, we do think that ACV will be a leading indicator of revenue even after the transition period, going out into the longer-term future.

Great. And I guess my last question. We can open up to the audience. If there's any questions, please feel free to put this in the chat box here. We talked about new product releases to tap into different areas of budgets for pharma. I guess my question is, is there a set plan of how many new releases we can expect from you guys? You did predictive toxicology and that went through beta for several years. And will that be the process, where you announce that you put into beta to test it with your customers to see? because obviously, putting it in beta actually helps you to get live feedback and make this the most monetizable product that you can and successful product that you can. But how should we think about that in terms of the framework? Because it seems like it is a growing part of your growth algorithm, and it makes sense because you're tapping into all these new budgets that you had never had access to before. But how should we expect the cadence of new product releases?

Yes. We did something with predictive tox that we haven't done in the past. We've had many, many new product releases over the years and new ones that have tapped new budgets. But we haven't talked about the beta. We did it this time because, first of all, there was a rather large grant associated with funding this project, which is kind of unusual. That had an impact on our gross margin. So we were sort of talking about that. But the FDA kept talking about the importance of computational methods for tox prediction and their attempt to reduce or maybe eventually eliminate animal testing, I don't think that's realistic, but certainly reduce it significantly. So we felt compelled to talk about it in advance. We don't normally do that. So it might look like something new was happening. It wasn't. This is the normal cadence, multiple new product releases or major enhancements to existing products every year. We have 4 releases every year. We have a large R&D effort in this area. So there are new products coming out all the time. We won't always announce publicly the beta release. But we will certainly announce the -- and we may not even publicly announce actually on that. I guess, of course, to customers, obviously they're hearing about the new products. But this is a pretty normal cadence actually. And it will continue into the future. There's a lot more to do. We're making fantastic progress. But as long as there's any failure in a drug discovery project or any project takes more than, I mean, even a few months, you should be able to get to a development candidate much more rapidly than we're doing now. There's still lots of new science to be done, lots more breakthroughs. And we'll continue to lead the field in those, in advancing the science.

And I guess maybe to end it here, we're coming up on the 35-minute mark here. Maybe for both Ramy and Richie, we've heard a lot about advancements in AI and pharma. This is a big application for pharma. What are your conversations like over the last 3 to 6 months with large pharma on this topic? Are there more inbounds than you've had ever before trying to understand your product offering? A lot of people will argue this is either a cannibalization or a catalyst for partnership. Maybe talk about what you're hearing from customers directly.

Yes. We see this as a tremendous tailwind, sort of the demand for AI is being used. That word is being used to really mean computation. So yes, the excitement around AI has over quite a number of years now dramatically increased the interest from traditionalists. Medicinal chemists are generally the ones running research groups saying, something is going on here. We need to be using computers. AI is just used as a shorthand quick way of referring to computation. They understand what everybody, I think, needs to understand. AI is powered by training sets. It has no utility without the training set. That's what AI is. You have to train. And I think they understand very well because they've been testing this for many, many years, that experimental data alone is not sufficient to train these AI models. You have to generate simulated data, just like in self-driving cars, just like in chip design, weather prediction, every field. And those are simpler fields than what we're in. You need to generate simulated data. They understand that they need our platform to generate the massive amounts of data that are required to actually power AI. So yes, everything is, I think, heading in the right direction and we're pretty excited about the future. And it's great that there's so much attention being paid to computation finally.

Yes. Well, that's great. I think this is a perfect place to end it. Well, thank you so much, Ramy and Richie, for doing this fireside chat with me.

Thank you. It's really great.

If the audience has any follow-ups or whoever wants to be in touch, please reach out to us. Thank you very much.

Appreciate it. Thanks, Scott.

Thanks, Scott.