All right. Good morning, everybody. Welcome to day two of the UBS Global Healthcare Conference. I'm Kevin Caliendo, UBS Healthcare IT and Distribution Analyst, and we are very proud and happy to have with us today management team from Schrödinger, Richie Jain, who is the Chief Financial Officer, and Pat Lorton, who is the Chief Technology Officer for software. Guys, thanks so much for having us.
Yeah, thank you for inviting us.
I should mention, I think you are the best performing stock over the last three months of the Russell 3000, or you're one of the two, up over 100%. So congratulations. I wish one of the three of us up here was smart enough to assume coverage with a buy rating three months ago when I did that. So thank you for succeeding and making me look bad. No, congratulations in all these ends. It's been fantastic to watch. Maybe before we get into the Q&A, why don't you talk a little bit and introduce everybody to Schrödinger more so?
Great. Yeah, I'm happy to do that. Thank you.
Thanks.
We are a computational platform for molecular discovery of high applicability to life science and material science customers. If you look at our business, it is weighted towards life sciences, but material science is a growing part of the business. We monetize the platform through licensing of the software. That is the majority of the business. We also monetize the platform by doing our own therapeutics, and by doing collaborations and NewCos with companies in life science. If you think about our outlook and trajectory, we are a high science, high innovation company bringing new products to our customers. We are continuously evolving the platform and expanding the domain applicability beyond just small molecule discovery, but even moving into new products such as formulations, expanding in from small molecule into biologics. We have also been innovating on the AI side.
We have rolled out an agent called Bunsen, which Pat can talk about it in more detail, but it is allowing our customers and users to just gain efficiencies in running the software. We have also been evolving the business model a bit. We are focusing on the platform-centric elements of the company, so licensing software, doing early stage therapeutics. We have also announced this year that we are moving towards more hosted contracts, and focusing on ACV as a result of that.
That is helpful. You mentioned AI. We should probably start there because it seems where everybody wants to start these days. It is clearly critical. We had a CRO panel yesterday where we talked about it. You cannot talk about drug discovery and drug development without talking about AI. It is clearly critical to the future innovation for drug discovery. Can you talk a little bit about your overarching AI strategy, how you approach it, how you use it, how you promote it, even within maybe? Why don't we start there?
Yeah, sure. I would love to talk about that. What is interesting is Schrödinger had a bit of a reputation for a couple years of being, we will say, anti-AI. It was not necessarily because we are anti-AI at all. I can actually say we tried to build the Bunsen agent, the first version of that, I think a little over four years ago now, I think when ChatGPT 3.5 came out. We have wanted this level of AI, but a lot of the hype around AI was being completely oversold, and we viewed it kind of as our job to stand up for the physics-first deep calculation approach. But that was using AI to make predictions, making AI to drive the actual discovery process and run the calculation engines, similar to how AI is able to code now, similar to Claude Code, how it is dominating all coding.
We viewed that very early on as something that could be transformational, and we were lucky because the second that that actually became true, that the harnesses became good enough, that AI was actually good enough to drive drug discovery, we started building Bunsen, like the current product, I should say, instead of experimental. We had early working versions of Bunsen more than a year ago now. I think we first started talking about Bunsen Q1 last year.
Q1 of this year.
Oh, I'm sorry. Q1 of this year, yeah. By Q4 of last year, we already had our internal scientists proving that this thing works, and we're extremely bullish on its ability to really drive discovery. We've mentioned several times that a big limiting factor we've always had is there aren't enough computational chemists, talented computational chemists in the world. We'll visit users, and we have introduced three new products, and they say, "I don't have enough time to use the products I'm currently buying from you." We think not only does this solve their ability to expand the scope of prediction and drug discovery, but it also solves their ability to run it at a much greater scale. That's why we're so bullish and excited about AI.
You're solving for the chemists? Is that in essence?
We are solving for all drug discovery, actually. I mean, you look at how powerful these engines are, and you can start thinking about biologists asking questions that only chemists could ask before. We are now at the place where we are comfortable with a biologist asking a question that not even a chemist could answer. A PhD computational chemist was required before. We are not putting computational chemists out of business. There is still a lot of stuff they have to do. But the ability to simplify some of the more straightforward tasks is really going to transform the predictive abilities in this industry.
I will just add on the AI framework, obviously, it is all about the data and it is about the model. Generating data in our field is extremely cumbersome and burdensome. There is not enough experimental data you can generate to run an AI model. You have to supplement it with simulation data, and that is what our platform does. We generate simulation data, in addition to experimental data, at a scale and throughput that you cannot replicate with experiment, and at the same accuracy as the experiment. That is how customers use our platform. It is generating simulation data to run the AI model.
To bring that home, our internal GPU supercomputer can run somewhere on the order of 30,000- 40,000 accurate binding affinity calculations a day. It is not just being used for on-target anymore, but in an entire five-year drug discovery program, they might only collect 3,000 data points. We are talking 12 drug discovery programs a day worth of data that we can generate on just our supercomputer. Most of our customers are using their own supercomputers too. So the gap between predictive data and real data is already, with current technology, orders of magnitude, several orders, and it is only going to get bigger as GPUs get faster.
I want to take a little step ahead. You are providing this 30,000- 40,000 data points. What does that mean for the pharma manufacturer who is like, "What does this mean for Bristol Myers or for Amgen?
Yeah.
Or somebody. Does that mean they have more targets, or does this mean they have fewer targets that are much more precise?
Let me clarify real quick. That 30,000- 40,000, that's for our internal programs only.
Oh, this is for your-
That's for our internal therapeutics.
This is not for your customers.
Yeah. Our customers have their own supercomputers generating their own-
Got it. Okay.
-bolus of But what it means for them is, every week, Bristol Myers on every program, I mean, obviously, it's generic pharma on every program, is probably picking 50- 100 molecules to synthesize. And each one of those costs a lot of money, and they take weeks to turn around, and the data to actually confirm that what you're designing is useful is very slow and cumbersome. If, when they pick those 50, they have looked at 50,000 and they are convinced that these set of properties, they're good, the success probability goes up dramatically. And success is speed, right? So when everybody is racing to get the first in class a little faster, speed is really what matters, and this allows them to operate much, much faster.
It's a lot.
Yeah.
I guess this is somewhat of a basic question, and maybe it's just my lack of understanding. What is Schrödinger doing that others can't? Why can't someone recreate what you're doing on their own? What's proprietary? I mean, without-
Yeah, no.
-giving away proprietary secrets.
I'm happy to-
Well, let me start-
Oh, sorry
-and then let me-
Yeah
-cover it from the business perspective-
Sure
-and then cover from the-
Yeah
-the details. One of the advantages we have is not only are we building a platform for molecular discovery, but we have an internal team of 200 people in our therapeutics business who are using that platform and giving us feedback on it.
We're not just building it and hoping it's going to work. We have a team of people who are using it and making sure that it works, so that when we bring it out to a customer, it is applicable to what they're doing. Go ahead.
You just covered most of what I was going to say, but it's great. What I'd say, what we have been doing is, we have a team of a couple hundred PhD computational chemists who are writing this code every single day. This is not the type of code that Claude Code is dramatically accelerating, to be honest with you. We have a lot of parts of our code where Claude Code is helping, but this stuff, there's not a ton of published work around it. It's in the nitty-gritty details. You actually need live feedback of assay data, which is what we get from our therapeutics team. We've just been doing this for 30 years. We've put over $1 billion in our platform, and we think we've actually done it quite efficiently, so it's not like someone's going to be able to shortcut that.
By focusing on squeezing out errors and inaccuracies, this is actually one place where AI really helps us. Let's say our product's 90% accurate and someone trying to come up has not put near the effort in and they're 70% accurate. You might be able to justify 70% versus 90% to save a lot of money if all you're doing is one data point. But when you're giving that tool to an AI and it's running for two weeks on its own and those errors are compounding, you need the most accurate tool. That's kind of what differentiates us. We've built, through a lot of blood, sweat, and tears, the most accurate tool.
Bunsen had some early success, especially Bristol Myers is one that we know about. Can you explain in greater detail what this product's doing with Bristol, and how it transforms your own workflows and your customer workflows a little bit? We talked about it a little bit, but what is Bristol doing with this that you know or that you can share?
I mean, to be honest, because of NDAs, I can't share a lot.
Yeah.
But I can speak generically about what a customer like Bristol would do that we expect almost all customers to-
A hypothetical pharma company.
A hypothetical Bristol-like company. Yeah. What it actually allows them to do is, we joke that every Bunsen session, which is not, honestly, does not look that much different than a ChatGPT session, aside from a lot of chemical built-in knowledge and visualizations. Every Bunsen session is a postdoc. So your computational chemist went from being a single person who is busy all day long to being able to have a team of postdocs running experiments for them, and that actually allows them to push more and more data into It's what I was talking about earlier.
It empowers them to start running more calculations, more risky calculations that would have been too time-consuming to take the risk, and to validate that our software and calculations in general are working more at a scale, and then push that into LiveDesign, our enterprise informatics platform, which they've also publicly talked about having. So that as you run all these experiments, the real data coming in is going into Bunsen, our experimental data is being generated and being pushed into LiveDesign, and the chemists have more information than they've ever had before to make the decision about what molecule is the right molecule. So it should be, for a company like Bristol Myers Squibb, the more they roll it out and embrace it, the more of an accelerant it will be for their discovery.
As that relates to our business model around Bunsen, as the customers are using the platform more, and they have higher throughput, they will need more licenses to run the platform. So, that is how we intend to capture the value of Bunsen, is through our throughput licensing.
Literally, and you are answering my next question, which is, how do you monetize? Is it licenses? Is it like on a per How should we think about it?
Yeah, I will cover it. We do not charge per seat. It is per license.
Okay.
The more licenses you have, the more calculations you can run in parallel. So it is quite simply, as Pat just said, we do not need more users. We need more instances of the software running. So it is a standard throughput-based license pricing model.
Got it. Okay. How can Bunsen be used outside of biopharma? I know you talked about this, we talked about this a little bit, and the idea that there might even be a greater unmet need outside of this platform. Maybe talk about that. I know this is a very big-picture thing, but it was fascinating to hear.
Yeah. I mean, my favorite joke, which I've been telling for a few years now, is that if you're on an airplane drinking a caffeinated beverage and reading something on your phone, every part of that involves a chemical discovered by Schrödinger platform. Or I shouldn't say necessarily discovered, but influenced by Schrödinger platform. We've been selling across many verticals in the material space for a long time. That said, materials businesses are not the same as life science, or especially small molecule discovery programs, where it's like you have to find a new billion-dollar asset every single year. For materials, it's more iterative improvement than it is new billion-dollar, well, hopefully more than billion-dollar assets every single year.
But the way we think Bunsen can really transform here is that where computational modeling was in life science, maybe in the early 1990s, admittedly before my time, but I've heard about it, in terms of there's one or two people fooling around with it in big pharma, but it's not transformative in any way, and it hasn't changed, it hasn't digitally transformed how things work. That's how it actually works in most materials right now.
And what we really think the opportunity for Bunsen is right now, where it might just be an individual off on their own, a single individual trying to make impact, if it can make that individual much more successful and steer them towards best practices, a lot of our materials contracts actually have to start out sometimes for years as service contracts, as we almost teach these employees how to use our software well, and we think Bunsen can actually speed up that whole adoption process so it goes to software sales faster than service sales.
Got it. When we talked after earnings last week, you had mentioned, and you kind of highlighted this earlier today too a little bit, that one of the growth hurdles is the number of computational chemists, or in general, people that know how to actually use this program. Is this a matter of training people to use it? Do they come and spend time with you for, spend two days to try to figure out with your team? Like, how does that work, and how big of a hurdle is it truly to get people? It's overwhelming to hear you talk about it, never mind trying to actually run it. Your passion about it is clear. How do you get somebody to actually be able to figure this out?
I'd say, why don't we tag team this one? But we have a lot of training programs and training courses available to academic institutions and non-commercial customers to have our platform be the standard as they learn about computational chemistry. There's also a domain of applicability, which we talked about a little bit earlier, but today, our platform is used mostly by computational chemists.
Especially those that have advanced training in the platform. Bunsen will make that adoption curve easier. But really for us, and this is not a today thing, but I think this is the next leg for us, is getting to the medicinal chemists who are working hand-in-hand with the comp chemists, but enabling them to be able to use our tools, so that not only are we growing with a smaller user base, but we're actually growing with a bigger user base. It's kind of where we want to be as we keep evolving Bunsen.
Generally speaking, just to get through numbers out there are about two orders of magnitude more medicinal chemists than computational chemists. So you'll have a fraction of a computational chemist on a drug discovery program, they're being shared across several, and then you'll have 10- 20 medicinal chemists on that same drug discovery program. So reaching them is a much bigger. To be blunt, medicinal chemists, because there are so many and they have such a long history of success, they run the chemistry departments, with a few small exceptions. So reaching medicinal chemists is very important. But Richie's absolutely right. We've invested a lot in training programs, our education, things like that, to bring people up to speed. But Bunsen really can be transformative.
It is similar. We see the exact same impact in our software developer organizations, where a junior developer comes in now and can actually start, with the right mentorship, can start doing projects that only a fifth-year could do before. I mean, the impact of Claude Code is tremendous in terms of the ability to accelerate people. We expect the, I mean, we are already seeing internally the exact same thing to be happening with computational chemists, where a junior person comes in and you might have only trusted them with this tiny task, but with Bunsen guiding them how to do it correctly, you can start introducing them to much more sophisticated tasks much quicker.
We also have insights from our therapeutics team, which would allow us to have confidence that not only the comp chemists, but also the medicinal chemists can use our platform and use Bunsen.
Computational chemists, medicinal chemists, decision-makers tend to be either in the C-suite or running the R&D programs. How do they understand what this is? Is the pitch to them, or is the pitch to the chemists? First to, like, with Bristol Myers, was it the C-suite who bought into this, or was it the R&D team, or was there a chemist who used this and said, "Guys, we need to have this"?
I would say it is all of those categories plus AI and digital transformation officers that are our customers today. You need all of those working together. You need the support of the actual users, but you need the R&D visionaries and the tech visionaries who understand the power of our platform.
Yeah, it was someone at NVIDIA actually said to me two years ago that every CEO is getting two questions from their investors right now, which is every CEO in biopharma, which is, "What's your pipeline look like, and what's your AI strategy?" There is definitely push from the top down to be forward-thinking in terms of AI. And we were really advantaged because our stance that AI was a little overhyped was pretty much completely agreed on inside industry. All of our users felt that in their hearts. So, we are in this position of not being viewed as an overhyped AI company, and now when the push comes for AI and we finally have an AI solution, they trust us. We're not an overhyping company.
We have both the incentives from the top and the people from the bottom saying, "Okay, let's do this." And I think that's actually, I've talked about how overwhelmingly positive the interest is as we show this to customers right now, and I think that's what's driving a lot of the positivity.
I think the NVIDIA C-suite's getting other questions today too, just as-
Yeah. Some other ones.
While Schrödinger is less focused on pursuing its own internal programs through late-stage development, the company is very active in partnering. You've obviously Bristol that we've talked about. Can you maybe, from a financial perspective, talk about the economics and how to think about It's hard to model from the outside world. Maybe talk a little bit about the economics, the revenue model of these engagements?
On the partnerships or-
On the partnerships, yeah.
Yeah, we're still very focused on using our platform to discover unique and proprietary therapeutics in the earlier stages of discovery, and then having the optionality to advance those programs either in partnership or through NewCos. We just announced a few months ago the formation of Tectora, which is a prime example of that. We contributed two programs into the company and raised $55 million and retained an equity stake in the company. In addition, I think what's really important is we not only formed the company, but Tectora is working with Schrödinger to continue advancing these molecules towards the clinic. So, from a P&L perspective, the investment we made into the programs as R&D, it's now transformed into an equity stake, and we will recognize revenue from Tectora for the service that we provide to the company on the programs that we're working on.
Should we start to think about your equity investments as being material to the value of the company in whole?
I would say not just equity investments, but also milestones and royalties. The way we go to market is, especially on the therapeutic side, is we do very intensive multiple-year collaborations where we not only earn revenue, but are generating IP alongside the customer, and we earn milestones and royalties as the program progresses down the clinic and into commercialization. So our equity portfolio has delivered substantial returns to the company since the IPO. I think that number is totaling close to $500 million now in distributions. We are very proud of the work we've done in collaboration with other companies and the benefit of working from a computational-first approach.
I have a question from the audience, and I want to get it in because it was more I missed it earlier. When and what do a ChatGPT look like for Schrödinger?
Sorry, what?
A ChatGPT?
Yeah. Because I mentioned it earlier. We talked about-
Sorry, what was the question?
Oh, I'm sorry. When and what do a ChatGPT look like for Schrödinger?
I mean-
What is it, like, visually?
Yeah, I mean, visually, you have your sessions on the left. We have a few things that are kind of unique to us. You want to monitor your isolated file system because there's a lot more IP concerns in our space. I should mention that we built this entire product knowing our customers' IT preferences and policies very well, which is something that a lot of the products built generically in the AI space don't fully understand. So, there's a lot of customizations and hooks in the backend that make R&D IT leaders just nod their head and say, "Okay, yeah, you can come in." But sessions on the left, file system on the right, a bunch of mini apps, but ultimately, you're talking with Bunsen. I believe we do have demos of it online, so you can see it if you're interested.
You'll say, "I'm looking at this target. Help me put together a virtual screening workflow." Before asking that question, might have been watching a one-hour video of how to do virtual screening in Schrödinger software platform, and now Bunsen says, "These are the four most important questions I have to ask you. If you don't understand them, let me know, and we can keep talking about it. Otherwise, answer them, and I will now go use your supercomputer to max capacity." That type of thing, I mean, it could have been a day or days of work for someone before, and now it is a conversation like that. Now, I don't want it to sound like this technology is absolutely turnkey at this point. It still often requires expert knowledge, and we are programming as much of that as we can.
Like, we actually task our therapeutics team to put their brain into Bunsen. It's not just an LLM repurposed for this. We're adding a ton of knowledge to it. There are places where Bunsen will still make mistakes, like every LLM. It is really important, we're calling it a co-scientist for a reason, that the scientists still really understand their space. But it's able to both guide you really well and often catch mistakes you were going to make. If you were going to kick a weekend process off your supercomputer that cost the company $100,000, and you forgot to set the right parameter for that type of protein, Bunsen is very likely to say, "Hey, I think this might be wrong, looking at the protein," and it just saved you $100,000.
I mean, that's the type of thing that we're super excited about that just nothing existed in the past, because in chemistry, there are so many gotchas. You're never going to hand code every single one into a UI. You need something with a little more sophistication.
Richie, I'm going to go back to the milestones question. Are these going to be, like, are these public that we'll know what your potential milestones are? Like, how much disclosure can you put around that? Because, hey, getting a $70 million milestone payment is material to the company. As analysts, how should we think about these, or how should we think about modeling them, or?
Yeah, it's a tough question. We've put out as much disclosure as we can for the terms of our collaboration agreements. We've summarized the total milestones and the general descriptions of the royalties, low single digit, mid single digit, high single digit. We've put out the disclosure that we can. Admittedly, they are not a milestone that we control directly in many cases, and it's hard for us to guide or forecast around them. But I think the track record is really what we would point to, is the persistent receipt of these over the past couple of years. As we continue building the portfolio and as the existing portfolio continues to mature, we expect that this will become more and more repeatable, more and more recurring. But we've put out the disclosure that we can around.
How important is it strategically or goals for the company to become profitable, turn EBITDA positive, cash flow positive? I know you've talked about phasing out some operating expenses potentially. Can you just talk a little bit about when you get together, is this a goal? Is it more around building the platform and getting the revenues and the partnerships, or is it profitability, and how do you, or maybe how do you think about both of those things together?
Yeah, I mean, profitability is obviously important, but we're going to be focused on driving innovation to our customers. By investing in the platform, investing in R&D, and growing the top line, that puts us on the path towards profitability. That is really the DNA and culture of the company, is investing in products and bringing new science to our customers. By doing that, we also can achieve profitability. So I'd say they're equally important to us. From a balance sheet point of view, we have the cash to get to profitability and look to continue making investments that are synergistic with the platform.
Got it. Okay. Is it fair to say that the transition from on-premises to hosted software is part of this overall strategy to become a software provider and partner?
Yeah, we started on that journey this year in a formal way. We've actually been building towards that for several years. But it is important to us for a couple different reasons. One is from a customer-facing point of view, we're better able to support the customer, better able to onboard the customer, and also able to send trial licenses and bring new products in and have them experiment with them. From a financial reporting point of view, this is the year that we are in the midst of the transition. It's obviously caused some noise with our accounting for the year, and in particular to software revenue, which is why we're focused on ACV.
But as we get through this transition and the hosted percentage increases to a place, we have targeted 75% by 2028, the revenue reporting of our software business will become a lot more clear and reflective of the actual fundamentals of the business.
Can you quantify what the software products and services growth rate would be this year if not for the transition?
Well, our ACV guidance is the clearest way to look at that.
Okay.
Our ACV is expected to increase 12%-17% this year off of $198 million.
If we were to break that down, how much of that? Are there any components of that, of the 12%-17%? Is it all one bolus or?
Where's the growth coming from?
Yeah.
What's been really nice year- to- date, and I guess for full year, we've had broad growth. We've grown within top 20 pharma customers, broadly across life science, material science. We've had growth from new products. We've talked about Bunsen, Predictive Tox. Crystal structure prediction is another area of growth for us. We've actually seen a better biotech market than we've seen the last couple years from a funding perspective.
How much is biotech versus pharma?
Large pharma is, call it, half of our commercial business.
Makes sense.
With top 20 pharma.
Yeah.
Outside of top 20, there's many large pharma customers that contribute to our overall commercial ACV. New customers has been stronger this year too. We've had some growth from new customers.
We spent a lot of time talking about Bunsen. It's obviously super exciting, but I'm guessing it's a very small part of the revenue drive or associated revenue?
Like we said, Bunsen for us is about throughput, it's about licenses.
I know, it's.
We're seeing license growth. And again, Bunsen is still in beta.
Right.
We have announced our transaction with Bristol Myers, but it is still in beta. I think we are pleased with what it has done on license growth, but some of that growth is still to come. Maybe Pat, you can talk about this, but customers are evaluating it right now. They are going through security reviews, but they have not necessarily transacted on Bunsen or increased license packages related to Bunsen.
I never thought to ask this before, and I can not imagine there is, but is there anything exclusive to the Bristol Myers agreement that others can not participate in, whether it be a clinical category or any kind of proprietary? You are shaking your head. I assume not, but for sometimes first customers-
It is more trying to think about what we can talk about and what we can not, but I do not think there is any problem saying no.
There is nothing that prevents you from selling-
No
-anything to another large customer that-
No.
Okay. Just wanted to make sure. You said biopharma's getting a little bit better. You've seen an uptick in that. How do you see the recent improvement in those end markets driving software adoption over the next? Is it a driver?
Yeah, I think broadly the industry is embracing a physics-first computational approach to drug discovery, and that is literally the definition of our platform. I think there's a lot of AI drug discovery discussion, biology, chemistry, targets. But I think the point is, if you are trying to design a molecule to drug a target and trying to optimize across the 20+ parameters you need to solve, our platform is the way to do that. I think there's a broader recognition across the customer base that this is the way to do drug discovery. I think our proof points across co-funded companies and the advancement of our co-discovered drugs through the clinic are demonstrating that. So, in general, from a user base level, there's a clear recognition of how to do this.
I think from a pharma R&D perspective, and even venture capital-based companies, you have to be efficient. In this environment, you have to be efficient. You have to be fast. You have to be able to be efficient with how you spend the dollars. Our platform enables you to do that.
Mm-hmm. When we think about the demand environment versus your execution, what is going to drive growth more? How dependent are you on the market? How dependent are we on your product, your ability to sell your product, and the effectiveness of your products?
Yeah, I mean, the market backdrop right now, it is a good backdrop. But we are bringing new products to the market as well. I think we expect to be a market plus grower, given the investments we make in innovation.
Are there any products that we should be looking forward to? Sounds like new products, you obviously have been developing a bunch of things, but is there any announcements we should be expecting? You do not have to give me specifics, but we are not asking for 2027 outlooks here, but it sounds like there are more to come.
Yeah, I mean, I think our investment in new product, it is not a destination and you are finished. It is continuous. The platform advancement, the accuracy of the platform, that will always be there. On Predictive Tox, we said on our call last week that we have now profiled 80 of the off targets. 80 of the off targets is impressive. There is still a lot more to go for it to be of more use to customers. Bunsen is in beta right now. We are continuing to evolve that product as well. And that is just a sampling of our new products, but for us, this is about continuing to advance the technology and bring in new products, which can drive long-term growth for us with customers.
Well, gentlemen, this has been great. Thank you so much for coming down to Florida, and thank you for the time. We really appreciate it.
All right. Thank you.