AppLovin Corporation (APP) Earnings Call Transcript & Summary

All right. Good morning, everyone. Thank you for being here. My name is Matt Cost, Morgan Stanley U.S. Internet team. Very happy this morning to be joined by Adam Foroughi and Matt Stumpf, the CEO and CFO of AppLovin. Thank you for being here, guys. On the Morgan Stanley disclosure side, please note that all important disclosures, including personal holdings disclosures and MS disclosures appear on the MS public website at morganstanley.com/researchdisclosures or at the registration desk.

All right. And with that out of the way, Matt, maybe let's start with you. I want to revisit the 20% to 30% growth target that you've talked a lot about over the past couple of years for gaming ads. You reiterated again at the fourth quarter. And obviously, you've exceeded it quite a bit over that whole time period. Has your thinking on that benchmark changed at all in terms of the baseline level of growth of the gaming ads business? And is there a potential upside to that number as we look out?

Yes. I mean there's definitely potential upside. Look, when we initially mentioned to investors in the public markets around the 20% to 30% we just wanted to frame for investors that there's a lot of opportunity there for continued growth with the core technology on the mobile gaming side of the business. We broke it down for investors very simply that between directed model enhancements and recursive learning that's happening on an ongoing basis, we should be able to get at least that 20% to 30%. At the time, I think when we initially said it a couple of years ago, we weren't even getting credit for 20%. I think people didn't think we could grow at all. So it was just a baseline and we wanted to kind of level set with investors. And look, over that period of time, over the past couple of years to your point, we've grown at a pace much greater than that. What investors need to understand is that this technology is very nascent. So our engineers are continuing to come up with directed model enhancements to make the technology better. And as we grow and scale, what that means for the technology is that we're getting more data to feed into it, which then improves the technology over time and continues to stack and compound and as we expand out on things like e-commerce, that will continue to grow the technology's ability to scale and drive spend at the return goals of our advertisers.

You mentioned e-commerce and web. Maybe I'll go to you, Adam, on that one. So on the last earnings call, you talked about expecting a faster ramp for web advertisers compared to gaming. I guess let's talk about what the drivers of that faster ramp are, the scale of the opportunity and then how that business is progressing as you move closer to the GA launch in the first half of this year?

Yes. I felt like with this question, it would be good to just talk through how do we do so well in building products in the markets that we operate in because there's always a question with our business of how do you outcompete everyone in this niche that you become so excellent. And over 13 years, we really understood with game developers what they needed from a marketing platform. The game development market is much easier to tackle at least for us, it was, then the e-commerce or these incremental markets because it was more niche, the developer mix was smaller and we were able to tackle it over a long time, really understanding that their biggest issue was how do you spend $1 and ensure you make more. These are not VC-backed companies, they had to know that they were going to get a profit on the marketing. In the absence of that, our business never would have scaled. So over the 13 years, we became excellent at building a recommendation system model to automate that problem for them and we do it today at larger scale than anyone else in the world. We became the #1 destination for game developers that required us to know our market. Now as we've gone outside of gaming, this is a new market for us. We got into e-commerce, we talk to clients, we try to understand the complexities of the market. It's a much more complex market because you've got two parts to -- maybe three parts to e-commerce businesses. They try to find new customers on discovery platforms. Meta is a fantastic example of this. We are trying to be another very large example of this. But they also mix in search advertising, soon there'll be large language model advertising typically called bottom of funnel. At the end of a conversion funnel, you run an ad, close the gap to conversion. And then they've also got CRM. They get a customer and they can convert their own customer. So when you've got a more complex mix like that, you've got to know what you're building towards. And when we got into the market, we rolled out one type of targeting in our system. Did this 1.5 years ago. We took our expertise in building a recommendation system and we were able to use the 1 billion-plus daily active users we add in games to convert them to new e-commerce brands that came on to the platform and ramped it really quickly. Now what we didn't talk about was what that first product was. It was -- we call it a universal campaign. But that product is a mixture of discovery for these companies and retargeting. Now when you talk to the customers, they all want incremental value, first, retargeting, second. We started with the first type of product offering that over the last 1.5 years, we built out more. A few months ago, and I think it was late October, we rolled out new customer campaigns. What this targeting did, all model-based, is have a model, be able to find them new customers that never bought on their site before, but might have visited their site. So if you think about what our offering is, it's a full funnel offering, have customers discover the product, but price it all the way to the point of transaction for them. We started with something that was further down funnel. We moved to the middle of the funnel. Now last week, we rolled out new visitor campaigns. This is super important because now we're able to drive a customer to their site that they've never seen before. That's the most powerful form of advertising for anyone. No one can debate how incremental something is if they find a customer they've never seen on their website before. We're able to do it really effectively and then pilot on that product, we got great results. We rolled it out a week ago with the product blog and adoption has been really swift. Now the reason I cover all this is we think about building product over time, and it's an iterative process. It requires us to really understand the clients that we have on the other side, deliver products for them that give them huge value from the marketing, profitable marketing that's measurable. If we can do that, we're able to scale our product out. We take our time developing the products. We're not rushed. As I think investors have seen, we really operate as a private company in the public markets when it comes to product development and product road maps because we're building for 5, 10 years from now. Now when I talk long term, it does not mean I'm not excited about the short term. In the short term, we're seeing fantastic metrics in an addressable market that might be 5 to 10x the size of the gaming market. As we go and get to a point where we open up our platform, it's not lost on us that now we're going to be able to make a much bigger impact on the overall economy and on our business and for a much bigger set of advertisers. Something that gets us really excited about. But I do want to remind investors that my role here is to build the biggest company possible 10 years from now, and I direct my team to think that way. We try to really understand our clients. We try to build great technology for the market that we operate in. We have one of the world's most powerful recommendation models. We've got amazing data inside that model. We're able to, at very large scale, #1 in the market, be able to provide value that's immense for the game developers, we're now starting to really ramp outside, and we think that's only going to accelerate as we get further.

Got it. And maybe building off of that, and you're just talking about the long term. I think when I talk to investors about your web business, I get a lot of questions about the short term. So people are asking about the number of customers being added, about sequential growth, dollar revenue and I think you pushed back a little bit on the earnings call, in particular, in that way of thinking about the business. I want to ask you, what should the market be focusing on? How should they be defining success? And what are the milestones they should be looking out for, for this business?

Yes, it's funny because I push back on short term for the reason I just gave because we think long term. But more importantly, the platform itself, the goal we have is to improve how well we monetize the impressions that we serve. We're serving over 1 billion users a day. It's a big audience. The ads are very much attention grabbing ads. Our ads, on average, are watched over 30 seconds. Roughly half the ads we serve, the user is actually opting in to watch an ad, that's called the rewarded video. Nowhere in the world are you going to find an ad that someone's asking to watch, except on our platform at the scale that we operate. Now what that means is we've got this opportunity to really expand the business if we execute as we go forward into these new categories. And let me break that down for you. We, today, have a 1.3% conversion rate on our ads that we serve. So said differently, we make money on 1.3% of the ads we serve. We lose money on 98.7% of the ads we serve. Now when the model knows a user is really primed for gaming, they're likely to churn the current game, they're likely to go to another game. We have well over a 5% conversion rate. This is what I call a high value moment. Most of the impressions that the model serves for gaming are not high-value moments. The reason they're not is that the user is playing a game that they like. And if you serve them 100 gaming ads in a row, at some point, that becomes annoying. Well, a very powerful recommendation system is meant to personalized content. But in the absence of content diversity, it has to stick to what it knows, which, in our case, originally was just gaming. Now it's gaming plus. So we've got the e-commerce brands, and we've got some other lead gen brands, but not a lot of them yet, not a lot of density in every single category. Now fast forward 5 years. Let's say, we have hundreds of thousands of customers, which I fully believe we'll get to. What's going to happen in our platform? Well, this powerful technology, which inevitably is going to get much better over time, is going to be able to serve a different product in every single ad impression. It's going to be able to take the 1,000 impressions that it serves on a unit. And when it believes that gaming is a good ad, nothing else will be gaming in that moment. And if the conversion rate of that is over 5%, you can think of some of the impressions that will be stuck on gaming, that will drive a lot of value for the gaming customers. They won't face cannibalization. But on the rest of the impressions, the model will get more precise, more personalized and drive a higher conversion rate. Our effective yield will go up materially because of it. And we fully expect that we're able to bring density to the auction, we'll get over a 5% conversion rate. Now we're starting at 1.3%. So just apply a multiplier on that. That doesn't necessarily mean that in our business, that's a 4x to the business because you also then have to understand the dynamics of how we operate. We live in an auction where we're paying out some portion, majority of the revenue that we generate goes out to publishers and I should say, majority of the spend that we generate goes out to publishers. So we pay for their ad space. And then we have spend on top of that, and the spread is what we report as a public company is revenue. Now if we 4x that. And as an example, last year, we gave $11 billion run rate in our business in Q1 last year of advertiser spend. We're materially bigger than that. As you all know, we've grown. And to conceptualize just how big, bigger than the sum of everything that's happening on Snap, Pinterest, Twitter, Reddit combined falls into the advertiser spend on our platform. Now if you take that number and remember, we're a 1.3% conversion rate there. What happens when you get over 5%? It becomes one of the largest advertising platforms around. We'd be able to generate tens of billions more of ad spend on our platform than we do today. The expansion opportunity, both to the economy, GDP, job production, everything from that is huge, gets us very, very excited. But the other thing that happens there is as our conversion rate goes up, we're not a fixed share to publishers. We end up in an auction price dynamic, possibly paying the same amount out to publishers in which case you'd say the business would grow, it would quadruple, possibly being able to take even a larger spread because we're accelerating so much ahead of peers in the marketplace, would you then say the business is going to more than quadruple. That's the growth opportunity in front of us. It requires us to get customers in and it requires us to improve our technology. We're super excited we can do all of that over the coming years.

Let's talk, Matt, about the investments necessary to improve the platform and the technology, obviously critical to the vision that Adam just laid out. What does that mean for the company in terms of investment and particularly head count growth? I mean I think you're notable as a company that's been very prudent and aggressive in managing head count through a tremendous period of growth. So how are you thinking about that going forward?

Yes. I mean when you think about investment, really the kind of core components of our cost structure that are important to understand are data center costs. So we've mentioned previously that what we've seen over the long term is that those costs have grown at about 10% of the overall revenue growth, really, in fact, we're beating that now. And that is just really, the efficient way that we run very disciplined as our team is launching new models, like Adam mentioned before, the prospecting model or making improvements. They're looking at the cost impact of that and making sure that those model changes they're making are profitable. And real time, because we run so lean, they're able to do that and really very closely monitor the cost impact of the changes that they're making. So we don't think that, that should change going forward. Then the other component you mentioned is head count, right? So I think we have, in total, around 900 employees at the company. But really, when you drill into what that is, about 400 only of the 900 are associated with the core ad tech business and all of our back office. So we already run extremely lean. We've got a business development team for the core mobile gaming side of the business, that's like 100 people. I think we have something around 15 people for e-comm. So very, very small team. So as we think about investment to grow to support some of these initiatives like the web-based advertising e-com, we'll definitely add headcount, but we're talking tens of people. So there really isn't any impact on the overall cost structure. The last category that we've looked at is -- and we mentioned, I think, on the earnings call as well is performance marketing. So now that we've got the product in a place that's very good. We're adding advertisers. We want to accelerate that. So we've started focusing on performance marketing to bring in new advertisers. And that is just as you would think, I mean, in our business model as well. It's running campaigns to drive users into the platform, but we're doing that similarly in a very disciplined manner. So we don't think that any of that should change the cost profile going forward either.

I do want to add a point too here for having a CFO that watches every penny we spend, Matt loves that our EBITDA margins are where they are. But you could take what I just said and you hear massive revenue opportunity, massive growth opportunities sitting right in front of this company. And he's saying, "Go get customers." Well, you can then follow that up with, "what are they nuts? Why aren't they hiring a bunch of business development people, ramping a sales force and getting out there and doing it?" and I get this question all the time. And there's a couple of reasons why we don't aggressively approach it with head count. One is there's a whole bunch of platforms that ramped up sales, brought in new customers and hit a wall and never grew their business. And why did they hit that wall? Because the product itself was not good enough to provably drive value for the customer, where they could scale in a way that they knew was profitable. And if they have that in front of them, they don't need to be sold, the advertisers are exceptionally smart in performance marketing. They know what's going on, on the other side of the dollars that they spend. We didn't scale gaming by begging for dollars, we scaled gaming by showing them that they make more money with their dollars spent on our platform than anywhere else in the world. So by constraining team, we're not front-running an opportunity. We're forcing ourselves to build the world's best product for the customer on the other side inside our space. If we can do that, the scale of the business is going to come. We're not rushed to go get there because we think it's going to organically come through word-of-mouth reality because our product is going to end up so good that these advertisers are just going to come into the platform and ramp on their own. We will pair that growth in product improvement, that growth from customers coming on with headcount, but never with that many. The other reason we constrain it is because we live in this world where we now have large language models to make people more efficient. A couple of years ago, we went through and with the EBITDA margins we have and the growth that we've had over the last couple of years, we cut roughly 40% of the headcount. It seemed insane to people at the time. Now why did we do that? We did it because we wanted to make sure the people that were at the company were the people that could take these tools and make themselves more effective. We've seen a 10x engineer become 100x in terms of output because of large language models that they pair with. Now not every engineer is created equal. 1x might become 2x because if they're 1x, they're not as capable of using the tools. They're still more effective. But then you have to ask, do you want 50 1x engineers around or do you want 5 10x engineers around and so what we decided is that we want a team of really high-caliber people who are going to learn how to use the tools, make themselves more productive so that they're not automated away. Across the organization, we made that the focus. We leaned up to those people, and that makes us really effective in the world we live in today.

Okay. I guess thinking about competition, Adam, you've pretty consistently framed competition as something that can expand the overall ad opportunity rather than compress it. As web advertising becomes a bigger part of the business, how should investors think about your differentiation versus platforms like Meta or Google? And what are some misunderstandings perhaps about the way you interact with those companies?

Yes. I think when we started the business, we were 1 of the bigger VC misses because we couldn't raise $1 million over 4 or 13 years ago. And the constant feedback I got is competition is going to eat you alive. 13 years later, we've been competing in some very big platforms, and we obviously have a great business. And people -- it's easy to believe that we have disadvantages to the largest platforms, therefore, we should lose. That's the easy answer. No one ever asks why are you winning? How do you do what you do? Well, let me break that down a little bit. One is, as I talked about, we really know our customer. If you talk about inside games, it's not like we're now competing with the largest companies in e-commerce. We've been competing inside games with the largest companies out in the world for a very long time and done it well and become the leader. Inside games and inside e-commerce in every category we tackle, we want to understand our customer, and we want to solve their problems in our space. Our space is a niche. It turns out this niche is pretty big. It's a billion-plus daily active users on our platform. It's still a niche. It's different than the other access points for the consumer. The ad formats are different, the experience is different, the type of user is different. This is not your high-frequency user of social media properties. This -- the person who's playing a Mahjong or a Candy Crush, is a different type of user. Fortunately, they're adults and they're great shoppers. And so you've got this great audience. You have this differentiated ad format. You've got a different reason why they're here. People who play games are doing it to relax and they're doing it to get psychological benefit in that moment. That's our framework. That gives us a very good framework. Now we went in and we built technology to execute in this space. I think there's -- and maybe in the investor community or people that don't understand the complexities of these models, an oversimplification on the problem set that we're solving. If our job was just to place an ad on behalf of the advertiser, a single prediction, analogy would be in a large language model that just had to predict the first word. Well, that's a pretty simple thing to do. If you just place the ad, you would lose a ton of money for yourself as a platform and the advertisers that exist on your platform. What are we actually doing? Inside gaming, for instance, we predict revenue all the way out to 28 days. If you back up what that means. These are companies that probably run 20%, 30% margin. Our model has to be precise that far out into the future, predicting revenue from a customer that didn't know they wanted this game to begin with. You have to predict engagement with the ad, you have to predict that the download is going to happen, you have to predict usage patterns, you have to predict propensity to spend, you have to predict amount of spend. You can't get any of the sequence of predictions wrong to deliver the value that you're trying to deliver for the advertiser. Now that's a really deep problem set to go solve. It took a lot of technology. There's a belief today, I think, in AI, and I don't tend to throw the two letters around too much. That large language models are the model that wins. Recommendation system models are one of the best commercial uses of deep learning models. And a lot of the same architecture and technologies apply, we've just built one of the most sophisticated ones to solve a very long list sequence of problem set. Now when we do that, and we scale our business, it gives us fantastic data inside that model that no one else has access to, to make our model better. The model itself gets smarter as it looks at that data, and so not only were we able to release something that was cutting edge in our space, we were able to solve very large problems for our advertiser set both in gaming and now moving into e-commerce, again, in this space. But our model is getting smarter every single day. This ad that serves our data that no one else has access to, the conversion funnels it builds and sees our data that no one else has access to. And that differentiation is very large when you think about modeling verticals. Some verticals are predicated on same data sets, different machine learning techniques that can create a ton of value as we've seen. Our verticals predicated on different data set and differentiated machine learning techniques and our models are exceptional at solving the problems that we do.

I know you don't like to throw AI around, but let's stick with that for one second. It's a major topic, obviously, but particularly in the past month or two for the video game industry. And I think we've seen some of those concerns spill over into the narrative about AppLovin. So I wanted to ask a fairly narrow question, frankly, which is what would it mean for AppLovin if there were many more games being created by many more people using AI and what if some of the incumbent game companies were disrupted, what would it mean for you?

Yes. I mean, look, the count of content is not really a KPI that matters because you can build a bunch of slop and it doesn't do anything. As we've seen internally, and I mentioned a second ago, our 10x engineers became 100x, our 1x engineers might become 2x, and we're mostly built around the 10x engineers. The efficiency gains are really much more magnified when you're sophisticated of what you do. So rather than think about a child or an adult who doesn't know how to code or is not a game developer, vibe coding a new game and somehow thinking they have a good experience. The application is much more likely to make the current game developers who are very, very sophisticated in what they do, get better. What does that mean? Content, development inside their current games is going to get cheaper. That's going to allow them to continue to expand their LTV. As their LTV goes up, their marketing dollars go up. That's very beneficial for a platform like ours and the consumer experience expands from that. Those same game developers are almost certainly going to be the winners when it comes to new game development as well. As the cost of content goes down, that count of high-quality games from those game developers goes up. the category is going to be able to use this technology to get to another inflection point of growth. Now you may also have someone like my 16-year-old envision a game, go write it in natural language and create something that's cool. And it's not slop. What happens at that point? No one in the world is going to play that game, unless that game gets discovered. And how does content get discovered today? The best form of discovery is natural to believe is search and large language models. That is absolutely one. But the other best form of discovery is through display advertisements that allow a consumer to take a break look at content and see if they want to go engage with that new content. And when we've got 30 seconds plus to show new content to the consumer on our platform, and we've got this powerful recommendation engine behind it, all of that new content is going to have to come to our platform to get discovered.

Got it. Let's talk about generating creative. I think it's something you've been piloting recently in terms of creating ad units or rather creating ad creative for some of your advertiser customers. So what feedback have you been getting from the people who are piloting that technology? And what performance or level of impact are you looking to achieve before rolling that out to a much broader group?

Yes. First of all, stating the problem. The problem we've seen is that game developers are extremely high-frequency traders of marketing platforms. The game developers that spend the most on our platform, spend a lot, but they've got over 50,000 ads in a single campaign. The new category, e-commerce and then onward, at best, we saw 1,000 ads in a campaign. At the high end of spend. Now if you think about that differentiation, our model is built to go test ads programmatically and find expansion of conversion rate and return on ad spend and more spend if there's more diversity of that. That is the most important variable that the marketer has at their disposal, but you've got a 50x differential. The e-commerce companies are not going to be able to spin up creative resources and production costs to go 50x their creative output on our platform. So how do you solve that? Unfortunately, the large language models have gotten really good at image generation and video generation. It's not so good that out of the box, you can solve it for a brand. If it was that simple, the brands were already up 50,000 ads. So we had to bring to market multi-agent approach on top of the large language models to go in and figure out how to create content, both in static images and video, and it satisfies the brand's needs. We rolled this out and pilot on the static part of the ad. Our ads go from a video to a follow-up to the video that think of it as like a 1-page animated GIF format. And inside that, we're already in pilot. We're seeing a lot of interesting success there with the customers that are adopting it because it allows them to have a lot more velocity. The video model is on the way as well, and people try to ask, "When is that going to come? When are you going to roll it out?" Well, I talked about when we go to general release of our platform, which we're still in a closed state on the platform we fully expect to be able to give tools to the customers to create ads for our platform. And I lock that data in this first half of this year. So if you think about first half of this year, we're 4 months away from that at the end. Some point in the next 4 months, we will have both these types of problems solved regenerative AI-based creative. Once that happens, the front end of the experience for the customer will be more personalized. Only going to get better over time. But as a starting point, it's going to be much better than where we are today, which is a 50x handicap to the gaming developers who are super sophisticated. As we see that, we fully expect that these customers will adopt it, they'll get a much higher conversion rate of their ad to user engaging with their product because of the diversity of content at the front end. That will drive up their spend, that will drive out their return on ad spend and should be a material unlock for us.

Great. Maybe turning to mediation. So there's been some news recently about a new player in the mediation market. I think a lot of investors have interpreted that as a competitive threat to MAX, which is your remediation product. So what attributes does MAX have that you expect to help it maintain its leadership? And how do you think that competitive ecosystem will change going forward, if at all?

Yes. Mediation is not a really well-understood technology or concept. So breaking it down a little bit, the market and mediation we got into in -- I think it was 2018 when we launched MAX. And we were competing at the time with Google's mediation layer and IronSource, which is now Unity's mediation layer and a few other ones. So this space has always been full of competition. The mediation plays two roles. One is this technology is meant to allow a publisher to serve ads but gain access to all the demand in the marketplace. We're a very large demand source for the publishers. As we all know, we're maybe the largest. However, there's a lot of other ones. There's Facebook, there's Google, there's Unity and if go down the last, 15 to 20 to maybe 100. and they get really small at the end, but at the head, there's a lot of diversity there. The tool has to give a completely fair unbiased auction approach to the publisher to get the best ad from the highest paying network on every instance to be useful. We built MAX completely unbiased, completely transparent on data to the partners that we have, fully audited solution. And when we brought it to market, we were 10 years later than Admob's solutions for publishers. We were maybe 8 to 9 years later than IronSource's solution from publishers. We were also competing with MoPub at the time. MAX went from 0 to probably 1/3 of the market in 2 to 3 years. This was before we had our demand-side platform strength. So if you just then go grade the technology that we built, we then bought MoPub, took over more of the supply in the space. Threw MoPub's technology out, replace it with ours. Publishers came over. They had every chance to go to any other platform back then. They all came to ours. The reason was is because the technology gave them the most yield. Now then fast forward to today, what's happened since then? We now not only have the highest monetizing in most dense auction inside the MAX auction, we paired it with the best buying tools. The vast majority of every publisher spend comes on our platform. Now this isn't the advertiser that is the in-app purchasing game. This is a publisher that's running ads inside their game. In order for that publisher to grow, they better be able to buy ads. We're the best destination for them. We built the best tools. It's a really big number in terms of percentage of their spend. So not only are we the best monetizing and most dense and offset competition every step of the way to get to that point. We give them the best growth tools as well. Long way of saying, our tools are really sticky. We've not been in a market that didn't have competition. So then if you go, "Okay, well, what's going to happen as we go forward?" There's inevitably going to be more competition. We live in an agentic world where it's really cheap to start products, it's really cheap to build tools. Let's not forget that we're also very good at using the same tools. We don't look at product innovation that comes from competition or within us as something that's static. We're always innovating our own products as well. So if there are breakthroughs in the space, and we have the best platform on both sides, that's already locked in, those breakthroughs will help us make our products even better for our customers. So we look at competition as inspiration, but we know that the strengths of our platform make it completely locked in because these companies that are on the other side of it depend on us for their growth.

Got it. Matt, maybe I'll go back to you and then we'll close Adam with sort of a big picture question. But before we do, Matt, there was an interview last month with your Chief product officer talking about some experiments the company is doing in social media. My understanding is that this is more of a -- less of a key strategic priority and perhaps more of like an other bet. Do I have that right? And how should investors think about some of the other smaller projects you're working on?

Yes. I mean there's a lot of talk about this question because it came out, I think, in an interview. And look, like we're always testing new things like this. We're just running small projects to see where there might be opportunity for us in the future. For us, it's less of like something that's core, so your characterization is correct. It's really an opportunity for us to bring in new talent that's differentiated. So something that's outside of our core kind of talent pool, bringing in new ideas and for us, I mean, obviously, we run very lean. So none of these types of other bets are bets that we're running from an R&D perspective really will materially change the cost profile of the business. We're going to keep them very, very small and tight, but we think about it more as an opportunity to bring in new talent that we can then cross-pollinate ideas to the other components of the business.

Got it. Great. So Adam, maybe to close. So obviously, a lot of AI talk today. I guess what would you highlight as the most underappreciated opportunity available to AppLovin in the conversations you're having with investors? And then maybe a challenge that you think is worth pointing out is something that you're going to execute through?

Yes. Look, we're in -- we're building models in a recommendation system. The structure varies similarly to the path of the large language models. And I think everyone expects the technology to be static for whatever reason. We're not out there boasting marketing terms like AGI, but recommendation systems are going to evolve on the same trajectory as the large language models. Not only are we developing with similar architecture. We're developing paired with the large language models to accelerate rate of development. If we believe that AI technologies are going to be 2x more efficacious in 5 years, just based off of that, if we do our job right, our system is going to be 2x more predictive for its task, the sequence of problems that it's predicting in 5 years. That would double our business or more. And so I think that's not particularly well understood. People really latch on to large language models, obviously, because they can interact with them and less so latch on to the strength of what these recommendation systems are going to be able to do as they evolve. The challenge we have is also tied to the opportunity paired with the technology. As we go get more customers as we open up our platform, we're going to really be able to expand this business as we talked about. Every new customer is more data and every new customer is more demand. Our data moat is already growing with the technology and the ads we serve and the more customers we get and the quicker we do that, that data moat will even expand more. And so our job is to do that. The challenge is we're clearly not very good at marketing ourselves. I named the company AppLovin. That's a handicap. Nobody knows about the business. And we've got to put ourselves out there. We've got one of the best solutions for companies in the world to market themselves. They got to find out about it. It's our job to make sure that happens. We do that right our customers grow to over 100,000, and I think it will be in the many hundreds of thousands over the next 5 to 10 years or more. And we improve our technology at the rate or ahead of the rate of improvement in where these AI technologies are going to go, "This is going to be a much, much bigger business in the future."

All right. Adam, Matt, thank you for being here.

Thanks, Matt.

Thanks for having us.