All right, we are live. Good morning, everyone. Thank you all for being here. My name is Matt Cost, Morgan Stanley U.S. Internet team. Thrilled this morning to be joined by Adam Foroughi, Matt Stumpf, CEO and CFO of AppLovin. Thank you so much for being here.
Thanks for doing this with us.
Awesome. Let's start on gaming. And then maybe it'll be the last time we talk about gaming advertising this whole conversation. But there's lots to talk about outside of it. But the vast majority of your business is in the gaming vertical. Obviously, you've been incredibly successful there for a number of years now. Talk about the durability of growth there, and then how the competitive environment is changing. And how do you feel about that 20%-30% growth number that you've talked about historically?
Yeah, so when you're looking at the mobile gaming growth, first, try to break it up into the various components. You have to understand that there's supply that's also growing, which I think a lot of people don't view that as a significant component. But it's very important to the overall dynamics of the mobile gaming growth. So the supply we've talked about that obviously the MAX Marketplace drives is growing on a double-digit basis year over year. So that creates a large opportunity for the overall market, right? All of the demand-side platforms that feed into the mobile gaming ecosystem have that as a base, right?
Then when you look at, on the demand side, Axon advertising, which is driven by the Axon model, the model continues to improve over time, which we've mentioned is driven by the technology improvements, which are the ongoing reinforcement learning, which is just a continuous driver of ongoing improvement of the Axon model coming from the data, the transactional information that we're getting from serving impressions, and then feeding back into the prediction model to make it better and better over time.
Then you have directed model enhancements, which we've mentioned to the public in the past that those are coming at least one per year, driving at least 10% growth. We've simplified that for investors. But really what's happening behind the scenes is that that growth is made up of various model enhancements done by our engineering team, dozens really of experiments that the team is doing.
And those add a lift over time as well. And those stack. And so when you look at the 20%-30%, we think that there's still that it's very, very robust in terms of growth because that technology is still relatively nascent. So we have plenty of opportunity for continued ongoing learning, as well as those directed model enhancements, because the team has this very long list of potential experiments to continue to drive growth.
Got it. Let's shift to Axon and the self-serve launch. That happened on October 1. Very excited, long-anticipated development. So I guess, how are you defining success for self-serve, for web? I mean, is it number of advertisers? Is it ROAS? Is it wallet share? And then what are you looking at internally to kind of measure progress?
Yeah, so with a new product like this, really it's not the number of customers to us. That's not how we're viewing success. With this new self-serve platform, how we view success is whether or not the platform is performing well. So that is whether or not advertisers that are coming onto the platform are able to see the same return on ad spend that's consistent with those return goals that they're putting onto other platforms as well. Are they able to hit those same goals? And then are they able to hit those goals at a scale that is consistent with what they're doing on other channels?
Because that provides them the opportunity for them, the incentive for them to move over and open up a new channel of advertising spend. And so we've seen that both with that pilot group of e-commerce advertisers.
So we added in Q4 of last year. And then similarly, this new cohort that we've started adding on through the referral program in October.
Got it. Maybe going to Adam. What sort of performance are you seeing from new versus existing advertisers in e-commerce as we kind of move through the holidays? And then you talked a lot recently about prospecting campaigns. It came up on the earnings call and at some other events. I guess, tell us a little bit about that product and what it's allowing advertisers to do that they couldn't do before.
Yeah, so I've talked about this in another chat. But what's cool for me is we're probably the world's largest startup because we're bringing this product to market that we just didn't have before. We've got this billion-plus users. And we've gotten really good at serving them gaming ads. In fact, we're the best in the world at driving value inside gaming. But these users don't want to see game, game, game, game, game inside this space all the time when they're engaged in the game they're playing.
And so the idea around this originally was build a product, bring it to market, and be able to drive value for companies outside of the gaming space to this audience that they weren't accessing before. Now we're the largest accessing this audience. The mobile casual gamer is an older audience.
This isn't the audience that's spending all day long, six, seven hours a day on social. And so this is an opportunity for a brand to be able to access a consumer that could be very well suited to them at a different moment in time than they're accessing them if they're buying on Meta and other channels. The other compelling factor for us here was we have ad space that we're using with full-screen video advertisements. And the world has gotten really high on ad opportunities. But in our domain inside gaming, the user has to watch the ad. They get a full-screen video. It's followed up by upsell screens. They can't do anything else but watch the ad.
So when we developed this product, we were really excited about the prospects. If the model could actually price it and target it effectively, we could scale here.
And we brought together this cohort that we took live last June. And we really have been developing against over the last year and a half. What's important for us when building a product is not to contaminate the results with new customer growth. We know that we're going to get new customers. And we're going to get a lot of new customers coming onto the platform if the product works really well. So how do you define product working really well? You take a fixed customer cohort that's diverse. We took companies that were 30 million GMV and up. And we really focused on this mid-market e-commerce business across a variety of categories.
And we said, can we deliver them a product that works?
About a year ago, I disclosed that it took us only a matter of months to scale 600 some odd advertisers up to a $1 billion run rate spend. That's a huge conversion rate of dollars to customer base. And we did that by delivering a very good product. However, I also said we weren't great in terms of working for every single type of company. The ones that we weren't working for at scale, and what we define working as is versus the market leader, are we delivering comparable ROAS as the advertiser measures it? And are we delivering comparable scale? If those two things are truths, we know we've got a big business opportunity in front of us.
What we realized is that a company selling a product that's new to the market but doesn't have repeat buyers, so think of one-off purchase, something like a fire extinguisher or a hearing aid. You don't need to buy three hearing aids in the next six months. You only need one. Those companies were scaling really large on our platform. But the ones that weren't scaling were actually the bigger brands, the ones with brand loyalty. When a customer has an apparel shop or a beauty shop, they have a pretty big audience that's going to buy their product on a recurring basis. You probably visit the beauty shop once a month, and you're making a transaction.
Where our first delivery in this market didn't allow the customer to differentiate new customer acquisition versus retargeting, retargeting being you're taking their own current customer base and you're serving them an ad and driving a transaction. Well, if you're a brand owner, the question you would ask, let's say you're selling shirts to the world, is that if we drove a sale of a shirt for an existing customer, yes, they get excited that the sale was driven. But then they question, well, would that sale actually have happened next week? Because the customer is loyal, and they were going to buy it already.
So therefore, we shouldn't give AppLovin credit for the full sale. You're going to ask that question. We don't have a way to answer that question. They don't have a scientific means to answer that question.
So they'll run incrementality studies, and they'll be slow to ramp up. Well, the way to address that for us was to take our time and figure out how to build a model that can allow them to deliver value on prospecting. Prospecting is to tell that same shirt seller you can only get new customers from us as well as the model can predict and drive to new customers. This requires them to do a bit more of a data integration with us, which is fairly seamless now with where we've gotten to with integrations with Shopify and other attribution vendors. And once they pass us the data, now the model can really drive to new customer acquisition.
And what we saw when we put this into pilot in October is that the results were universally phenomenal.
You had customers where they were getting, as an example, 40% new customers, 60% retargeting. Click a button in our dashboard, instantly go to 80% new customer rate. So if you end up doing industry checks and talking to a customer that's buying on our platform in this method, they'll say the results are great. They do measure up really well against the market leader. Why that's exciting for us is that if you take that disclosure that we made with the 600 advertisers, billion-dollar spend, and understand the biggest customers in that cohort were not maximized in spend potential, and that this prospecting tool will maximize their spend on our platform as we go forward,
it really makes our ability to extract dollars from customer base even larger than what we put out there.
Which then, as you say, we have an opportunity in front of us to go acquire hundreds of thousands of customers, makes the potential future opportunity even bigger than it was before.
Great. A lot in there. I want to change gears, though, before we revisit some of those points and just talk about AI-based creative, because that's something I think really important. You've talked about a lot in the past. What is going to change about how creative is deployed, how it works in your platform, and what's available today versus over the next 12-24 months?
Yeah, this gets us excited as an advertising platform because you've got these categories on our platform. Gaming, for 13 years, these game customers have gotten really good at building creative for our platform. In fact, today, they build for our platform first, given we're the largest in the world driving mobile gaming advertising dollars. So the gaming companies have optimized creative. They know what works on our platform. The Generative AI opportunity for ad creative is not all that valuable there. But where it's valuable is we're bringing on these new customers.
Whether it's e-commerce today or tomorrow, it's a small business, a local laundromat, or restaurant, they don't make creatives optimized to our platform. Our ad unit is as long as 60 seconds on video. It's followed up by an upsell screen.
And it's followed up by a catalog ad that shows their products run through our recommendation engine to the user. The last screen's built. But the first two screens require a lot of work to make optimized to our platform. We also serve a lot of ads per day to the users. We serve a lot of ads over time to the users. So there's a lot of frequency. The best customers know you have to deliver a lot of ads into our platform, let's call it 50 a week, to deliver the highest result for each individual customer. Today, if they don't have the resources to build those ads, they just don't build them. They bring the ads that they're running on other channels, put them on our platform, and hope for the best.
Since the ad unit is unique and new and really high impact, if the brand understood they can utilize as long as 60 seconds, the way to unlock value here is give them generative AI-based tools to get creatives built in an automatic way and in a much cheaper way. This requires us not to be the only solution provider that gets there. The large language models that are generating videos and images have to also improve. Now, the good news is in the image generation space, the large language models are really pretty good at this point. Internally, we've written a model on top of the most popular image generation model that's out in the world.
And we're now creating part of the ad, the static part, on behalf of our customers in bulk using these tools. That's going to release to the customer pretty shortly here.
The video generation side is a little bit further behind. It's a much more complex undertaking. If you've played with Sora 2 and Veo 3, they have a limit where if you type in a box and you get an output, they have a limited time that they're going to give you a video for. So call it, let's say, 15 seconds of high quality. If our ads are 60 seconds max and we want more time, and the first version of an output off that model in 15 seconds isn't even good enough, you can imagine we have to write a model on top of those models that keeps re-prompting the LLM to get enough cuts that are high quality enough where we could automatically stitch together a video and deliver that to the advertiser.
We're at the point where, again, internally, we have tools to do this.
and this isn't that it's at the point where the video ad is great. It's at the point where the video ad is good enough for our platform to be better than what they have today. So their result can improve dramatically. Because we have to keep reprompting the LLM automatically to get an output that's satisfactory, the cost is not very low. The image generation model today for us to get a good image is about $1. A video output to get a really good video with all these prompts happening is about $100-$200.
Now, if we opened up this tool to all of our advertisers and said, you can create unlimited videos, well, they start creating 1,000 a week, and we're racking up a $100,000 bill per advertiser. It's not a great economic model.
So what we'll probably do is very shortly here roll out the image generation model to customers. Those cost $1. That's very scalable. That's going to be a pretty sizable uplift in the e-commerce and broader categories that come onto our platform for those customers. And we'll probably give them a quota of videos that they can go create. Now, take that to the current new advertiser flow on our platform. When a customer comes in, signs up for our platform, they need ads for our platform. If the local laundromat signed up, they probably don't have ads for our platform. We can automatically take their site that's uploaded to our system and generate them their first couple of ads.
And that's a very powerful thing.
If you're a customer and you get into a platform and they say, here's some ads that look like they run by the best practices of our platform, you'll be very excited about putting your first dollars into the platform because it makes it a very quick start.
Matt, maybe I'll shift to you on direct payments, so this has been a big topic in the industry, specifically gaming, but I think it applies to a lot of app-based businesses, but when we think about the opportunity for companies like game companies to circumvent App Store fees and do direct payments, what impact could that have on your business? Because a lot of investors are asking me, well, hey, they have more gross profit. Are they going to throw that at advertising?
Yeah. And it's a great question. Obviously, it's going to have a positive impact on us and everyone within the ecosystem. But I think what's missed is the timing of the impact and then also the dollar amount. So first, on the timing, this isn't a change that's going to happen immediately. We think that this is a change that the industry is going to make over a longer period of time. Firstly, I mean, mobile gaming companies are apprehensive about being the first mover about particularly when it's regulators on the other side and then the ecosystem like Apple and Google.
So we think that this is a change that's going to happen over a longer period of time, over 2026 and the years after. So I think that's important to understand.
And then secondarily, in terms of the dollar amount and that impact, not 100% of the payments are going to shift. We think that there's still going to be a portion of those payments that are still going to be run through those much easier payment platforms that Apple and Google have. So the impact on the actual percentage and the dollar amount of those costs are not going to come down to zero. So it's going to come down from that 30% that it is today to somewhere probably in the middle, probably in the 15%-20% range. And then that delta, that incremental profit, some of that is going to end up in the pocket of the mobile gaming companies.
They're going to decide at first, maybe I want my profit to increase slightly. And then some of that's going to hit the advertising companies.
And then over time, as that industry becomes more and more competitive because there's more profits to spend on user acquisition, a lot of that will then end up with the advertising companies like us.
Let's talk about competition for non-gaming. So another question I get a lot is, how does AppLovin stack up versus other performance ad channels out there? So when you think about that world, what differentiates what you offer from the Metas and the Googles of the world? And are there any misunderstandings about how you interact with them that are worth clarifying?
Yeah. So let me define performance first, because I think a lot of advertising companies say performance. And I'd argue that there's different levels of performance. What we try to do is take a top-of-funnel user, so help a user find a product that they've never seen before, or they don't know that they want to buy, get them engaged in it, drive it to the bottom of the funnel where the point of conversion happens, and price it all on revenue. The only other company that has top-of-funnel, the bottom-funnel price on revenue is Meta. Google is a bottom-funnel company. You go to search, you have a sense of what you want to buy.
Google closes the loop. Amazon is structured very similarly as well.
So when we think about our market, what we're going after, it's much more aligned with the type of customers that Meta has really helped scale businesses. Now, you then go and take that and say, well, we're competing with Meta. If we were to take a product to market, and Meta is as great as they are, they've built maybe the most innovative ad solutions that the world's ever seen. If we took to market a product and said, you've built your business on Meta, move some dollars over to us, because when I think about competition, I think about it as you're shifting from one versus another, we'd have zero shot. We're coming with basically no salespeople.
We have the name AppLovin. You're a media buyer. You're not going to go, I work on Meta. I built my business on Meta. I'm going to shift to AppLovin.
What was important for us is that we build tools that are so good and tap into an audience that's actually incremental, that they have not discovered the products that we're showing them on the social network. If they haven't, there's an opportunity to go to the same customer and say, you're spending $100,000 a day on Meta. You built a great business, but you're capped in your growth opportunity if your marketing mix is limited to what you know. All of a sudden, for the first time in a very long time, there's a new solution that can unlock incremental spend for you.
Instead of spending just $100,000 a day, you can spend $150,000 a day at the same return you were getting at $100,000 a day. That means that this same customer in this example can grow 50%. That's a huge economic opportunity.
What gets us excited about where we are as a business is we've gone from what I'd call a niche, executing exceptionally well inside gaming, to now proving we can build products for any type of customer in the world to market on our platform. If we can do that and we're proving that it's incremental, we're going to be able to influence economic expansion in these categories that are all much bigger than the gaming market we operate in. Do that at scale, and this company is going to be much bigger in terms of business opportunity than the one we've executed against the past 13 years.
Got it. Matt, maybe on capital allocation. So you've been very active in repurchasing stock over the past few years. Remind us again of your philosophy and your targets for capital returns. And then separately, just on headcount and reinvesting in the business. I think when we were here probably 12 months ago to the day, you were anticipating reducing headcount in 2026. I guess where does the balance fall? Sorry, in 2025. Where does the balance fall for 2026 in terms of investing in the business and hiring versus headcount?
Yeah. I mean, I think the plan now is obviously not to reduce much more. I mean, we're very low and very lean in general. I think it's missed because we don't disclose a lot about the headcount. The total company headcount's around 900, and then of that, there's only about 400 of those employees that are focused on the ad tech business and then the entire back office of the company, so we're already running very lean and efficient, so we will be growing headcount, but at a very low level.
We focus on hiring the best quality talent that we can get, so we'll be adding headcount to our growth initiatives, those being obviously the business development team to focus on e-commerce and then engineering as well, but it shouldn't have any material impact on the margin profile of the business.
So then what do we do with all that excess free cash flow? We're going to be continuing our buyback program. We announced that we increased the authorization by $3.2 billion. So we'll be very active in continuing to invest in the company and buy back shares, which we've done very well for investors by returning that to investors over the last few years.
Great. So Adam, maybe looking ahead into 2026 and beyond, as you think about the full launch of SellServe in 2026, I guess, what are the gating factors to get there? And then what is the pace of improvement that we should expect for the web targeting model over time? Should it be akin to the gaming model?
Yeah. So the web model is starting earlier. So as you've already seen with prospecting coming out, we're making really big innovations quickly. The underlying model has improved over the last year as well. And if you talk to the customers, you get a sense that their performance keeps improving. We're also starting at a point of data penetration that's really low relative to where we are in gaming. So as we get more data that comes from more customers, that model is going to keep improving. So what really excites us is that we're already really good, yet we know that the model and the data position is going to get only much stronger over time.
On the former, we're building tools really fast. The dashboard, the Axon Ads Manager, is very effective for the customer. They're having a good experience. Customers can sign up today.
I mentioned on the earnings call that since we went to referral state at point of earnings, we were seeing 50% roughly a week, week-over-week growth on that business line. Now it's starting small, but I've started multiple businesses. If you're ever going through a period of 50% week-over-week growth, you get ecstatic about a business opportunity like that. VCs would be lining up with a lot of money to fund something like that. So we're seeing that the opportunity is going to be really large.
Now, what gets us to open is that we're able to execute on marketing the brand, being able to tell small businesses around the world that this Axon Ads Manager and this Axon opportunity could be a game changer for them and their business. Of course, this brand didn't even exist a couple of months ago. So we have to nail the marketing side.
The way I think about it, and this has been helpful for investors to understand, is that I talk to our team that's doing this growth marketing testing right now and say, look, a great outcome next year would be if we could spend $1 million a day in advertising, and we'll do sponsorships, we'll do advertising, we'll do more media, we'll do podcasts, but if you combine it all and we spend $1 million a day, and it costs us $2,000 to sign up a new customer that goes live on our platform, you'd spend $365 million over the year, and you would get 182,500 new customers in one year. Now, recall that we've disclosed we only have thousands of customers.
If the number can ramp up that quickly with a number that's very immaterial against a revenue base, so you wouldn't even expect much of a margin impact, this business is going to be off to the races. We think it's doable. The second we're ready to do it, we'll open up the platform. We'll be talking about it with investors. And certainly, the small businesses in the world will start seeing that opportunity out there for them.
That's a great note to close on. Adam, Matt, thank you so much.
Thanks a lot.
Thanks, Matt. Appreciate it.