AppLovin Corporation (APP) Q3 2025 Earnings Report, Transcript and Summary

Welcome to AppLovin's earnings call for the third quarter ended September 30, 2025. I'm David Hsiao, Head of Investor Relations. Joining me today to discuss our results are Adam Foroughi, our Co-Founder, CEO and Chairperson; and Matt Stumpf, our CFO. Please note, our SEC filings to date as well as our financial update and press release discussing our third quarter performance are available at investors.applovin.com. During today's call, we will be making forward-looking statements, including, but not limited to, the future development and reach of our platform, including the expected timing of product launches, our share repurchase program, the efficiency of our operations, the expected future financial performance of the company and other future events. These statements are based on our current assumptions and beliefs, and we assume no obligation to update them, except as required by law. Our actual results may differ materially from the results predicted. We encourage you to review the risk factors in our most recently filed Form 10-Q for the second quarter ended June 30, 2025. Additional information may also be found in our quarterly report on Form 10-Q for the fiscal quarter ended September 30, 2025, which will be filed today. We will also be discussing non-GAAP financial measures. These non-GAAP measures are not intended to be superior to or a substitute for our GAAP results. Please be sure to review the GAAP results and the reconciliations of our GAAP and non-GAAP financial measures in our earnings release and financial update available on our Investor Relations site. This conference call is being recorded, and a replay will be available for a period of time on our IR website. Now I'll turn it over to Adam and Matt for some opening remarks, then we'll have the moderator take us through Q&A.

Thank you all for joining us today. First, I'd like to recognize our inclusion in the S&P 500, a huge milestone for our company and a strong acknowledgment of what we built. It's a privilege we do not take lightly. It also means we now carry the expectations of a much broader set of investors, and we must push even harder to continue delivering. Turning to our business. Q3 was another very good quarter. Our performance was strong with gaming advertising continuing on a solid trajectory. Our teams delivered multiple incremental lifts in our core models this quarter. And our MAX supply-side platform, one of the best indicators of our end market growth continues to grow at very healthy rates. We also opened up international traffic for advertisers promoting websites or shops in Q3 ahead of schedule. I'm particularly proud of our team because even while executing a strong quarter, we also delivered our major October 1 launch of our self-service platform and referral form. We did so without any significant hiccups, no major bugs and effective filtering out of low-quality ad accounts, something I was personally monitoring closely. This speaks volumes about our ability to automate and execute. I know everyone wants stats on how self-service is going. And instead of something specific around accounts or ramp-up since we're still very early, I'd like to point out a stat which I watch very closely. While it takes a while for new customers to get going, to integrate, to learn how to use our system and to ramp spend, we're already seeing spend from these self-service advertisers grow around roughly 50% week-over-week. It's too soon to be significant, but this type of early growth gives us even more confidence that our platform will excel at being an open platform to any type of advertiser. Our focus for Q4 and 2026 will be the following, with priority always given to improving our models for all advertisers. We'll continue tuning our onboarding flows and ramping more AI agents into the workflow to support a seamless experience for new advertisers. Once we're satisfied with the quality and experience, we'll open the platform broadly beyond referral basis. We'll be testing generative AI-based ad creatives. Over time, if we can move to mostly automated creative generation, we believe user response rates to more customized ads on our platform will materially improve. We are actively testing paid marketing to promote the Axon Ads platform to new customers. We'll continue tuning this acquisition method so that when we launch the platform beyond referral in 2026, we can scale advertiser count without a reliance on a large sales force. If we maintain execution discipline, we are well positioned to acquire a large volume of new advertisers in the coming years. We believe that giving our powerful recommendation engine, a more diverse set of advertisers to recommend will dramatically improve conversion rates, paving the way for elevated growth rates for years to come. It's worth noting the backdrop. The market is recognizing our platform, our scalability and the reach we offer our partners and the institutional dynamics that come with the S&P 500 inclusion are already in motion. At the same time, we continue to operate in an environment of heightened scrutiny around data, privacy and ad tech practices. We remain committed to strict compliance, transparency and execution excellence. To conclude, we delivered a very strong Q3. We are executing on our strategic priorities, and we are confident that our best days are ahead as we broaden access to our self-service platform and scale globally. With that, I'll turn it over to Matt for a deeper dive into the numbers.

Thanks, Adam, and thanks, everyone, for joining us today. Q3 was another exceptional quarter. Revenue was approximately $1.405 billion, up 68% year-over-year due to model updates in the core gaming business, while adjusted EBITDA was $1.158 billion, up 79% at an 82% margin, up 1% quarter-over-quarter from operating leverage and a modest reduction in operational FX. Quarter-over-quarter flow-through to adjusted EBITDA was 95%, slightly above Q2. Free cash flow was $1.049 billion, up 92% year-over-year. Free cash flow margin improved sequentially given no semiannual cash interest paid on our debt this quarter as those payments occur in Q2 and Q4 of each year. We ended the quarter with $1.7 billion in cash and cash equivalents. During the quarter, we repurchased and withheld approximately 1.3 million shares for $571 million funded by free cash flow. Over the last 3 quarters, we have reduced our weighted average diluted common shares outstanding from 346 million in Q4 of last year to 341 million this quarter. During the quarter, our Board of Directors increased our share repurchase authorization by an incremental $3.2 billion. Finally, turning to our financial outlook for next quarter. In the fourth quarter of 2025, we anticipate revenue between $1.570 billion and $1.6 billion, reflecting between 12% and 14% sequential growth. With adjusted EBITDA between $1.290 billion and $1.320 billion, targeting an adjusted EBITDA margin of 82% to 83%. Now with that, let's move to Q&A.

[Operator Instructions] Our first question will come from James Heaney.

Could you just start off talking about the characteristics of the advertisers that you've onboarded since October 1? Would you say the GMV of the advertisers are smaller than the initial 600 that you had in the pilot? Or are you going down -- more down market? Just any help there would be great.

Yes. Sure, James. They are obviously filtered set of advertisers when we curated the list last year. That was filtered by our team. And then this year, it's filtered through referrals. So these aren't like your local dry cleaner trying to come on to the platform yet. They are predominantly shops. They're not going to be as large as they were in the cohort last year, but they're not going to be materially smaller either. So think of them as comparable in mix. And then it's a broad set of categories. There's no limitation when you're open the way we are through a referral to the type of customer that comes in. So a broad set of shopping categories being represented.

Great. And then just one for Matt. Could you just talk about guidance philosophy for Q4? Just curious how you've used e-com seasonality from last Q4 as a proxy for this year? And just interested to hear kind of what's being assumed from sort of the current customers versus new customers on the e-com side?

Yes, sure. It's not the best comp. Obviously, last year, we had an entirely new e-commerce business that was ramping. And then this quarter, right, we had an existing base. So it's not best comparison year-over-year. So within the guidance, we took an approach where it reflects a combination of different factors that we have going on at the company, obviously, the optimism around the e-commerce referral program, continued model enhancements, the updates that we talked about previously within the Q3 period and then also kind of normal holiday seasonality. So the combination of those factors led to the larger guidance that we're projecting quarter-over-quarter.

Your next question will come from Omar Dessouky with BofA.

Adam, I wanted to get back to your comments about substantially higher conversion rates. So am I to read that as -- that a significant growth in impressions would not be required to absorb a significant increase in e-commerce advertisers in 2026?

Yes. I mean, look, we've always said we serve a lot of impressions to a lot of users today, over 1 billion users a day. So we're in a world today where the biggest lever for growth on our business, given we report on a net revenue basis is increasing the conversion rate. And that happens from a couple of things. You've got the model enhancements, which we always talk about. Those are super impactful in increasing conversion rate. That's a continuous effort. We are in the very, very beginnings of understanding how to work with neural nets and these AI technologies. I mean if you think about this industry and the core AI industry is only a few years old of engineers really being able to extract this kind of value out of these tools and technologies across a broad range of industries. So as this goes forward, we're going to have consistent incremental improvements and sometimes large, sometimes small, but additive to high impact on driving up conversion rate from technology lifts. Then you're also going to get advertiser density expanding paired with our recommendation system, giving the model the chance to personalize the advertising to the user better. If we have less advertisers and less categories, we just have less to show. So you can't get a diverse set of content to the customer to maximize that conversion rate. Both those things are just going to naturally happen as we go forward. The third piece that I touched on to your phrase you repeated, is that generative AI-based creative. Today, in our advertising system, the advertiser can do almost no targeting. They can pick their country, they can put in their economic goals, they can put in a budget and off they go. The one manual lever is creative. And in particular, in the shopping category or in this like website advertising business, a lot of the customers come on board and they don't have a creative that's adapted for our platform. The average viewership of our ads is roughly 35 seconds. The average viewership of an ad on social is roughly 7 seconds. So a lot of these customers are coming in and just porting a short ad and trying to replicate what they have on social on our platform, and it's mismatched. It diminishes their possible conversion rate. So what does that mean? Well, when we get into a world where we can use generative AI tools to automatically create ad creatives on behalf of these customers, they're going to get to a point where they can actually expand their conversion rate. We're doing nothing more other than just expanding the count of creatives into our system and ensuring that the types of creatives in our system follow best practices on our platform and doing that at no cost. And so we're really excited about where the tools in the marketplace are going. That's something we're going to be testing in short order here.

And if I could just follow up. So you've addressed the conversion rate question I had very clearly. But I did want to touch on how you're thinking about supply, even though, obviously, you've clearly -- you said that the conversion rate is a big driver. So in the past, you've talked about double-digit growth in publisher revenue in MAX, over the past few years. And I'm wondering if you expect that to accelerate as e-commerce ramps. Do you see supply driven primarily by higher ad load as dictated by publishers, higher engagement with mobile games overall or improvements to MAX like those you shipped in 2024? Like which of those factors would be most important?

I mean it's a combination of all of the above. The MAX platform ecosystem is growing really quickly. happens from a couple of different factors. One is, as the ads become higher quality, and we look at e-commerce shopping and just demand density, it's higher quality because the user stops seeing game, game, game, game, game when they're stuck to a game that they actually like. So if you bring more diversity of content, you'd expect retention to go up and ad supply to expand. So that's going to be a natural tailwind inside this ecosystem that will unlock as we get more demand outside of just core advertisers that we have today. Second part of that is what we've talked about in past calls is the unlock of getting these publishers who are in-app purchasing predominantly don't tend to run ads or run ads at a very, very small percentage. If you're monetizing a high LTV game and most of the gaming customers are residing in these deep games, you don't want to run ads for your competition. Well, they didn't have a way to monetize as well as gaming ads in the past. So if we can unlock this material demand shift, then we're going to be able to bring more supply into the ecosystem. That's very advantageous for supply as well. So it's a combination of those 2 things. And then, of course, you've got -- as our models improve, the same customer, that publisher can buy more users that are retained in their game, so they can create more growth in audience, and they'll get better tools as well to better monetize that audience.

Next, we'll go to Jason Bazinet with Citi.

So a quick question. I appreciate that 50% growth in week-over-week spend from these e-commerce customers. Is there any sort of context you can give us of like when you looked at that same metric during the pilot phase, what was it? Or how do you know that that's a good number or a bad number? I mean it sounds great.

I mean, look, like one of the simpler mathematical functions is extrapolation, right? So like we're starting pretty early here. It's a month in. And it does take a while for these customers to ramp up. Remember, ours is not a plug-and-play solution. They got to come in. They have to integrate pixels. They've got to go live. All of that takes time. So just to get to a point of go-live is a week plus usually. Some can do it very quickly, but it's not common to be able to do it in a day. So you have a period of lag time from October 1 to even a cohort going live. And then this cohort is not as big as what the absolute numbers were that we reported, I think we disclosed in Q1 this year of how big that cohort last year swelled to. But given the ramp-up, the ramp-up is really swift. So we're excited that if this continues to compound at the rate it is and then you keep adding new customers, this thing is going to then snowball on itself. And the true value for us right now is not go, okay, how do we extrapolate this out to when customers in this category become a bigger and bigger part of our business. It's more are we actually building a tool that we have confidence is going to succeed in converting a lot of customers over time. And that breaks down into a few things we look at. Of course, I get excited by seeing scaling spend, but I'm more excited about the fact that we didn't introduce a ton of bucks. We didn't have a bunch of customer complaints. We didn't bring in low-quality advertisers. Anything that was low quality was automatically kicked off the platform. So the team enabled a whole bunch of tools to get a product release that was not immaterial into market in a seamless manner. And now we're in a point of optimization. One point of optimization that's key is when a customer signs up, is it easy for them to understand the value proposition and get through integration to the point of go live. That's something we're really focused on optimizing because as we mentioned, we do want to eventually promote the Axon platform to future customers. And if we want to do that, just like any advertiser in the world, we've got to optimize our conversion funnel. The other piece that's important is our teams are constantly working on embedding tools into the interface, so large language model powered tools that allow the customer to get support without interfacing with us. We're not seeing a huge influx of customer support tickets. We're seeing these customers go live, be able to manage themselves, get best practices extracted out of the tools we've enabled in the dash, and that's fantastic to see. So what I look at when I see this ramp-up is not I'm extrapolating to how is this going to become billions and billions of dollars. It's more that the fact that it's working already a month in, and we're not getting bombarded with customer complaints, and we're not seeing a ton of issues implies that we're on the right track to eventually get open in '26 and really be able to bring in a ton of advertisers over the following quarters and years.

Next, we'll go to Clark Lampen with BTIG.

Sorry, I have like 4 mute buttons right that I had to take out.

Complex setup over there, Clark.

Right. It's a little too complicated. Okay. So as we're thinking about, I guess, the growth of what has the potential to be like a really big business for you guys over time, billions of incremental as you just sort of talked about. I'm curious how you think about balancing growth, chasing sort of new pockets of potential supply and building up demand to go after that with displacement for your core gaming customer because you are introducing a new bidder with potentially higher transaction value or consumer LTV. Is that something that model improvements and a faster pace of model enhancement will sort of take care of along the way? Or how do you, if any, think about gating the growth of this business if you have to for managing the core -- for the purposes of managing the core?

Yes. I mean, look, one, we don't try to gate growth. So we sort of look at a platform as it's going to develop and evolve as it does. But understanding these models gives me confidence that as we get more density of advertisers, we're actually going to have expanded spend for gaming customers, not diminished spend. And this is a bit counterintuitive, but here's why. The model today, if you go back a year prior to us getting into e-commerce and shops, you ended up having 1,000 impressions to show a user and you bombarded them with games. Well, that's not a great offering. It's as if you had a social network that was showing short-form video and said I'm only going to show golf videos. Everyone who's on the golf videos at that moment would churn. Well, in our case, the customers aren't churning. They're playing games, but they're not going to convert. And our conversion rates are really low. There are moments when the model knows a user is going to be in the game. When those moments happen, the CPM for a game advertiser is phenomenal. It's really, really high. Our average conversion rate is that 1% that we've given a year plus ago. So maybe it's higher now, but just for this example, let's use 1%. We're driving 10 game installs over 1,000 impressions, but we're probably wasting 80%, 90% of those impressions because the model knows in a tight percentage of impressions, games are going to convert, this user is ready for something new and the CPM there will beat anything else that comes along. So you go and bring in demand density. What happens? Now the model can better use all that access impression. And so if it does, what's going to happen is your gaming customer is not going to diminish. It's just going to be more targeted. So maybe impressions go down, CPM goes up for them, but everything is priced to revenue for our customers. So they get the same revenue out. And then these new customers better monetize the user and give them more diversity of content, which hopefully will train them to better engage with our ads. And then you take it to the next level, which is now all of a sudden, we're getting data from way more types of customers. We now know, let's say, tomorrow, someone buys a $5,000 handbag on a website, that's the data point we didn't have a year ago. That data point, my simple mind, can probably tell you that's a good user for Candy Crush if they haven't played Candy Crush. The neural net is going to tell you a lot more than that based on its correlations. And so you're building up a data set that doesn't just limit itself to the shopping category or the website advertising category. It helps enable better advertising for the gaming customers as well. So you put all those pieces together, and I'm really confident that we're not going to squeeze anyone in our platform. We're probably going to have expansion across the board as we add more demand density and get more data into the system.

That's helpful. Maybe just as a very quick follow-up. You guys highlighted tuning the onboarding flows and generative AI creative. How far away are you or sort of at what sort of rate are we making progress to sort of getting one of those tools live? Would either of those things, I guess, be a gating factor to launching or introducing GA at this point? Or are they vital to going to a broader customer base?

So the Axon Ads site is a prompt at this point. And part of the logic of doing that was to get inputs into that prompt, so we can tune a bot that's external facing and then bring it internal. So we're in the midst of doing that, and that's not far off. So we'll have different implementations of bots inside the site. I mean, as you know, and we've talked about last quarter, when a customer uploads ads or uploads a website to promote, we're not manually checking it. There's a bot there that's automatically checking for our quality and ensuring that the quality meets the standard that we need on the platform, both for the creative and for the website being promoted or the application. So we already have various points of bots into the tool, and we'll have more. It's just going to get better as we get more data in, and we can tune it so that it's actually accurate across the board. The generative AI-based ad creatives don't depend entirely on us. But Sora 2 came out this past quarter. That was another step forward. Veo 3 keeps getting better. So you can assume these tools are getting pretty close. We're not -- we're probably far away from the point where the model itself can recursively just go and create more content and just bring more ads into the system. But we're probably not very far away, and I'm hoping, in a matter of weeks or months to be able to test generative AI-based creative that we create with a little help, some tooling on top of the large language models and then submit to the advertisers for approval. That in itself can really explode the count of ads on the platform. And so that's not far off. So on your last question of are these gating to general release, I don't think so. I think the main thing that we care about there is how we tune the flows. Is the conversion funnel appropriate? Are customers getting confused or are they getting a good experience from onboarding to ramping? If they're getting a seamless experience, and we're not getting overly overwhelmed with inbound complaints or concerns or need for support, then we're ready to go open.

Our next question will come from Alec Brondolo with Wells Fargo.

I appreciate it. I think as we've kind of spoken about direct payments and this transition from kind of paying the App Store and the Play Store 30% to going to an O&O payment product, we've kind of talked about this, I think, historically as more of a medium- to long-term tailwind. Do you think that might be manifesting sooner than expected? And did it contribute to third quarter results?

I don't think it's contributing much at all yet. I think it is going to be over the quarters that it's going to take impact. I don't think it's realistic that 30% tax goes to low single digits. So let's just take the midpoint, 15%. That's a material lift in LTV for a lot of these in-app purchasing games, roughly 20%. Some portion of that is going to go into development of more content, which is great. They're going to make better games. Some portion of that, they'll bank into the bottom line. Some portion of that will go to marketing companies. I mean I'll say the one thing about our business always is we don't try to think about what's happening outside of us. This is something that's outside of our control. It's up to the platforms, regulators and then the content creators. It's not up to us. What's up to us and inside our control is how good our tools are. Q3 was driven by what we said on the earnings script. The models continue to get better, iterative improvements in the template, more advertisements on the platform, more advertisers on the platform, all of that's compounding to really quick growth rate. even in the core category. We're still believing very confidently in this 20% to 30% long-term growth rate in our core category. But even in the core, we're beating that. And then now you're layering on, on top of that, all this opportunity with the self-service platform. So we're really excited about where we are focused on what we have underneath our control.

Perfect. And maybe if I could ask a follow-up. I think there's always been this talk about eventually extending the reach of supply. Like right now, most of the ads are placed in mobile games and then the idea is perhaps over time, we could go to other surfaces. It seems like some combination of AdX and Google Ads Manager might come up for sale as a function of the Google ad tech, antitrust trial. Would you be interested in those assets if they were available?

I mean, without commenting on what else is out there, like commenting about our business, the reality with our business is we think about providing the best solution to our partners. For the advertisers, we believe we're on the track to really give them a good way to access a really large audience playing games. For the game publisher and then expanding publishers as we get to more publishers on our platform, we've built them very good tools to monetize and promote their products and grow their businesses. As we think about going broader than that, the open web publishers and other app publishers that currently can't access our tools the same way could use better monetization. We all know that category is pretty slow growth. And then we talked about CTV in the past, too. Everywhere else is struggling to monetize except in the walled gardens and except on games, predominantly because of the success of our platforms. And so if that's the case, we look at that as our potential prospect clients as well. We should be able to extend our product offering over time to those folks. We need more demand. We're not demand -- we're not supply constrained today, we're demand constrained. But if we do our job right and bring on a lot of advertisers, it serves us well because it serves them well to be able to extend our offering out to more publishers.

Our next question will come from Vasily Karasyov with Cannonball.

Adam, I wanted to follow up on what you said earlier about LTV calculations that you think your advertisers do. So if given the variance in growth rates in in-app purchases market and in-app advertising market, right? So can you comment on what you see if the -- your advertisers' LTV calculations are evolving, let's say, compared to a year ago compared to now, I would think that the advertising component would be much higher now, right? Going and how do you -- what it means for you?

So look, generally, as a whole, the in-app purchasing market is more mature than the in-app advertising market. So we're seeing much faster growth rates on the MAX platform. We've said multiples faster than the in-app purchasing market grows because those publishers, both the older publishers in that category are getting better tools, both for growth and monetization. And then newer publishers see these other publishers scaling, so they integrate more ads. So you're seeing more supply come online and existing supply monetize better. In-app purchasing market isn't really creating that many more monetization tools. It requires for growth, this tax to go down if the 30% goes to 15%, that creates a really big tailwind, and it creates better monetization mechanics. That's really hard in that category. It's up to the game developers, but it's not uniform across the platform. And then it requires more IAP games. There's a really big set of games that are just mature in that category that frankly don't really have much of a way to grow anymore. And then there are newer ones that constantly come live. If you look in the top 10 to 20 in-app purchasing -- top grossing games, there's a huge mix of games that have launched in the last 2 years that are now on the top 20. So there's new ones that go live that help grow there. But for us, what we're focused on is where we can drive an impact. We can help the in-app purchasing ones promote themselves. The in-app advertising ones are really exciting for us because they power our true market. The true market is the supply side on MAX. And as we give them better monetization tools and better growth tools, we see that supply expanding. That supply expanding the growth rate there is the direct market that helps us grow. And then hopefully, we can grow on top of that even more because of improvements in all the other parts of our business.

Our next question will come from Matthew Cost with Morgan Stanley.

Thank you for the 50% week-on-week growth metric that is really interesting. Is that the metric that you're managing to try to find the point at which you're going to go general availability? And if not, what are you looking at? Like what will be the things that you need to knock down for that to happen?

All these things take time to build. So I don't like spend -- obviously, if spend wasn't going up a lot, I'd be a lot more concerned when you got a business in any scale growing 50% week-over-week, there's a reason to be excited. But what I care more about is that we have time to optimize the funnel. We need to make sure the conversion funnel is optimized. This is just like launching a B2C property. Your first funnel is not going to be your best funnel. We've already optimized it once. So what's today on the site is better than what was on the site 4 weeks ago. But there's room for improvement there. Your communications with the clients, the e-mails that the potential clients get after they sign up, we can improve that. We can improve the tooling inside the dashboard, all the AI bots that we've talked about. And so there's different aspects of this that we just need time to get to a place where we go, this meets our quality standard, then we're ready to open up. Because if we open up today, we may be okay, we may not. We may get inundated with user concerns as you shrink the size of the advertiser. So one day, I'd like to make sure that, that local laundromat that signs up gets a great experience promoting themselves to this gamer audience on our platform. And if we're not ready yet today, we're going to take our time to get there. We don't think it's a long time away. I think I'd said on our prior earnings call for '26. So this isn't a very long time away. We're just going to take our time to ensure that we've got the product at the level that we want.

Great. And then on the paid marketing front, I think you talked last quarter and maybe even mentioned in passing this quarter, the opportunity to do more of that. It looks like based on your sales and marketing budget in the third quarter that it wasn't something that you started to lean into. So how should we think about the timing and potential magnitude of doing more marketing?

Yes. So we're testing right now actively. Testing budget is going to be not large. Even if we ever scale this, the scale of our business is quite large. So like some of the largest advertisers in the world can only spend a couple of hundred million dollars a year. So as you think about the scale of our business, it's never going to be a very large line item against the revenue potential. But because our LTV is so high, and we're optimizing our conversion rate, and we think we can get that to be really compelling and the brand isn't known, the setup is really good for promoting our product to potential end customers. And so -- if that's the case, and we know we have a great LTV to cost of user acquisition, we'll spend. We'll break it out, as it scales, we'll show the unit economics so people can understand what we're doing, but we're really good performance marketers, I think we can all agree at this point. So we're not going to waste money on this. We're not brand marketers. The dollars will be much greater than the dollars we spend on user acquisition. And it's -- for me, it's the full-blown automation of a sales force. We'll need some salespeople, but we can keep a very lean sales team in line with our traditional culture if we can automate onboarding through advertising all the way to point of go live.

Our next question will come from Benjamin Black with Deutsche Bank.

Is there any reason to think that the take rate or revenue margin from your e-com spend should be any different to that of the core gaming business? So any structural differences, I guess, to the advertising credits you're offering folks early on, perhaps lower the conversion temporarily. But longer term, are there any differences to be aware of?

No. I mean, look, the advertising credits we offer is such a tiny fraction to overall value of a new customer. So it's no different than like a cost of user acquisition if we got the customer through paid marketing. So it's just really low. So consider that immaterial. The business is not built to, say, web advertising or shops is treated differently than games. It's one unified auction. We're a single platform. So when we get a higher conversion rate, whether that comes from gaming or shops or any category, it's going to have a constant take rate across the board. It just implies that more density equals better conversion rate, possibly higher take.

All right. Great. And then second question would be sort of you're clearly growing your ambitions within AI automation more broadly. So maybe talk to us about sort of your investment priorities as we sort of look ahead to the next year. I'd imagine your sort of compute capacity requirements are likely scaling. So sort of how does that play into sort of your expense outlook as we look ahead to the next year?

Yes, we're pay-as-you go. So like I mean, we try to project and buy, in particular, GPUs, that being the more lead time dependent part of the stack. We try to buy those a year in advance. So if you've looked at the financials, you'll see spikes in infrastructure investment, but it runs through the P&L. It's not capitalized, and we do plan it really effectively. And we don't try to really overinvest ahead of revenue. We want to make sure we're really disciplined. And that's completely aligned with our culture where we want to be cost disciplined in every aspect of the business.

Next, we'll go to Chris Kuntarich with UBS.

Hopefully, you can hear me. Just wanted to ask on web-based becoming available to EU advertisers. Any update there? And then, Matt, just a quick follow-up. Are you making any assumptions about advertisers that aren't currently onboarded in the 4Q guide?

Yes. On the EU side, so we can work with EU advertisers today. We just don't open up our inventory for website or shop advertisers in the EU region of our audience. So just a clarification bullet. It's -- EU tends to be somewhere in the low teens percentage of our business, if I remember off the top of my head. So it's not a huge priority versus expanding out the business. GDPR rules are more restrictive and require a build-out for us. So we'll get to it in due time. It's not a priority against getting to general release of our platform and building out the rest of these tools we've talked about.

Yes. And in terms of guidance, I think we've done pretty consistent to our approach thus far. We've communicated before that we guide to kind of where we feel very comfortable that we could potentially land. So we guide to what we know. We don't guide to try to estimate for something that's unpredictable. And in this case, we can't predict the number or the volume of new advertisers coming on to the system through the referral program and how that potential ramp in spend could happen through the quarter. So there is no incremental assumption built into the guide for onboarding of incremental customers.

Next, we'll hear from Rob Sanderson with Loop Capital.

I have 2. Just in terms of kind of understanding more of the sort of the current points of friction to bringing people on, it sounds like you're doing a lot of work to tune the onboarding flow. But sort of what other points of friction are necessary to address to just further optimize? And then you've also said that you're not seeing a ton of complaints, but I'm sure you are getting asks for features and things. So maybe what are some common sort of asks for feature add to Axon Ads Manager?

Yes. So the first question, I mean, like you've got 2 points of this funnel, pre getting-to-us. So eventually, we've got to get the brand out there. We got to be able to market the platform. We've got to -- we're constrained by how many referral codes we gave out. So like just advertisers coming to the Axon platform and signing up. So let's set that aside because we purposefully constrained that. After the sign-up, we want to make sure we have as little drop-off as possible. Of course, with any product, when you get a sign-up, not every sign-up is going to be qualified, but we think a lot of these are qualified. So we're optimizing to as high a rate as we can from sign-up to go live. In terms of feature requests, surprisingly, not a whole lot. This is like, I would say, hard to guarantee that it will be that going forward because we're only a month in. So like if there's a lag time to integration and ramp up, and we're seeing the swift growth, you don't have time for these customers to really understand the platform yet. So that may change, but we get more feature requests from our current cohort of customers, the ones that went live a year ago, much more so than what we brought on in the last month.

If I could ask on international. You mentioned that you launched a little early. And it sounds like no EU. So maybe where are you available? Anything surprising or different about the behavior from this cohort? And then just anything you can share on next steps to expanding? I mean, you probably have some language optimization to do, I'm sure, and channel development work, but kind of what are -- sort of what are some of the next steps to make international a much bigger component?

Yes. So first off, it's everywhere in the world, except for EU traffic for web shops and web advertisers. App advertisers are everywhere in the world. We don't operate inside China, so -- but the rest of the world is there. In our business, the customers that we have today are mostly Western shops. And so those Western shops aren't likely to go into Japan or Korea. Japan being our second biggest market and promote themselves. They're just not going to have a localized product offering. So the countries that have really been successful for this current customer base are the obvious ones, Canada, Australia, New Zealand, et cetera. So English-speaking, similar makeup to the U.S. As we go open up the platform, that's when we'll go try to get local presentation inside Japan, Korea and other markets that are more closed off, but very large markets for us. And in the Western markets, we already have a presence. So it will be faster to get going. I think over time, localization is fortunately not a challenge anymore with LLMs. Language is pretty easy to solve. So we're fairly solved over there. It's just much more built around getting the system to be workable the way we're talking about for the West, goes global, right? Like all users around the world that we see playing games, the human beings behave similarly. They might buy a shot from different shops because there's local merchants in each one of these markets, but humans aren't behaving much differently in Japan or Korea or Canada or Australia to the U.S. So the model translates all human behavior to math. Math is universal. And so as we launch and broaden out the platform, we don't think there's going to be some big lift to really see a lot of success internationally. And we haven't seen that as we've opened up these markets and have these same customers expand out.

Our next question will come from Martin Yang with Opco.

Sort of related to the last question, can you maybe talk a bit more about your current cohort, how they have been performed in 3Q? For example, are they more actively spending, given your improvement in tools, having more features, et cetera?

Yes. Look, over the last year, I mean, it's been a year since we've had this product, right? So the team is continuously improving the product. So the return on ad spend for the customers have gotten better. The tooling has gotten better. We went from the old dashboard to the Axon Ads Manager. So the tools that they have at their disposal has gotten better. The customers' understanding of our platform has gotten better. Just that nuance on ad creative that ours are 35 seconds on average, whereas social is 7 seconds. It took months with a lot of customers to explain that fine detail. And that if you're not building a 45-second ad, you're going to lose to your competitor. So all these things just compound. We're 1 year into a product in a very large advertising market, competing with other companies that are years or decade plus into that same market. So as you build better tooling and as you get a better understanding of your tools, you see the effect compounding, the knowledge compounding, the usage compounding. And so we're seeing trends that are positive. But this thing is going to take time to build at the level that we want to build. We've been in the gaming business for 13 years. We're clearly the best channel for the gaming customers at this point. We think we can replicate that success across all these other categories. And as we build to the scale that we're accustomed to operating at, you'd expect a lot of compounding success over time across knowledge, usage and tooling.

Got it. I have a quick follow-up on the PSU issued in October. That's for engineering employees. Can you maybe give us more context? Is it for a small handful, single recruiting, retention? Any additional details would be helpful.

Yes. It's a pool, Martin, for a group of engineers, but it's also a future tool that we can use for recruiting as well for new hires into the engineering team.

Next, we'll go to Jim Callahan with Piper Sandler.

I just had a follow-up on the referral codes. I guess, are all the codes so far given out? Or is this something we can expect the partners to sort of continue doing through early 2026?

Yes. Generally, if there's a partner that we gave X codes to and they run through X and they deliver quality leads, we give them more. And we can measure back to every referral partner success of the customers they bring. So we don't want to constrain good referral partners. The gate is solely to slow it down so that we have time to build the tool the way I've been talking about. But if someone is bringing us good leads, we're going to take good leads. So we're going to be dynamic in the codes that we issue across the board if we see success coming in.

Got it. That's helpful. And then I guess just a follow-up on -- you talked about low quality and like making sure you're selective and having the right kind of advertisers. I guess what would you define as like low quality or an advertiser that wouldn't work sort of with the platform?

Yes. I mean -- so over time, you broaden out the type of advertisers you have to basically everything because you have a lot of density and a customer is never going to see 200 impressions of the same thing. Where we are today, we don't have a lot of density. So in certain categories, a customer may see 200 impressions of the same thing. Well, the bar we set right now is if our team is willing to buy that product or not. If they think it's a good product, they're willing to buy it, great, we want to run it on our platform. That doesn't mean we're going to maintain the same standard forever. Most of the companies that today we call low quality or we're not letting on the platform, they're running ads on social, they're running ads on search. These are real businesses. They're substantial businesses in many cases. But we just want to make sure that we think about our audience as a very consistent large audience, and we want to train them that our ads are really high quality. But in the absence of that competition, if they're getting bombarded with single offers, we want to make sure each one of those is really, really high value for that customer.

We'll take our last question from Nat Schindler with Scotiabank.

I'm going to go really high level and touch back on an earlier question someone asked about whether or not you were interested in some of those Google assets if they ever came up. You're growing at obviously absurd rates and you're doing it to -- it sounds like you're going to continue in your core market in gaming and you're adding obviously e-commerce, which is an enormous opportunity. I assume a lot of this is your conversion rates keep improving. You guys have been great at that over time. But some of it has to do with inventory itself. So at some point, conversion rates can improve too much. And if e-commerce is a lower converting area than gaming, when do you run out of inventory on your core gamer market?

Yes. I mean, look, we don't know is the simple answer. There's a long way to go, we think, just because we have so little advertiser density today. There's never been any company that's been set up like ours in the advertising space in history. Where -- what we reported, I think it was in Q1 was $11 billion plus of ad spend. And then the disclosures we've given you across web advertisers and gaming advertisers puts it in the low thousands. So you've got such a high amount of spend for such a low amount of advertisers across over 1 billion daily active users. Well, what happens when we get more density? You get more data, you get more density, you get more time to improve your models, that conversion rate is going to go up. If you think about social. It's not like there's more users on social. The growth in social has consistently come from the last few years, more customers getting a higher conversion rate because just the technologies are getting more powerful. And so we think we're going to see the same thing, but we're going to be able to pair it with this advertiser recruitment. And we're starting from such a low point, there's going to be quarters, possibly many years of growth in conversion rate to come before we start worrying about supply. Now that said, we've talked about -- it is interesting to us to go help the broader set of publishers, both in the open web and connected TV to better monetize. We get pinged all the time. It's no secret that we're really good at performance advertising at this point. So at some point, we would like to broaden out the supply base as well. Because, why not? That builds a really good growth catalyst into the future. But today, we're really heads down focused on the demand side of the platform because there's so much work to do there. At some point, you'll see us talking about both sides.

And that concludes the question-and-answer session for this quarter. We thank you all for joining us today. Have a good afternoon.

Thanks, everyone.

Thank you.