Hey, everyone. Thanks for joining us for a fireside chat with PubMatic. I'm Jason Helfstein, Head of Internet Research for Oppenheimer & Co. Very excited to have CEO Rajeev Goel and CFO Steve Pantelick. The format is a fireside chat. If anyone would like to submit a question, use the link just below the webcast or email me at jason.helfstein@opco.com. Gentlemen, thank you for joining. For those not, let's just start off with an overview question. For those not familiar with the company, what does PubMatic do and why is the company's offerings unique within a pretty competitive ad tech space? Yeah. Hey, Jason. Thanks for hosting us. We are a global platform for digital advertising, and we focus on connecting content creators, you know, this can be streamers, publishers, mobile apps, ad buyers, data owners, and commerce media participants, all on our owned and operated infrastructure so that they can grow their ad businesses. We're really enabling the primary stakeholders in the advertising ecosystem to grow their business. We started narrowly on the sell side as an SSP about 19 years ago, exclusively looking after the needs of publishers. Over the last five to seven years, we've broadened our platform significantly to facilitate transactions between all of these stakeholders with specific product capabilities and offerings that bring them to our platform. I think some of the capabilities that are unique, we're a leading sell-side platform. We drive significant yield for publishers. We lead in what's called supply path optimization. That's where we work with buyers. These are major agencies and advertisers to consolidate their ad spend onto PubMatic. We're extending that into direct buying in the SSP with our Activate product. We have a leading offering for curation. That's shifting targeting from the buy side to the sell side of the ecosystem, where there's a lot more data given the way privacy regulations are moving. We have a fast-growing commerce media business with great customers like PayPal and Instacart and many others. We do all of this on top of owned and operated infrastructure, which I think is an important differentiator. That means we own tens of thousands of ad servers. We own the network infrastructure and, of course, the software that's providing all of these capabilities that I described earlier, all on private cloud infrastructure that we own. That provides a significant source of outperformance and also profitability, which we can get into later. Great. Steve, can you maybe just broadly break down today what percent of the business is coming from the legacy SSP, retail media, and then your newer buy-side product? There's other, but just broadly. Sure. Just on the point of the diversification of our revenues, since we went public, we've dramatically built out our platform, as Rajeev just described. We have a fully scaled omnichannel platform. Over the last couple of years, we launched a CTV business, which is now about 20% of our revenues. We scaled a mobile app business, which is about 20% of revenues. The part that you're describing, we just launched sort of a whole set of new revenue streams within the last two years. As of this past quarter, second quarter, it represented 8% of revenue. Dramatic growth in new areas. Great. Rajeev, maybe give us an update on the ad market. You just reported results, but I think most folks got the sense that the quarter started out a little more cautious, but finished strong. What are you seeing right now, like third quarter and kind of feeling toward the end of the year? Sure. Yeah, I think in terms of the macro, our view is that it's been relatively stable. I think Q2 had a little bit more ups and downs, but we're anticipating a relatively stable environment going forward. We do see a tremendous shift across the industry towards advertisers looking for performance, for transparency, and for control. I think the open internet is becoming more and more about performance. Data is a big driver of that. There are many more logged-in environments. If we think about CTV, if we think about commerce media, if we think about the mobile app environment, all are significant growth drivers for us. AI is also changing workflows and reducing the lock-in of legacy UIs. Traders that might have been accustomed to learning a particular platform and that had created lock-in, now through a simple chat or text prompt, they can, in our platform, for instance, set up deals, do targeting, and optimize those deals. What we're seeing in our business is really strong growth in CTV. We shared that it was up over 50% year-over-year. The emerging revenue streams, as Steve just called out, are doubling year-over-year. As we look to the second half of the year, some of the key strategies that we're focused on are diversifying our DSP mix. We have some concentration there, which I know we'll get to, accelerating our investment on the buy side, getting deeper with advertisers and agencies, advancing our leadership in CTV, scaling our emerging revenue streams, and integrating AI across our tech stack and operations. We've made four major AI releases in the last year, and we continue to plan to accelerate that. A bunch of topics in there, which we will unpack. Just first, just to kind of level set, what do you think your rough market share is right now within the SSP community? Yeah, so we think we have about a 4% share of the market. We estimate Google is the largest at roughly 60% share. I think when you consider scaled, think about scaled SSPs, if you consider them to be global and omnichannel, meaning having CTV inventory, mobile app, and web, the things that buyers are looking to buy when they're trying to reach a particular consumer, there's really only three, right? Google, PubMatic, and one other public SSP. Obviously, Google was found guilty of having a monopoly in the publisher ad server and exchange market earlier this year, as well as illegally tying the two together. We think there are significant reforms on the horizon that should benefit us. Maybe let's kind of get into that. Obviously, no one knows what's going to happen, whether they have business remedies or potentially forced divestiture, and if they choose to appeal a forced divestiture or choose to appeal a remedy. Just even before that, do you see clients moving away from Google right now just because they're like, hey, there's an organization that is going to feel culturally interrupted? I'm not going to wait around until the remedy happens. I want to move my business now. Just talk about what you're seeing there. Yeah, I think that's exactly right. What we saw well before the verdict is that buyers and publishers that may have been using different parts of Google's tech stack could see that it was not objective or unbiased. I would talk to an agency buyer that would say, hey, somehow when we put money into Google's buy-side stack, DV360, magically 90% of it would flow into AdX. No matter what we did, we couldn't shift that. Because of that, I think people have been moving away from Google over the last couple of years, and certainly we've been a beneficiary of that. I think since the verdict, to your point, Jason, there's a big question of if Google is going to have to spin this thing out or undergo some behavioral remedies, then Google's own incentive to invest in this portfolio of solutions is likely to go down. They're unlikely to maintain the share that they have. If their investment is going to go down, then I better get ahead of that as a publisher or a buyer and move to independent solutions that are innovating and investing in the future as opposed to shrinking. I think that's a tailwind for us for sure. Maybe to that point, specifically on the SSP, how do you think about Google's relative pricing versus yours? I think most people assume they price at a discount because they can afford to do that. How dramatic is the discount? Yeah, so I think in the exchange market or SSP market, our understanding is their pricing is meaningfully higher than where the market is for SSPs. That is a result of, obviously, this tying and the illegal activity where market participants are forced into that platform. That's how they would justify a rate that's well above. When we look at what are the implications for us, we estimate each one-point share of the market is worth $50 million- $75 million in revenue for us. Because we're already processing all of the inventory from publishers, we have a very high degree of overlap with Google. It's a very profitable, very marginally profitable set of revenue. We think about 80% margin. I think that's one aspect of upside for us. We could significantly grow our market share. As you said, nobody knows for sure what will happen and what the exact timeline is, but it could be as early as the first half of next year. Second, we saw one SSP already has filed a civil damages claim against Google. That's something that we're also, we haven't eliminated any options at this point. In the quarter, you did highlight, I guess let's talk about, it's like you talk about diversification of DSPs as something the company's focused on. Last year, there was some interruption where one of the large DSPs made an adjustment to their auction buying behavior, catching up to the industry, but it still affected how they bid on inventory from PubMatic. In this quarter, you highlighted a DSP which was shifting the type of inventory they were looking to target. As a result, you've had to change the way you were making certain inventory available or showing different inventory. The point is, one is how do you diversify the DSP mix? By definition, there's only so many of them. Is there a cost? Is there going to be a margin headwind to making that adjustment for the most recent DSP if you have to, let's say, look at more inventory? There's more potentially maybe processing costs to that or something like that. Let me take the first part of that, and then I'll turn it over to Steve on the margins. Historically, our main source of demand has been DSPs. There is a degree of concentration, just given how the industry is built, which, Jason, you noted. We saw an issue with the top one last year and then more recently this quarter. While we are resolving the latest one, which is important, a top priority for us is also to accelerate the diversification of ad spend on our platform away from legacy DSPs. We've been making great progress, but we plan to accelerate our strategy here. Just a couple of data points. In Q2, we expanded the share of spending from DSPs outside of the top five, with performance marketers, mid-tier DSPs growing 20% year-over-year. A couple of examples that I mentioned in the earnings call: MNTN, which is a performance CTV ad platform, recently went public; tv Scientific, performance CTV ad buyer; China-based DSPs, where we're helping them, whether commerce or non-commerce, with their non-China business. In July, excluding those top five DSPs, all other DSP spend accelerated to over 30% year-over-year. When we step back, when we look at the ad market, what we see is that the ad budgets outside of the Fortune 250 advertisers are growing faster than they are in the Fortune 250. The mid-market of buyers, and mid-market can have many different definitions, but let's just look at it outside of the top 250. That cohort of buyers is growing at a significantly faster rate. We think that there's a lot of great opportunity to grow our business while also diversifying away from the kind of the legacy DSPs. I mentioned AI earlier, but we see that AI is starting to change the paradigm where people are not as locked in to a particular user interface because the learning curve is a lot lower. It is clear that the concentration of our legacy DSP relationships is a significant factor constraining our growth, and we intend to address that head-on. Let me turn it over to Steve on the margin part. Sure. With respect to that, the good news from our perspective is that the cost is really de minimis to open up access to a new DSP. It's really identifying who are the right partners, the performance marketers that Rajeev just described. We're already processing and incurring the cost for about 900 billion impressions per day. What we do is identify who the targets are, establish a commercial relationship, and get that DSP buying on our platform. Through the first half of this year, we've actually increased the number of new DSPs on our platform by about 20%. It's really a function of focus and then identifying the right partners to work with and then building out the business, developing business plans with them that meet their criteria, their goals, and obviously exposing them to all the various inventory sources that we have. On the interruption with the most recent DSP, where they're effectively looking for different kinds of inventory, is this something you can correct or we just have to lap past it? Yeah, I can take that. What happened with this latest DSP is that as they made changes in their platform, the type of inventory that they're bidding on has changed. The parameters of how they evaluate and value inventory have changed. We are working to optimize the inventory we send this DSP accordingly. Steve mentioned we process about 900 billion ad impressions. We don't send each and every one of those impressions to every DSP. It's just too much volume, so we have to select, OK, which impressions are we going to send? We are working very hard and very diligently on software iterations on our platform. This is machine learning algorithms and other approaches that determine which impressions, which traffic to send to this DSP. That work is ongoing. I am hopeful that we can reset that and correct that and revise that. I don't think we're in a situation where we necessarily need to be waiting for a year. That is ongoing work, and the results are still to be shown. Got it. When we get past that, how do we think about the normalized growth when we get past that? Yeah, I mean, from our perspective, all the things that we've been investing in over the last several years are really working quite well. Rajeev hit all the headlines. Our CTV business is growing over 50%, has done for the last four-plus quarters, emerging revenues doubling. When we get past this DSP impact that Rajeev just described, we absolutely see our total revenue growth reaccelerating. As a point that we called out earlier, we now have a full end-to-end platform. We've never had in our company's life as many ways to grow as we have today. I fully expect that as the DSP impact normalizes, we'll get back to the mid-teens to 20% year-over-year growth. Talk a bit about pricing. These are typically kind of take-rate businesses. That being said, when you have SPO, it makes it more stable and less volatile. On an overall basis, how do you think about revenue as a percent of gross spend? How stable has that been if you go back over the past two years? Sure. When we went public, we did comment that given sort of the scaling of the industry and the continuing evolution of mix, new formats, video formats, CTV formats growing, which typically have lower revenue share rates because of the high value of the CPM, $15, $20, $30. We assumed at the time of going public that there would be a gradual decline in revenue share rates across our platform over time. That is exactly what we've seen. There hasn't been any dramatic drop. It's been completely manageable from our perspective. Point number two is we've always been very focused on driving down our unit costs. From our perspective, our goal has always been to look at what are the absolute dollars that we're bringing down from revenue and then make sure that the costs against that are as low as possible. We focus on the marginal profitability of that. We are very comfortable with the evolution of the take rates over time. From our perspective, we are also doing many things to take advantage of the innovation that we have built over time, where now we have incremental revenues that are sitting on our platform without incurring significant incremental costs. What I mean by that is we already process 900 billion impressions per day. We incur those costs. With the increased functionality that we've created, whether in data curation, whether in commerce or Activate, we don't incur significant incremental costs because we've already spent the money to process. Those revenue streams, as they grow, add incremental absolute dollars to our platform. We feel really good about the evolution of the business, the foundation of it, and now sort of non-platform spend-related investments and outcomes. For example, we have an enterprise-grade wrapper software that is fee-based. That is a very important piece of software inside of publishers, and that's growing very sticky. That sits outside of the bid stream. We're focused on developing more and more revenue streams like that. Rajeev, you mentioned we talked about data before in your opening comments. Let's talk about how that ties into what's now referred to as supply path optimization or SPO. What is it, and how are you able to leverage your data to drive incremental revenue through it? Sure. Yeah, so supply path optimization or SPO is a process business engagement model that we created about seven or eight years ago. What we're focused on with SPO is making the case to a buyer as to why they should consolidate their ad spend on our platform. When an advertiser or an agency bids through a DSP, that DSP has choice in terms of the exchanges or SSPs that it's bidding on. We made the case to advertisers and agencies that, hey, you ought to consolidate your spend onto our platform so that you can increase your return on ad investment. What we've been able to do is grow from 0% at that time to now about 55% of our business is through supply path optimization. We were at about 35% just a few years ago. It's really important because of our ability to consolidate the market and build deeper and stickier buyer relationships, which is obviously a driver of growth and profitability. With SPO, what we are doing is whether it's data, so it could be targeting data, viewability data, log-level transaction data, workflows, integrations, commercial incentives, we are helping the buyer spend more intelligently and increase their return on ad spend when they buy through our platform. That, by the way, increases the yield or CPMs for publishers. It's a great win-win where the buyer is getting more return on ad spend, so they move more of their spend there and they buy more, and then the publisher is getting more yield. Data is a key part of this, Jason, as you mentioned. As I talked about earlier, open internet is moving more and more towards performance advertising. With performance, data at scale is obviously a critical input into that performance equation. What we've seen is just an explosion in the number of data sets on our platform. We launched a business called Connect a number of years ago where first-party data owners can bring their data onto our platform, and then the buyer can use that data to target within the SSP. Some examples of that: Instacart data is available on our platform, so a buyer can come in and say, hey, I want to target households that buy premium cereal, and I want to show a CTV ad across 50 different streamers on the PubMatic platform. We have Nielsen data, Experian data, PayPal data. We announced PayPal very recently, 430 million consumers. They have obviously all sorts of transaction data that segments are built on top of. This data, I think, is really key to driving that stickiness and growth in SPO. Why, I mean, a lot of that data is also available through the DSP or the buy side, but why is it working better if they do it through you versus do it through the DSP? Yeah, so there's a couple of reasons why targeting is moving to the sell side of the ecosystem. Let me just go through a handful of those. First of all, compared to 10 years ago, we're now in a world of consent, meaning in more and more locations, I'm sitting in California. If somebody's in Europe, somebody's in Australia, there's about 20 states now in the U.S. with privacy regulations. Consumers increasingly have to consent to having their data used in targeting. Consent sits at the intersection of the consumer and the publisher, and that's exactly where our technology sits. We have a higher scale of consented users on our platform. Second, regulators are wary of data being spread across the bid stream, meaning if a publisher has data from the consumer and they make that available in our SSP, regulators are more and more tamping down on that data being sent to the 100 DSPs that Steve mentioned on our platform in every individual ad transaction. Regulators would prefer that that data is used and targeted on the sell side of the ecosystem. Third, we talked about the volume of ad impressions that we have and how it's not feasible for DSPs to take all of those ad impressions. When a buyer does targeting within our SSP, they're able to increase their reach, meaning they can find more users and more ad impressions that have that data than they can in the DSP. We've published a number of case studies with pretty much all the major agencies and major advertisers where when they target in the SSP as compared to in the DSP, they're able to find more audiences and make that process more cost-effective, meaning lower CPMs and higher return on ad spend. We think that is a secular tailwind that we're only in the third or fourth inning of. Sometimes it's referred to as curation. Sometimes it's sell-side targeting. We think we have a leading product capability and platform in this area. Got it. Basically, supply path optimization kind of may have started a little bit of a volume discount, but then ultimately show the value of the unlock of the data. Now we kind of bring it full circle with Connect as regulations get more and more stringent. Yeah, I think, go ahead. No, you go ahead. Finish up. Yeah, sorry. I mean, that's exactly right, which is that we are continuing to broaden and deepen our value proposition for SPO. Maybe in that context, it's relevant for us to talk about our Activate product. Yeah, I'll go on there next. Yeah, perfect. All right, I'll let you tee up the question then. I was going to say before we got to that, I wanted to hit on, so the idea then that, look, you brought it up multiple times, it's not practical to send every impression to all the DSPs. It would just be like. Too much cost for them. Too much volume overload is driving costs up to everybody and reducing latency. Maybe with kind of like the step function we've seen in LLMs, as you deploy those LLMs, I guess how can you leverage LLMs to, let's say, solve this most recent DSP issue, just to almost like process more and kind of just do that faster? Yeah, we have been significant innovators and users of machine learning, which is, let's say, a different branch of AI than generative AI. Machine learning is, I would say, the relevant branch here because it deals with transactions as opposed to text and content or images. With machine learning, what we are focused on is a number of things to improve the outcomes on our platform. Traffic shaping is one of them, selecting which impressions we think a DSP is likely to bid on or win. We use machine learning in setting price floors, helping publishers determine, hey, how should they appropriately price their inventory? We use machine learning to match the appropriate buyer with the appropriate seller in terms of ad quality. There are a lot of ongoing uses for us with machine learning. With LLMs, there is no end to the opportunities. One great example is last year we shipped a creative classifier using generative AI technology for the very important political ad season. We were able to drive significant political revenue. The way that classifier worked is we would look at an incoming ad creative and classify the ad and say, hey, is this ad a negative ad or a positive ad? Is it about a candidate, a tax question, or maybe something controversial like a gun bill or an abortion bill? Many publishers do not want controversial ads, but they are OK with local policy issues. Using this creative classifier, we were able to open up a lot more inventory to political buyers and bring in significant revenue as a result. OK, let's jump to Activate. Activate is kind of, I guess, your newest product for buyers. What percent of the revenue is this today? Why is this important? I guess you could tie it into you've already had relationships with buyers through SPO. What percent have you been able to convert to Activate? I assume that's the sales pipeline. Yeah, let me talk about that. Then I can bring Steve in as well. With Activate, the challenge that we're solving is we talked about performance on the open internet, targeting, moving to the sell side of the ecosystem, and supply path optimization. What we discovered is that as we went deeper into SPO, buyers started to say, hey, why can't I just buy directly in the SSP? You see all of the ad inventory. You have more data there. If I could just buy directly within the SSP rather than having to go through an alternative buying platform that then integrates in, where there's discrepancies, there's latency, there's privacy and data concerns, there's higher costs, there's operational overhead. If I can just buy directly, I can increase my return on ad spend even further. That led us to building Activate, which is what's unique about it. Rather than an SSP and a DSP, so two kind of parallel platforms, there's just a single technology stack. We don't have to process inventory multiple times as would normally happen. We've been able to bring Activate into many of our advertiser and agency relationships. For instance, I cited a case study with Omnicom Germany on our earnings call. We've released prior case studies with Mars, for instance, where they've seen a significant increase in return on ad spend. I'll turn it over to Steve in terms of the scale and growth of that. Sure. We have MSAs with all the major holding companies. Many advertisers are using the Activate platform. From our perspective, it's now a matter of usage and then RAM. For example, when we launched, Mars, a global company, really saw the advantages in terms of transparency and control. That business in the second quarter doubled over the first quarter. The total revenues are embedded in our new revenue streams that I referenced earlier. We do expect Activate to become a very important part of the future of our business, as are all the other key relationships that we have with DSPs. Think of it as an important part of our portfolio. We believe it's going to be one of the fastest growing and a complement to everything else we do. I would anticipate that in the coming quarters and years, it'll grow to over 10%, 15% of our revenue over time. I guess, then the technical question, if someone's looking at this and saying, the benefit of using a DSP is I can understand unduplicated reach across my spend, across multiple SSPs, and obviously many, many more publishers. If I'm an advertiser and I use Activate, how do I still see unduplicated reach between my spend that's still going through my DSP? Yeah, I mean, the unduplicated reach, I think that's a bit of a myth, right? Because if we think about an advertiser's world, they've got to use DV360 to buy YouTube. They've got to use Amazon's DSP for Prime Video. Maybe they're using a general-purpose DSP, and they're already very much in a world where there is no unduplicated reach. With Activate, we're giving the buyer the opportunity to see all of the open internet, so we neither hurt nor solve that problem today. I guess I'm going to like to say, but then wouldn't a simple math pitch be, hey, look, the DSP is going to charge you 15%- 20%. Even if you have a 10% waste factor in duplicated reach, you're still saving 5%- 10%. I'm not sure I followed. Where does the savings come from in that? In other words, when you use Activate, you don't have to pay your DSP. Let's say the average DSP is charging 20%. I don't know what you're charging for Activate, whether you kind of build it in or it's incremental, but presumably, there would be just some take-rate savings not using the DSP and using Activate. Even if there is going to be some amount of waste through duplicated reach, you can mathematically, there's going to be some savings. Yeah, I mean, what we've been able to very clearly prove is that we can increase the return on ad spend with this approach. I think what you just said makes sense, Jason. Also, because there's no traffic shaping, the buyer can see all of the open internet inventory, which means they typically are able to find more users, more households at a lower price than if they just go through a traditional DSP. I think there's a lot of puts and takes, but ultimately what it adds up to is a significant increase in performance when buyers are buying directly in the SSP. Gotcha. OK, we probably have time for two more questions. If we kind of, you know, Steve, the question for you, where are you looking for like the most leverage in the model over the next two years? I mean, is it fair to say that like Activate requires a higher level of sales and marketing investment? Do you like offset that in other areas? Margins this year are obviously down because the revenue is weaker. Just even if you kind of just look through that, where do you see leverage in the model? A couple of places. I mean, we absolutely are going to continue to invest behind Activate. Our core strategy is always to find efficiencies and leverage those. As Rajeev described, we've been utilizing AI now in the engineering organization for a couple of years. We have now rolled that across all of our business functions. My expectation is we're going to continue to unlock efficiencies there and use a portion of those savings to fund sort of our fastest growing areas. I would expect leverage out of G&A, leverage out of team technology and development, and certain components of sales and marketing. My expectation is that we're going to be well positioned because as we've evolved the last couple of years, everything we do now is AI-first strategy. It's how can we use it? How can we leverage it before we go out and start to add incremental resources? The last thing I'll comment on, there's a couple of other things that are going to help us over time. I did reference them briefly. That's just our margin mix. We're moving towards higher margin offerings while our costs, as you noted, are highly cost leveraged. As we move to higher margin, we'll be able to continue to drive our margin rates. Just a kind of final question. We have seen kind of a tick up in CapEx and capitalized software, again, funding the data centers. We've talked about how certain video is more data consuming, et cetera, or data center consuming. Are we nearing, do you think we're nearing an end of an investment cycle around higher CapEx and higher capitalized software development? Yeah, I'd say just to frame it out for the folks on the webinar. Over a number of years, we've made really significant progress on reducing our CapEx. This year, our target is about $15 million, and that is about half of what we invested three years ago. Clearly moving in that direction. I won't say we're sort of moving to an era where we don't do any CapEx at all because we own and operate our own equipment. We really take the energy to get the most out of that equipment, keep it in service five, six, seven years. We'll have periodic replacement needs. Overall, I would say our level of investment will be lower going forward than it has historically for the reasons that we've cited, the software optimizations that we're able to achieve, the AI leverage. That leverage is not only from an equipment perspective, but also you referenced the impact on technology and development and capitalized software. I do anticipate that to not grow at the historical rates and potentially stay flatlined in the near future. OK, maybe final question on buybacks. I think the stock at these levels is kind of like below the average buyback price over the last several quarters. Just what should investors expect around buybacks maybe through the end of the year? Do we wait for the stock to get back to kind of just your thoughts on buybacks from here? Our capital approach has always been to what is going to deliver the greatest value to shareholders. First and foremost, it's making sure that we have the resources to drive innovation, which we have been able to get that incremental resources to put there. We've also been buying back shares. I shared at the last earnings call that we've bought back about $12 million Class A common shares. We're always taking a look at the balance between investing in innovation and buying back shares. It's going to be the same equation going forward. We have roughly $100 million left in our authorized program through the end of 2026. We're going to be judicious, but our priority is to keep innovating and growing the top line and then make an assessment on buybacks. Great. I think we'll leave it there. Thank you both for joining us today. If anyone has any more questions, feel free to reach out to me, and we can connect you with the company. Thank you, and everyone have a great day. Thank you, Jason. Thank you.