Conference Transcript - PUBM 2026-06-03

Rob Kulberth, Analyst — Evercore ISI

Great. Good afternoon and welcome back to the conference. This is Rob Kulberth from the Evercore ISI Internet Research Team. Rajiv Gohl, CEO of Pubmatic, is here with us today, despite what it may say on the title card that you see. So, Rajiv, great to see you and thanks so much for the opportunity. Thanks for being here.

Rajeev Goel, CEO

Yeah, thank you, Rob. Excited to join you today.

Rob Kulberth, Analyst — Evercore ISI

So, why don't you take this opportunity outside the quarterly earnings cadence to step back and talk about some of the big picture reinvention of programmatic that I think you're driving at Pubmatic. So, for starters, and maybe people aren't as familiar with the story, maybe you can just sort of level set for us the role that Pubmatic plays in the programmatic ecosystem today and what the vision is for Gentic OS and how that potentially expands the ecosystem and your role in it.

Rajeev Goel, CEO

Sure. So, Pubmatic, just in terms of who we are and where we are, we're an AI native advertising technology platform company focused on performance advertising. So AI-native means we're focused on AI internally as well as externally, and that gets to your question on Agentec OS, and we're really thinking about the opportunity as how to deliver performance advertising in the open internet across streaming TV, mobile app, and web or, you know, browser-based monetization. So the opportunity with generative AI is, I think, a really, really significant one. You know, we've seen two major transformations in our ecosystem before. One is real-time bidding, programmatic, and the other is the shift towards mobile. And this, I think, is going to be much, much bigger. And the reason is that today our ecosystem has increasingly complex workflows, data, a variety of systems that publishers and advertisers have to navigate in order to successfully deliver advertising campaigns and to drive a relevant ad experience or revenue from a publisher perspective. And Gen AI creates a huge opportunity to really collapse the value chain and bring the publisher and advertiser much closer together. And that's a huge theme, I think, of what we'll talk about today, Rob, but what we're focused on as a company is how do we use Gen AI, how do we use new technology to bring the publisher and the advertiser much closer together and drive performance and drive efficiency in the advertising process. So, with Agentec OS now, we have over 20 different AI agents that buyers can use, that publishers can use. So, some examples of buyer agents, inventory planning, audience discovery, media activation, fee transparency, guardrails, and governance, so, you know, those are four or five agents that I just cited, that our clients are now using to plan and execute their ad campaigns. I think one of the things that's most powerful is that we're breaking through the UI lock-in that many of the incumbents in the industry have had historically, where a big advantage was that traders or AdOps people, they knew how to use the software interface, the UI of a particular provider. And now with the Gentic OS, you don't need to know a UI. We can significantly increase, or rather decrease, the amount of time that it takes somebody to set up a campaign, to manage a campaign, to troubleshoot it from hours to minutes and the buyer can be often executing. They can use a lot more data, they can use machines to automate things, and so that drives the performance.

Rob Kulberth, Analyst — Evercore ISI

So Genethico seems to be changing almost by the day, and that's some very interesting stuff that we want to dig into in a second. Want to get back to, you've already executed thousands of campaigns on the platform. uh you just address you know some some of the benefits just a few of the benefits that you convey uh to advertisers and publishers from agentic os but you can maybe dig it a little bit more into the performance benefits so you talked about yeah okay you're a performance advertising focus

Rajeev Goel, CEO

company yeah how did agentic os express that yeah so we've seen you know great ramp up and usage of agentic os obviously it's still early in terms of the the ai transformation in the ecosystem We launched the industry's first Agenta campaign in December with an agency called Butler Till. And we shared in our last earnings call, which was, I think, about four weeks ago, that we had 30 campaigns that were now live and running at that point in time. Publishers have created over a thousand AI-powered deals on our platform. So it's still early, but it's starting to scale and you can see the the direction of travel very clearly so let's talk about some of the benefits so with butler till they're an independent agency in the u.s they had a alcohol beverage client we were able to show them using agentic os and activate which is the the buying direct buying interface into our ssp that they could reduce fees that they paid to run this campaign by 40 percent which in turn led to 40 percent more impressions or media delivered for their budget another example is abovo max lead in the netherlands i'm sure you have not heard of them before they are the largest independent ad agency in the netherlands so they're actually important in that market and with abovo what we were able to demonstrate is an 87 reduction reduction in campaign setup time and over 70% reduction in campaign troubleshooting time. So meaning using AI agents to set up the campaign, instead of having to click through multiple pages of complex UI, multiple screens, a complex campaign can take a seasoned trader over an hour to set up. Now they can do it in minutes, just through a chat interface or through an AI agent. for campaign troubleshooting. Okay, great, I've got the campaign set up. It's not scaling or it's not delivering the right audiences or, you know, my click-through rate is not quite right. How do I fix that? How do I address that? We're able to collapse the troubleshooting by over 70%. So pretty significant gains. And you're going to see, you know, more and more case studies coming from us. And I want to just comment for a minute on how we're able to deliver, you know, these uh uh these benefits and it's really um a number of pieces of technology that we've been working on for three or four years that are all coming together now so first is the scale of our ssp so 20 years of history there over 2 000 premium publishers over 100 000 streaming services mobile apps and websites you know we have 28 of the top 30 streamers globally so this is the head the market premium content number two is activate which is direct buying in our ssp so again collapsing the value chain bringing the publisher and the advertiser closer together third is agentic os which we've been talking about over 20 ai agents for buyers and sellers to use fourth is connect our data platform we have over 300 data partners uh commerce media demographic you know behavioral data contextual data and then finally is our own and operated infrastructure So we own tens of thousands of servers around the world where we process a trillion ad impressions per day, petabytes of data on a daily basis. We have co-innovated with NVIDIA, which I'm sure we'll get into, around a GPU architecture that really drives the execution of the AI of the data and of the inventory. So it's really this unique combination of capabilities that, again, we've spent years building that are kind of coming together at this moment in time when Gen AI is really disrupting our industry.

Rob Kulberth, Analyst — Evercore ISI

I guess, you know, stepping back a little bit, agentic seems to mean different things to different people, right? And so I think some people in the market, when they're talking about agentic opportunities, it seems to be more about replacing, you know, that manual process that accompanies, you know, traditional insertion order, Which is still you know significant part of the market. Yeah, I think that that your view is you know that plus You know significant automation benefit for the programmatic ecosystem

Rajeev Goel, CEO

Can you talk about what enables you to do sort of both sure? Yeah, so and and I'm just on the definition of agentic I think about agentic as an AI agent That is autonomous in terms of its decision-making capability, right and that's different than let's say we go So, you know, pull up Gemini, OpenAI, whatever, and we're doing a text-based chat and saying, okay, hey, I'm going to Spain next week, like, give me an itinerary for my family. There you're getting, you know, data back, but it's not making a decision on behalf of you as an economic actor. So we're really focused on that autonomous decision-making. So I see, you know, two big opportunities. The first is, within the programmatic ecosystem, you know, our industry has been very focused over the last 10, 15 years on real-time bidding, right? And that is this, at the point of a decision of, hey, this ad impression needs to get auctioned. Consumers on an app or a website, an ad is going to be shown, let's send this ad using a real-time bidding protocol to hundreds of different demand-side platforms, DSPs, that can bid on it. If we step back and look at that workflow, there's actually a lot of upstream and downstream work from that point of activation. So there's discovery, who are the audiences that I should be targeting for this campaign. There's planning, okay, what inventory, what price points, what properties, what third-party audience data should I overlay. There's the activation piece itself, the real-time bidding. And then after that, there's the measurement and reporting. So with AI, what we're focused on is autonomous decision-making across that entire end-to-end workflow. And it's clear that AI works better when it's not in silos, right? When the AI can work across the workflow entirely, and when it has access to all the data, meaning there's data on the buy side, there's data on the sell side, there's other data that sits outside of that, bringing all of that together for AI to act on. So that's a key area of focus for us. The second piece that you mentioned is, you know, in environments like streaming, for instance, there's still a lot of I.O. transactions. And historically, we have not participated in those I.O. transactions. With our agenda capabilities, that is now part of our TAM, our addressable market. And so I think what that feature looks like is, you know, today an advertiser launching an ad campaign, they might email out an rfp to 20 top streamers it's okay i'm launching this campaign what offers what audiences what inventory what price points do you have for me and then you know sales rep you know fills out the spreadsheet responds back there's some back and forth that can all be done with ai and i think the opportunity when you do that is an ai agent sends out solicitation the publisher's ai agent can respond with all of that information coming from its underlying systems ad server and inventory and data systems and now we can drive a lot more performance because while a human planning process can maybe take 20 30 40 rows in a spreadsheet with ai you can envision that that ai agent for the advertiser can go reach out to hundreds of different publishers so maybe the golf channel which is more of a niche audience has got great inventory for that campaign but because it was too small for a human to engage with you know it got skipped but now we can open up a lot more audience a lot more inventory and drive a lot more performance so we have launched a seller agent so an agent that plugs into publishers and can can integrate into their systems and be able to identify these opportunities and respond so we see that as a as another great growth opportunity around AI got it so back to

Rob Kulberth, Analyst — Evercore ISI

What's the innovation hand that we wanted to talk about? Just on Monday, you announced something called Decision Fabric. Programmatic decisioning logic sitting on top of EjeticOS in your environment. So maybe you could just help us dive in and tell us exactly what that means.

Rajeev Goel, CEO

So there's a lot of buzzwords in our industry, as you know, Rob. So really excited about Decision Fabric. You know, what we're doing with Decision Fabric is opening up the stack that I talked about earlier inventory activate direct buying in our ssp the agentic os the data partners and the infrastructure we're opening that up for our partners to be able to fully leverage and so what that looks like is a dsp a curator a custom algorithm company they can take their logic and they can run that logic in a container a container is an isolated environment sitting inside of one of our servers, and they can run that logic inside of our infrastructure where they can take advantage of all of the inventory that we have, all of the data that we have, all of the AI agents that we have, all of the infrastructure that we have in order to create a better outcome for their client. And why is this important? Well, in the current paradigm of SSP and DSP as it exists today, there's a lot of limitations that DSPs face. So one is that most DSPs don't see most of the inventory. We process a trillion ad impressions per day. Some large DSPs, they might take 30%, 40%, 50% of that traffic. A small DSP might take 5% or 10% because, of course, it's expensive. They've got to have servers to process all of this. So now when you run a container with Decision Fabric inside of our infrastructure, you can see 100% of the inventory. So now you're opening up a lot more users, offers a lot more impressions. Number two is latency. So in the current paradigm, end to end, we have about 250 milliseconds to process a transaction. And in the bid request going to a DSP and the response coming back, we might eat up about 150 milliseconds, so the majority of that time. Now, with the container and decision fabric sitting in our infrastructure, there's no network time loss. So that means the buyer and the seller have more time to run their application logic. And so that means you can use more data, you can apply more processing power, so you can create better outcomes. And then third is privacy regulations. So the direction of travel with privacy regulation, whether it's, you know, U.S., different states, GDPR, you know, Australia has privacy regulations, etc. et cetera, less and less data is now available inside of DSPs because regulators want to limit the spread of data across the ecosystem. Obviously, the source of data is the consumer on the publisher's website. That's where the data interchange happens, and the DSP is several steps removed. So we have access to more data inside of the SSP, and we can make that available in a privacy-safe way when we're doing that inside of decision fabric and containers. So, you know, really the opportunity, and we have four launch partners, and so it's early going, but I think it's an opportunity to rethink what does real-time bidding look like and how can we open this up to many more partners. And I think there's a future where a DSP will think about it differently. It won't be a monolithic system like it is today, but it might be an AI agent where an advertiser, depending on the different campaigns they have, the brands they have, They might be running dozens of different AI agents that are actually containers with proprietary decision logic that are accessing our data, our infrastructure, the impression volumes, and our AI agents.

Rob Kulberth, Analyst — Evercore ISI

And maybe it gets to your partnership with NVIDIA, which I do want to hit on, but this idea of running potentially the decision logic a trillion times, times the number of DSPs sitting on the platform. Give me a sense of, I mean, I would guess that's a big number in terms of, you know, the processing, the cost of processing, the inference, whatever. Give me a sense of what, you know, what you're going to be able to do other than you're not traffic shipping. You're showing everybody, you know, the sort of the full, the full QPS, the full hose. So why is that? How can you do that more efficiently with, you know, somebody residing within your platform?

Rajeev Goel, CEO

What makes that make it sort of work? Yeah, so so the way we have to think about that is there's a lot of infrastructure costs today For us to process an ad impression Send it out across the network to a DSP for the DSP to listen right meaning process that inbound Impression and then to send back a bid response Yeah, so if we add up all of that cost and we say okay now we're just going to process inside of the SSP inside of our our environment our infrastructure we can eliminate all of that network time the the back and forth that means we can eliminate a lot of the wait time on a cpu a lot of the networking bandwidth costs data egress costs all of these things we can offset we could say okay there's all of that savings now we're going to allow you to run on the full set of ad impressions which is going to you know take some amount of money to be able to do and because we own and operate that infrastructure and we use an extensive gpu architecture to parallel process we think we can do that in a uh profitable manner meaning the the the total sum of processing inside of us versus the sum that an advertiser or a publisher was paying between the ssp and the dsp is going to be lower and so we're going to drive better outcomes For our advertiser clients in terms of more performance more efficiency. We're going to lower the total cost We're going to generate more revenue in the process the advertisers are going to generate better outcomes And so there there will be money there to pay for them on it. So so big

Rob Kulberth, Analyst — Evercore ISI

Big overhead savings and part of you know key parts of what happens in terms of processing and impression and then

Rajeev Goel, CEO

Accelerated computing the beauty of accelerate computing. It's right. And and by the way, this is how walled gardens operate today right so there's no ssp dsp paradigm inside of a walled garden their single layer of technology that manages the inventory on the one hand you know think of meta and then they have apis where advertisers upload campaigns and they run that all in their own infrastructure and so they're you know they're you know great proof point of of the efficiency of that uh of that approach

Rob Kulberth, Analyst — Evercore ISI

i think you've made the point before that you know um bringing the the sell side and the buy side closer together, both in terms of the infrastructure and the ability to communicate data, all these sort of key inputs, that that can make this ecosystem much more competitive with the walled gardens. Maybe you could expand on that a little bit.

Rajeev Goel, CEO

Yeah, yeah, absolutely. So, you know, just picking up on what I just mentioned, right, when you look at the most performant ad businesses, ad platforms, you know, for sure it's the walled gardens today. and those walled gardens in part they're so performant because they're not trying to match users across disparate systems right there's no ssp and dsp they're a single layer of technology the user logs into one of those walled gardens meta instagram you know tick tock etc advertiser uploads their campaign and who they're looking to target and then you know that walled garden runs in that single layer of infrastructure. They can communicate all data about the user, about the context, about the environment, targeting performance, et cetera. We, on the other hand of the open internet today, live in this very constricted environment where we have an open RTV protocol. Only certain fields are allowed to pass back and forth. The SSP has to match the user with the DSP, so you see a fall-off in the users right away. Often that fall-off can be 50%, 40%, 50%, 60%. you see because of traffic shaping not all of the signal gets there we don't have a way in the rtb spec to communicate the click-through rate or the conversion cost or how a campaign is pacing so many of the basic parameters are not there and so with ai we're going to change that game with what we're doing with agentic os and activate with direct buying inside of the ssp we don't have any of these limitations anymore we can communicate whatever we want between the buyer and the publisher we can go well beyond the rtb spec which is which is in fact what we're doing and we're able to drive you know significantly higher performance and we're seeing that you know day in and day out um and i think at the end of the uh at the end of this process the open internet is going to be much better off because it's always had much better content than the walled gardens right the open internet is the home of you know paramount nbc disney right great streaming content It's the home of, you know, great mobile apps from, you know, Zynga to Wall Street Journal. It's the home of, you know, premium websites. So that content plus performance, I think, is going to be able to shift market share back towards the open Internet.

Rob Kulberth, Analyst — Evercore ISI

Got it. And here I'm going to present, you know, the devil's advocate section of the program. So, you know, if you were to rewind maybe 12 months or so ago, I think, you know, around the early development of ad context protocol which you you guys have

Rajeev Goel, CEO

you know implemented i guess is maybe the we're a founding member of adcp so we've co-innovated that with many other great companies in the ecosystem it wasn't entirely clear at that time

Rob Kulberth, Analyst — Evercore ISI

you know what the vision of agentic advertising was going to be you know something that that is built to integrate with the current programmatic supply chain paradigm or you know try to overbuild it try to displace it you know the the idea being okay we're going to replicate something like that I.O. system, you know, the young media planner or intern, you know, picking out sites on a ranker, but sort of infinitely long ranker until you get the right number of total rating points. You know, and then that, okay, you'd lose maybe sort of, you know, the precise efficiency of programmatic, but you'd also, you know, dramatically reduce, you know, the supply chain cost, if you will. Is anybody sort of seriously pursuing that vision of Agentic anymore? I mean, I think there was a little bit that was somewhat in vogue, like I said, 12, 18 months ago. But maybe you can tell me what's out there in the market right now.

Rajeev Goel, CEO

Yeah, I don't think there's any, or maybe there's few, if any, serious thinking along those lines anymore. I think we should think about this as some transactions are done via insertion order. Some transactions are done via real-time bidding. And there's good business reasons to have both. AI and agentic can sit on top of both of those sets of rails and really turbo charge those two types of transactions in a significant way. I took you through the, you know, the RTV process earlier, right? Planning, sorry, discovery, planning, activation, measurement, and we talked about how AI can meaningfully make that process more efficient and more performant. The same is It's true in IOs, the spreadsheets being emailed back and forth, AI agents can transform that. So I don't see a third type of transaction appearing. What I see is AI running those workflows and breaking down the silos of data, silos of systems where people have to work across different user interfaces. I see AI is really breaking that down. I have put out a prediction that by the end of, I think I did this two quarters ago, by By the end of 2028, I expect 25% of media buying in our industry to be agentic. And by the end of 2030, I think it'll be 50%. And two quarters into that, I stand by those numbers, and if anything, I may be conservative.

Rob Kulberth, Analyst — Evercore ISI

I'm going to put it on my calendar. So again, to play devil's advocate for a moment, has it been 20 years? 20 years, yeah.

Rajeev Goel, CEO

This fall, it'll be 20 years. You look exactly the same.

Rob Kulberth, Analyst — Evercore ISI

Rewind 20 years at this point Pubmatic, you know, I think was the innovation leader in the SSP space I mean, so I remember distinctly writing about the dynamic waterfall and say this is what an SSP is And people in the industry coming back to me and say correct me and say no, that's not what an SSP is That's what Pubmatic is. That's what an SSP would like to be, you know, the category writ large But you know So you guys have always been I think at the forefront of innovation at the space what's your plan to sort of make sure that you fully capitalize on your first mover advantage

Rajeev Goel, CEO

in agentic this time around? Well, I think there's a couple of aspects to that. So first is, I think we've established ourselves as agentic leaders in the industry. You know, it's early in that space. So this is not, you know, a lead that you can measure in quarters or years by any means so with that leadership position though you know a key part of a key reason to do that is we want to be at the forefront of customer conversations and innovation and if we position ourselves in the right conversations then we'll be on the the leading edge of innovation so that's that's number one is we got to position ourselves in the right conversations so we're leading the way in terms of these conversations and i think we're doing that number two is uh investing in uh the right team composition so I started by saying we're an AI native platform meaning AI internally and AI externally we're using AI intensively internally and changing the mix the composition of you know what is our team what does our organization look like in terms of where it's focused so as we deploy more and more AI internally and automate many different processes it's allowing us to increase the size of our sales team so you know if you look at our numbers what What you see is that headcount growth is flat to down in aggregate, meaning our overall headcount is flat to down, but actually our sales team is growing in size. So I think in an AI-driven world, human connection, human interaction is going to be even more important. And so we're investing more in our sales team to cover more of the market, more advertisers, more agencies, more independent agencies. uh and then i think the third piece is to continue to invest in the infrastructure and the data that's required to power uh this agentic revolution and so we'll get into the nvidia partnership but we continue to invest into our infrastructure and we're now at over 300 data partners uh that we work with on our platform just a couple of weeks ago we announced paypal and and Walmart connect select as new partners. So providing the data and the infrastructure that can power this agentic revolution. I think if we do those three things well, then we have a really strong shot at extending our leadership position.

Rob Kulberth, Analyst — Evercore ISI

Maybe a fourth one. Yeah, maybe this time around you won't be competing against a legal monopolist. But we'll get back to that.

Rajeev Goel, CEO

We'll get to that, yeah. That would help.

Rob Kulberth, Analyst — Evercore ISI

So wanted to ask about the NVIDIA partnership, absolutely. Can you help us dive in there, tell us a little bit more with the genesis there and what that means for the business today?

Rajeev Goel, CEO

Yeah, so we've had a multi-year collaboration with NVIDIA, and it's quite interesting because, you know, obviously we're not going to be NVIDIA's largest customer. You know, we're not going to move the needle in terms of the amount of chips that we can buy from them. But what we can do together with them is to co-innovate and really apply their hardware and software stack uh to the advertising domain in a you know domain or industry specific manner i think that's really the the root of the of the partnership so i'll give you a couple of examples so we talked about traffic shaping right that's the process by which we're processing a trillion ad impressions per day and we can only send a subset of those on to dsps because for cost reasons they don't want to they don't want to take all of it so key question that we face um on a continuous basis is which impressions to send to which dsps so this is an inferencing model right we have an inferencing problem we have a model we've trained that model with historical data and then as an impression comes in in real time we need to run inference and make a decision and so we have worked with nvidia software and hardware stack to to optimize inferencing on their architecture so that we We can make the best possible decisions at the least amount of cost in real time. So that's one example. Another example is as the industry turns towards more and more private marketplace deals and more and more audience targeting, if we get an impression that comes in and we have hundreds of data partners on our platform, which data should we attach to an ad impression before we make it biddable for a DSP? So that's again a heavy infrastructure problem where speed, parallel processing, cost, all of these things matter and so our collaboration with nvidia is really to identify industry-wide challenges and for us and them to co-innovate put engineers from both both companies on these problems and you know solve them uh in compelling ways for for our joint set of customers got it

Rob Kulberth, Analyst — Evercore ISI

so yeah i wanted to step uh into you know some of the recent history um i think we could pull it You didn't say, you know, you've had some dislocation DSP relationships over the past couple of years. One of those, I think, now seems to be a pretty solid footing. The other may be a work in progress. You know, if we read the trade press, we tend to believe, you know, DSPs and SSPs are at each other's throats or at this, you know, really heightened moment of strategic conflict. How overhyped is that from your perspective? Like, I mean, the day-to-day.

Rajeev Goel, CEO

I think the day-to-day, I think that is overhyped. But, you know, we look at it not from a DSP versus SSP conflict perspective, but rather how do we enable advertisers and publishers to generate better outcomes? It goes back to, you know, bringing them closer together. And so I think there's, you know, a couple of things I call out here. One is last summer we shifted our roadmap and our focus significantly towards diversification within our DSP base. We had a couple of DSPs that were quite large, quite concentrated, and so it was clear that we needed to make changes. And so we set that process in motion. You know, last year we added 50 new DSPs onto our platform, so a pretty significant number, and we shared in our last earnings call that that, you know, mid-market DSP cohort is growing at over 20% year over year. So we're, I think, executing very well from a diversification perspective the second thing is that we're seeing a huge fragmentation across the ecosystem of dsps right so whereas i think a number of years ago was all about concentration and consolidation within the dsp ecosystem now we see you know very specialized dsps that are going after specific verticals or specific segments of the market and i think part of that is that the some of the incumbents they've been focused on the fortune 1000 advertisers and when you look at those brands there's not a lot of growth in the fortune 1000 brands a lot of the growth you know go back to covid is record small business formation is coming from the mid-market upper mid-market brands you know that are growing much faster so uh a chime bank uh when you think about banking or vory when you think about clothing as opposed to you know fortune 1000 clothing or banking brands You know those upper mid-market brands are growing their ad budgets as they're growing their revenue And so there's a lot of growth there that really hasn't been covered by the the incumbent DSPs So some of the things that we talked about earlier decision fabric our AI capabilities The the data Partners that we have these are all Designed to help empower DSPs to grow their business on our platform and we're seeing a lot of good success with that which is growing our business and also solving for that diversification need that we have got it i want to change gears a

Rob Kulberth, Analyst — Evercore ISI

little bit talking about commerce media that's been another big opportunity that you've highlighted can you talk about what's driving some momentum you're seeing um you know for for both on-site

Rajeev Goel, CEO

and off-site commerce media activity yeah we're seeing terrific growth uh with commerce and our focus is is twofold it's uh both on-site and off-site you know every uh retailer or commerce media network, retailer that's participating in advertising, they are focused on, number one, how do I grow my on-site ad revenue, and then number two, how do I extend the value of my data and the value of my user base off-site, where then they're not constrained in terms of the ad impressions or the user experience. And so we built a number of products, our first focus was on off-site, given the capabilities that we have, where we have a fantastic set of commerce partners, PayPal, Instacart, Intuit, Klarna, just a couple of examples, where their data sets are available in our platform, and publishers can sell against it. Those data partners, those commerce companies can sell against it. Buyers, DSPs, can sell and target against it. And they're using that data to deliver a more performant ad experience so just to dive deep for a second into the paypal example you know paypal has 25 billion transactions a year they have 400 million uh you know paypal and venmo accounts so so that you know rich data in a privacy safe way is available on our platform uh for you know for all those constituents to use and when that data gets used we generate a data fee in addition to our ssp fee we might also generate an activate or agentic os fee so these are all incremental transactions a couple of weeks ago we also announced a partnership with walmart connect select which is their retail audiences and obviously you know they know more about uh consumers than just about anybody else their retail audiences they're now targeting those on our platform as they sell CTV, online video, display ad campaigns. So these are, you know, a set of accretive businesses for us where if you take them together, Commerce Media, Activate, Agentic OS, our broader data business, we categorize these in emerging revenue streams, and we shared that last quarter. That segment of our business is now 14% of our revenue, so we've been nurturing these over the last couple of years. Now they're very much material and at scale, and they grew 80%, 8-0% year-over-year last So, you know, key part of our second-half growth plan. Got it.

Rob Kulberth, Analyst — Evercore ISI

I also want to ask about CTV, you know, another big opportunity for the company. But you did see a little bit of deceleration there. I just wanted to ask you, is that, you know, related to the dislocated DSP relationship? How is the underlying business growing in CTV? You always made that just a sort of function of, you know, the sort of outsized growth and the expansion of the partners that you had last year. Maybe you could just sort of give us a sense of the dynamics there.

Rajeev Goel, CEO

Our CTV business is very strong. I think the quarter to quarter is just a little bit of noise in the numbers, but, you know, it's a very material part of our business. We work with 28 of the top 30 streamers globally, so, you know, all but two. And we continue to work on those two as well. But we're seeing tremendous growth in CTV, and it's a key part of why we've told investors you can expect double-digit growth from us in the second half of the year. Live sports is a key part of our focus. So we've been powering Cricket World Cup, IPL, Wimbledon last year, NCAA finals earlier this year. We're gearing up for FIFA World Cup soccer, which will bring a lot of very high value inventory to our platform. We participate in NHL, NBA, so just about every major sports league that you can think of. And that, again, is driven by AI, our agentic OS, by the data that we have on our platform, by the owned and operated infrastructure, given the spiky nature that can happen in live sports. So I think you can continue to expect strong CTV performance from us.

Rob Kulberth, Analyst — Evercore ISI

Got it. And let me check the date. It's June 3rd. And so I guess by my calendar, that's about seven months, eight months almost, since closing arguments in the ad tech trial and remedy space. Let's take it so long. And so seriously, hopefully we'll see something any day now. You guys have, you know, sized what a point of share could mean for the business, and it's pretty significant. But depending on what this is decided, you know, how material do you think that could be? And how quickly do you think that we could see some share shift in the market?

Rajeev Goel, CEO

Yeah, I mean, as you said, we're expecting a verdict, you know, imminently. And we think it's a huge opportunity. And there's two fronts to this. One is the remedies, and then the other is our civil suit against Google in the wake of the DOJ's securing a guilty verdict against Google. So in terms of the remedies, there was a lot of discussion in the remedies trial about structural versus behavioral remedies. Behavioral is, of course, Google changing some of its activities, its behaviors to reduce reduce the impact of its monopoly, the structural remedies would involve spinning out or shutting down certain parts of its business. So the behavioral remedies can be enacted, we think, quite quickly and, as you said, can be quite material. We've sized each market share to be $50 million to $75 million in revenue, and 80% of that That would be profit, given we're already processing the inventory from publishers and we're already getting bids and integrated in with advertisers and DSPs. So I think that could happen quite quickly after the verdict is introduced. Obviously, we've got to see what exactly comes from that. The second piece is our civil damages claim against Google. We filed a lawsuit last fall, and that case is proceeding in court in New York, and we're looking forward to having our time in front of the jury with that.

Rob Kulberth, Analyst — Evercore ISI

Depending upon what we do see from them, and of course, the EU is running its own process on some of these issues as well, to certain markets, whether you're talking about further the vertical integration, the ad server, or regional expansion in a bigger way, do certain markets become a bigger opportunity for you, are those things that you're beginning to Not just the incremental point of share, but just structurally, big chunks of market opportunity that could open up for you.

Rajeev Goel, CEO

I think there's kind of two ways to think about that. That one is the technology providers themselves, so to your point. And the other is around privacy regulation. So with some of these other markets, they're intensely focused on large US technology companies, including Google. And so there's changes that are being rolled out. right? EU has the DSA, you know, gatekeepers law that, you know, prescribes certain types of behaviors if you have a certain power in the market. So I think that absolutely creates opportunity for us. We're already pretty well entrenched in Europe and in many parts of Asia, except for China, and increasingly in South America. And so I think all of those markets benefits and then the second is around privacy regulation where you know many other countries are much further ahead of the u.s in terms of restricting the flow of data and given everything that we've talked about uh with activate agentic os you know data coming more and more into the into the ssp and being targeted there and the privacy benefits that that creates i think that's

Rob Kulberth, Analyst — Evercore ISI

another big uh tailwind for us uh well rajiv it looks like we're out of time so we have to leave there but you know fascinating discussion great thank you thank you so much for the time really

Rajeev Goel, CEO

appreciate it thank you rob appreciate it okay