The Internet Analyst here at ROTH Capital Partners. Welcome to our fifth annual AdTech Summit. This is where we talk about big picture themes affecting open internet. I think this year is, I hate to bring up this, but kind of an inflection in a lot of new things happening in open internet, and I feel there are opportunities and challenges that we should talk with all the ecosystem players. Importantly, this webinar is live and being recorded, and these are specific disclosures that compliance has asked me to share. There are six other sessions after this, and if you just Google ROTH AdTech Summit 2026, you'll see a microsite, and you'll see the rest of the sessions after the current one. There is Criteo later today, and then we are spreading the rest like peanut butter and jam on bread over the next seven days. Most importantly, welcome Tim and Chris from Viant. While I do that, like all my webinars, I'm going to launch a quick poll. For people who are online, please take it. What I'll do is, I'll close the poll at the end of this session and share the results, and we'll keep using the same poll over the next seven days, and we'll have a kind of a collective wisdom of people, and get a good pulse on what the sentiment is of investors towards a bunch of topics on advertising and advertising technology. With that, somewhat of a long-winded introduction, thank you again Tim and Chris for doing this. Thanks, Nick, for making this happen. Just opening, kind of for investors who may still think of Viant as a DSP, maybe how would you describe the company today? What are the biggest changes in Viant's positioning over the last, call it, 24 months or so? Yeah. Rohit, good to be here with you. Thanks for hosting us. Looking forward to a great discussion. You should have one disclaimer up there of just the Vanderhooks in general. We need our own disclaimer. Yeah, let's talk about Viant and how we've changed. There's been lots of changes in the open internet advertising industry. When I look at the DSP, we've really morphed the business into what we describe as an advertising intelligence company that has our own proprietary data, independent measurement, and media execution all on one platform. We've made a series of acquisitions around proprietary data. Some of your investors may know we acquired IRIS.TV, which gave us a content identifier. We are uniquely the only company in the space with a content identifier coming through. We've scaled that from only 7% of ad requests to now over 50% of ad requests, and we should be at 75% later this year with the new content owners that are installing that. Huge advantage when it comes down to content targeting. What we're known for when we went public was Household ID, our identity targeting solution. It works 95% in the CTV channel, 80% across all channels. I'm sure many are familiar with the LiveRamp acquisition. Very similar in the way that we approach. Ours is physical home address, and we attach the various device identifiers, but enormous scale. The last thing is just around independent measurement. Just recently, we acquired the second-largest U.S. TV panel, TVision. Most people are familiar with Nielsen in the linear space. TVision is an incredible new technology, started at MIT, and it's got a webcam inside of the TV homes that are part of their panel. We're able to detect unique things that didn't exist before, like is anyone in the room before the ad break hits? If anyone is, what's co-viewing? What are the demographics of the people sitting in front of that glass? Most importantly, are their eyes looking at the TV or are they distracted on their mobile phone? We can run through those. That's what I mean by an advertising intelligence company, Household ID, the content identifier with IRIS, and now all types of new data from the TV panel we just acquired in TVision. Okay, fantastic. That's a good setup. For people on the line, the way I hope to structure this conversation is I'll talk about, obviously, AI and impact of AI on you, on your customers, and on your engineers. We'll talk about agentic AI, we'll talk about data, which seems like a very big deal at Viant. Obviously, we'll talk about other growth drivers in the business. I guess just at a very high level, when you talk about AI, where are you seeing the biggest P&L impact, costs versus revenues, and where is it manifesting today, and where do you think 12 months from now you would be able to, say, point to one single number and say, "This is AI"? Yeah. I would say, two years ago we launched ViantAI. We led there. We did a public kind of webinar. We showcased the product, which is very unique because you're going to hear a lot of companies talk about AI, and you have for the last two years. It's pretty interesting that they don't show you anything. We launched ViantAI, and we've had great success with it. Today, we have kind of four phases of ViantAI. We have AI Bidding, we have a set of agents that do bidding on the human's behalf. We have AI Planning that does the media plans, a set of agents that do that. We have Measurement Analysis that does, instead of humans trying to figure out what they're getting for their money and get insights. We have a set of agents who deliver that. Our last one was AI Decisioning, that we recently launched at CES this year. We launched what we called our first fully autonomous product. Fully autonomous, meaning it doesn't require a human. The human sets the guardrails up front, but the AI takes over and makes all decisions, and that's called Outcomes. We've had great success with that. If I start at the top, AI Bidding that we launched about two years ago, that has nearly 90% customer adoption. What that does is it goes out and procures the inventory and the audiences, whatever the marketer looks for. It goes out and procures it for the best possible price. On average, we're saving greater than 35% compared to the human picking price. Tremendous value for our customers. They get cheaper CPMs, more impressions, more reach, therefore it boosts return on ad spend. We have our AI Planning product that's used by 30% of our customers. Why are they using it? Because it produces better media plans faster, does it in under 60 seconds, and it drives return on ad spend. We have about 30% customer adoption there. Our Outcomes product is really about us just aiming at getting into the performance-based advertising business. We can talk more about that. On average, compared to humans, it delivers customer acquisition costs 65% less than humans managing campaigns. Tremendous success there with our ViantAI product. Just to add to that, the impact on the P&L AI Bidding, we keep a percentage of the CPM savings. It's been a large contributor to our P&L. Since there's 90% adoption, it's been driving revenue for a number of years now and continues to deliver. What you'll see in the back half of this year and going into 2027, the Outcomes product, the fully autonomous product, is going to be an enormous driver for mid-size marketers and down. I would say they're adopting AI in a fully autonomous solution at a much faster pace than, say, Enterprise Fortune 500, which has a lot of governance questions before they will enable or allow a fully autonomous product. When it comes to a fully autonomous product, the thing I hear in my mind is it a black box or to what extent is there transparency and how much control does one have or have to give up? What pushbacks do you anticipate for your AI Outcomes product over the next 12 months? That was a great question. Typically, to frame it up for everyone, this product is analogous to Google's Performance Max or PMax they call it or Demand Gen. They have two products there. Basically all what Meta does in their system. Where ours is different is those are black boxes. You basically tell them what your customer acquisition target is that you want to spend, your budget, and they deliver those customers. We do it, but we offer a tremendous amount of transparency. We show you where you're running, not just the channels, the sites, the apps, the audiences. We give you that transparency. The human sets the guardrails upfront. You give us four pieces of info, who you are as the advertiser, you give us the URL of either the advertiser or the product, budget, flight dates, and what your CPA goal is. The AI basically just goes out and executes. It does give you the transparency that you don't get in Google or Meta. Okay. I got you. I guess, in terms of expanding the pie, is there evidence that your end market is expanding while your market share possibly is also growing as a result of the various efficiencies are being brought in from these products? Yeah. With our aim for the autonomous products is that mid-size marketer and down, who are we going after there? It's the 10 million advertisers on Meta. You're seeing AppLovin and some other companies start to put in their focus there. Certainly expands the market and our capabilities. When you have a do it for me solution, it's that small and mid-size business can really take advantage of it. Whereas today, they really only have one channel that they can use to market their products, which is Meta or sometimes search. We're seeing search continue to get deflated and lose market share because everyone's recognizing, if I'm buying my own branded search terms, these are not net new customers, they're the same customers that I've already won. You're seeing Meta's adoption grow, AppLovin as well is going after that, and our goal is to go after that same 10 million advertisers. Okay. I got you. From a cost standpoint, it seems kind of the flavor or investor sentiment changes on a week-or-week basis. Last year, it was all efficiency gains. This year, it seems like there is token maxing in 1Q, and now there is a hangover of tokens in 2Q and beyond. Where is Viant in that? Where do you see more room for efficiency from AI? How much, if a new engineer joins Viant, how many tokens do they get? Yeah. We don't have a set limit on tokens. It's kind of by seniority. We'll go kind of bottoms up on your question. Our chief architect has unlimited use, and most of the code by, I would say, our top 1/3 of engineers are being produced by foundational models today, and they have become professional editors, and how do we implement it from there? When we look at AI usage internally, tokens are up about 10.5x, but cost is down per token 78%. It's a blend of what you're doing. You have new reasoning models that get deployed by the foundational models, very expensive token usage. The smartest approach to that is to use open source models for one chunk, and then actually use the higher reasoning for our Lattice Brain, the decisioning architecture there. You have a blend of models and our general takeaway is that the commodity in this setup is the foundational models. The companies with proprietary data to train those foundational models on data that's exclusive to an organization, that's really where most of the value will accrete to. We call that the application layer of AI. I think most investors have this wrong today, that the foundational model is the unique one, and they're going to eat the world. They know they can go kind of in any direction. They lack domain expertise. They lack proprietary data that's critical to these transactions. We think investors will wake up to recognize, "Oh, the foundational models that compete with open source models really are the commodity," and that application layer is where most of the value will flow to. Awesome. I guess just switching to agentic AI, I think we already addressed some of the topics I had on my mind. In terms of truly autonomous, you mentioned there is adoption and AI, kind of the performance product is geared more towards SMBs. Where do agencies and large holding companies play a role in that? Is there a scenario where you could actually have a disintermediation of some of the agencies or how does their role evolve in this space as agentic AI becomes more prevalent? No, I think that there's currently a bifurcation in the space. I think it's going to remain for some time. The bifurcation that I'm talking about is when Tim said, what we're currently doing and who are the adopters of our Outcomes product, that's the performance-based advertisers. Most industry bystanders will say, "Okay, well, that's SMB." Well, actually, no. It's probably 65% of the total amount of money spent in the U.S., call it $400 billion. These are all companies. Where's that money going? Today, it's going to Google, Meta, and Amazon. That's performance-based advertising. It's all things performance. Those are the adopters because they don't care as much about control, they need performance. If you go to the other end of the spectrum and you look at the Fortune 500 or the Fortune 1000, those dollars that are managed typically by ad agencies, it's not that they're not looking for performance, but they do want a level of control that they really just haven't completely adopted yet. A lot of that money that agencies manage is what I'll call mid to upper funnel. They do manage search budgets, yes. Again, that money's going straight into Google. If they're managing social, it's predominantly Meta. When we think about the agentic opportunity, that I think is in that kind of mid to upper funnel. The question is, when does that budget kind of migrate to an agentic future? We think that that is a ways off, just because predominantly of organizational behavior, both in marketers and in agencies. The big thing that Tim just touched on is around governance. I think largely what you're seeing that's happening right now in the space, there's a lot of talk about it. We have many pilots going on, both with brands and with agencies, for what we call a third-party kind of agentic protocol, where they're creating agents that, let's say they're using Claude, and they want to use ViantAI. We are enabling that through a lot of proof of concepts. I think that that's where the industry is right now, is proving these out. There's a lot of work going on that still has to be done around industry standards. There's multiple protocols that are trying to be standardized in our space, but I think we're not going to see that in full on. We're not going to be seeing that for some time. I think that just to add to that, the only thing, what's changing, it used to be integrate via APIs. Now what the Fortune 500 is looking for is us standing up MCP servers for them to interact with their own instance of Claude or ChatGPT, depending on everyone's got their own specific LLM that they believe. Some are even using DeepSeek, as well, from an open source. It's really the introduction of MCP is where the change is happening. Okay. I guess, how far are we to a point where the holy grail of omnichannel platform for advertising and marketing where you magically, as in perhaps you're already there in certain pockets, but where you magically kind of manage campaigns across connected TV about all the various kind of channels that you have? All you set is one final objective, and the agents take care of everything, and all I see is transactions on my platform. When does that happen? I think we're there from a technology and product perspective. Certainly Viant is with the fully autonomous product of Outcomes already there. It's just the organizational behavior, the adoption, the comfort with AI. What happens if someone makes a mistake and you spend too much money or what, and you don't get the returns and the results? Who's on the hook for that? There's probably just some open-ended questions that people are still getting comfortable with in there. I think from a product perspective, we've been there for quite some time. It's really just adoption, and changing status quo human behavior that's going to take probably another year or so. Interesting. Okay. My guess is if that were the case, probably there is a pretty significant market share shift that could happen as people look at these channels holistically, and there is a significant uptick in ROAS from kind of perceived expensive channels that we know of, the ones that you mentioned, Amazon, Google, Facebook, where they take an arm and a leg. There is obviously a lot of low-hanging fruit out there. Yeah. Let me just add, there's one area that needs to be broken down, which is the walled gardens have walls for a reason. It's because they claim they drove all the sales when in fact they drove very little. That was part of our TVision acquisition, which was to give us the measurement capability that sees through those walls. We can supply an independent measurement report or let's just focus on attribution, what I just described there, across linear TV, open web streaming, YouTube, and Prime Video. Once marketers now are getting new product offerings like our TVision panel that sees through the walled gardens, they're getting much smarter on which transactions were going to happen anyway, which transactions truly were open internet versus a walled garden simply showing an ad when a transaction was going to happen anyway. As that level of knowledge continues to get pushed into these big marketers, you're going to see a reallocation of budgets. That's why measurement and attribution are so critical and why Google and Amazon and Meta don't allow anyone to measure their platform, is they don't want them to know the secret that they're really just targeting the same customers you had previously. Okay. Interesting. That's a nice kind of segue into some of the things that you talked at the beginning, your IDs, your identity, and measurement kind of capabilities. Maybe level set, if I understand, for somebody who's new to the story, there are three IDs you have, Household ID, then you have the IRIS_ID, and now TVision. How do they come together? Actually even before that, how does Household ID compare to UID2 as such? Yeah. I'll start on Household ID. Household ID, as Tim said, is your physical home address. That's what we do. We basically look at all the adult individuals, all their names, their emails, their phone numbers, and we link that into an address. When that then comes across the bid stream, we then know who's in the household. That's Household ID. You have the IRIS_ID, which is the CTV content, right? Most people are completely unaware of this. Today, when a marketer goes to buy CTV in a platform like in any DSP, you're only able to buy at the app level. You can only buy Paramount+ or Netflix, but you don't know, is it "Bridgerton" or is it NFL football in Netflix? You have no idea. We saw that that was a massive gap, that's why we went out and acquired IRIS. They work with the content owners. They index every video in their library. Some of these, like Paramount has 25,000 videos in their library. Each marketer, one message may be right for NFL football but isn't the right one for "Bridgerton". Understanding that intelligence, what we call the content intelligence, not just the show name, but all the stuff around it, the actors in it, the contextual, what's happening in the scene right before the ad break, those are areas that we know drive big campaign performance. The scale of Household ID is up so dramatic this year, over 50% from 7% when we acquired it in 2024. Of IRIS_ID. Of IRIS, sorry. When we get a bid request, not only do we know the household and we know who's in it, we now know the exact content right before the ad pod is going to come, so we can then match the ad to the content. Then now with TVision, we have the panel going in real time. We link the TVision attention scores to that content ID right before the ad break, and we know how much attention is being paid to this program. Is it 40% or is it 75%? That'll change what we actually bid, and it's what we call our attention-adjusted CPM. Marketers will bid up for higher attention, bid less, or maybe not bid at all for low attention. Just to put that into a real-life example, let's say the advertiser is Cabela's. They sell outdoor fishing, hunting, things like that. If we see, if you're in The Trade Desk, you may have an email address of the logged-in user, and that is all the targeting capability that you have there. With Household ID, yes, we can identify, okay, which ones are outdoor enthusiasts? We see the content. Okay, this is "Landman." It's not a random impression in Paramount+. It's actually "Landman." In "Landman," right before the ad pod, it is a fishing scene, and they caught a really large fish. We also know the TVision data. There's two people in the room, both are males, 18- 54. So we know the value of that impression might be worth $150, but we're able to clear the auction and buy it for $30. The return on ad spend to Cabela's, given all that proprietary information, is a massive value increase. Okay. I guess with regards to having all three IDs, how would you have investors think about your ability to outperform, say, a The Trade Desk or some other kind of peers like that? What is it that is irreplaceable that makes your value prop even stronger? I think it's pretty stark. If you look at Household ID, just on average, let's just say not CTV, it's at 80%. UID2, which is the The Trade Desk solution, is only at 20%. LiveRamp, by the way, to put it in perspective, is at 37%, just in terms of penetration. The marketer believes that when they come in, they want all their ads to be addressable. In the The Trade Desk, it's only going to be addressable 20% of the time. If you're using LiveRamp, it's only addressable 35% of the time. If you're using Viant, it's 80% of the time. We're going to show relevant ads right up front based on the household demographic makeup and who they are. We're going to be doing that at a rate of 4x that of The Trade Desk. Again, knowing what content the household is consuming at that moment is a huge driver in performance. It's not even close. The Kardashians is not going to perform well for Cabela's compared to "Landman" scene that Tim just described. It's going to drive much greater attention, and we see the downstream sales impacts from that. Again, it's another leg up on performance. Then, of course, the TV side, are people paying attention or not? If someone's in the household and an ad plays and no one's paying attention, well, the ad never actually happened. You paid for it, but it had no impact. Interesting. Okay. I guess you mentioned LiveRamp, you mentioned The Trade Desk a couple times. Maybe sticking on LiveRamp, the acquisition of LiveRamp, that has created a lot of discussion around identity infrastructure and how do you think that deal changes the market, and how does that impact you guys? LiveRamp was a great acquisition by Publicis. It gives them a proprietary identifier. They've attached that identifier to a data marketplace of many third-party data companies. We offer something very, very similar, but just at twice the scale when it comes down to it. For Publicis, it was a great acquisition. If your name is Omnicom, WPP, Dentsu, Havas, any of the others, it was a bad acquisition for you. What you're going to see, and we've already seen the rumblings of other holding companies, other large independent agencies looking for an alternative, which opens up the door for Viant as well, if LiveRamp was your provider or your preferred choice that was there. I think what you're going to see, although a great acquisition for Publicis, on a revenue multiple, it's probably much more expensive than when it gets marketed when you fast-forward one year and you see the churn of customers. There's lots of grumblings in the space on how many new customers of LiveRamp are looking for an alternative solution, and I think you're going to see this play out over the next year. I would largely say UID2 is not considered. It's just an email-based identifier. It doesn't have the penetration in the marketplace, and I can't describe it enough how bad The Trade Desk product is in-market when it comes to streaming in CTV. They probably have a last-place product, which is really driving their financial results south. When it comes to making decisions, TV is about half the budget. Our thesis is whoever has the best product in CTV is going to win the omnichannel DSP buy. We don't believe SSPs are a threat. We have really very few competitors. The Trade Desk, Google, Amazon, and Viant. Those are the only real enterprise-grade DSPs on the market. When you look at Amazon and Google, they have a natural conflict of interest. Their whole goal is to tell you that YouTube drove all the sales, or vice versa, Prime Video drove all the sales. They're very, very expensive. Everyone claims Amazon DSP has a 1% take rate. It's really not true. They run 80% of your money within their walled garden of Prime Video, and they have a 100% take rate. That conflict of interest, they're doing what's good for Amazon, not what's good for Cabela's. When you look at the buy side-only players, there's Viant and The Trade Desk. The Trade Desk has a subpar product in-market. They don't know where the content, they don't have a content identifier. They don't have a user panel showing eyes on screen attention or co-viewing or demographics, all of these important critical signals. We think Viant, we have very little competition. We have the best product in market and I think you're going to see wild market share swings, just like the acquisition of our new customer, Molson Coors, who considered The Trade Desk, Amazon, and some others, and came out that we were the best product in streaming. Very cool. I love the pitch and really love the energy there, Tim. I guess we are at 30 minutes. Last 5-10 minutes, I want to talk about secular and cyclical growth. More on secular growth. Where is the next wave of open internet advertising going to come from? Is it CTV, retail media, you have mobile, you have AI advertising? Or is there some new chunky thing that you think that not just overall ecosystem, but Viant could also benefit from? Yeah. Well, it's important for investors to know where we're at in the adoption of streaming. There's still $50 billion being spent in linear TV that is going to provide enormous tailwinds for streaming over the next couple of years as that continues to degrade and streaming continues to rise. That is probably the number one tailwind. You touched on retail media, bringing in commerce transactions to get a better understand of what products and services. That is really just unlocking in streaming as well, and we're very early there, and it will bring in a whole another wave. Where AI provides the benefit is it's the orchestration and decisioning layer to consider all of these data points. When you construct an ad campaign, we did the math, there's over 93 trillion ways you could combine data, media, creative formats, channels. The humans are just far outmatched in how many choices that are. AI adoption providing a third leg of performance, whereas a human used to log in two to three times a day, adjust campaign settings, pricing, channels, formats, data use. You have, every second now, AI making decisioning on your behalf, driving actual performance outcomes. It's not emotional. It never sleeps. It never has to drink. It is running 24/7, always on, and has the ability to look at all possible combinations, which is providing a third leg of growth. I just want to remind investors, we are still AI very early, where you have governance questions. Even linear coming into streaming, we're still only third inning of the money flows. Okay. Very cool. I love it. To baseline that is very important. People think this industry has been around for a very long time, now we should be at a point where secular growth needs to come from new and incremental opportunities. Just more of the same is still something that you can grow 2x, 3x from here on out. In terms of cyclical things, I know you probably won't talk about impact of geopolitical things, but just second half, we have a lot of new cyclical drivers. How do you play into that, like elections, sports, lectures on? I would say at the top, I would put political out there. We previously stated on our earnings calls, we expect that to be a 200-300 basis point tailwind for us. We don't think it'll be as big as the presidential cycle, but we think it's really healthy out there. You have major sports things. You have World Cup kicking off here. That also brings in new dollars into the space. I think the setup for the back half, I'd say industry-wide is pretty good. As far as geopolitical and things like that, we haven't seen it yet. Obviously, the longer that this goes on, with fuel prices staying elevated, I don't know. If it goes on forever, there likely is some impact. We just haven't seen it from our customers. Okay. There are two new players, I would say not two completely new players, but if we talk this time next year, probably there is TikTok with its U.S. incorporation and ChatGPT, they're looking for new places to make money to support their valuation. How does that impact open internet and Viant, if at all? Yeah. ChatGPT, the way I would describe it is it's primarily going after search budgets, which was all behind a walled garden. It opens up a brand new set of money that was previously closed off to open web DSPs. Yeah, certainly going to be an amazing brand new channel. I would call it enhanced search, is the best way I would describe it, where you do have more context of what that consumer, because these are long discussions that you're having with the LLMs, and it opens up what was search money will be migrating to ChatGPT. Obviously, Gemini at Google will stay behind the walled garden. The consumer adoption of ChatGPT with hundreds of millions of consumers that are active users of this, and their adoption of the open internet or open web players like DSPs like ourselves, actually bringing money to them. It's all brand new money that we previously were locked out of and should be yet another growth driver for open web DSPs. Okay. Very cool. Yeah. This is a topic that we'll probably talk more over the next 6-12 months as how that balances out some of the market shares that have been pretty asymmetric with the time spent and whatnot, no matter which way you put it. I guess that brings us to 35 minutes. I'm going to end the poll and share the results. While I do that, just to keep it light and fun, I love to do this word association at the end. First word that comes to mind when I say something, without blinking your eyes, Chris and Tim. We are in an AI bubble right now. Yes, no, maybe? Yes. Foundational models are a large bubble. The data centers will never get built. It lacks the energy production. We own a solar farm. To get energy onto the grid takes three to five years. All the timelines are drastically overstated. We did that in 2020. I had no gray hair. Tim had hair. That's how long it took. Okay. I think we could do this all day. Human media buyers? You can't compete with AI. It's impossible for humans to even play a role at some point in the future. Okay. Token maxing? Token maxing on open source, great. Token maxing on very expensive reasoning models are going to be decreased. Okay. Last thing I have to ask you, Viant? Viant, bullish as could be, baby, on the Molson Coors big win coming on, linear TV coming to streaming, and all the proprietary data and measurement capabilities we have with the highly strategic acquisitions we make. I would classify what's going on right now as an old industry term, but it's Pitchapalooza right now. I've never seen so much business that is on the DSP side that's looking for a home, they're certainly not looking to go to another walled garden. Okay, fantastic. That's great. I closed the poll, for people who are still online, I'm just going to read out. How worried are you about a recession? It's kind of balanced. I don't see a specific, but there are people somewhat concerned as well. Incremental growth driver in advertising, connected TV is a clear winner. Which AI applications to advertising and marketing in their ad targeting seems to be the clear winner. Where do you see meaningful consolidation or M&A? I think identity is number one, then maybe there's independent DSPs, AI workflow. Finally, which company do you think is best positioned in the AI-driven future of advertising? I think two top would be identity data companies and CTV companies. We'll get more results as I do more of these sessions. Most importantly, thank you, Chris and Tim, for spending the time with us. We'll have a replay, we learned quite a few things, I think more people should look at Viant and the opportunities ahead. Thank you very much. Thanks, Nick, for making this happen. Next session is with Criteo starting in about 20 minutes. Thank you. Thanks for having us, Rohit. Great to see you. Thank you, everyone.