Okay. All right. Thank you everybody for joining us here at the end of the day here for our Age of AI tech conference here, the August version. We get to have a fireside chat with Nick Kormeluk, who is SVP of Investor Relations, and I think you are still in charge of real estate, for Magnite, which is a leading SSP, about $4 billion market cap, with a little over $300 million of debt and cash, basically. We cover the company as our top pick with a buy rating and a $40 price target for a stock recently north of $24. The shares have had some volatility and sympathy with post-SaaSpocalypse relief rally and also some positives in the latest earnings report, up about 50% year-to-date after a 15% dip last year.
Our buy rating ascribes a ton of value, we would say about $20, to the benefits of antitrust remedies and private litigation to recover antitrust damages from Google, and another $20 for the core business at a beatable 11 times EV to EBITDA multiple of 2026 E. Magnite has reasonably, we believe, argued for $50 million of essentially sales and EBITDA lift, given limited variable costs for each one percentage point of share shift from 60% share SSP Google to 6% share Magnite, from antitrust remedies. Given that Magnite is 15% of the non-Google SSP market, we see several percentage points moving over time as behavioral remedies are implemented. Behavioral is what is key here, as that can be implemented immediately, even as appeals are outstanding versus structural, which in other words is a spinoff, which can be appealed, and so getting to that takes time.
With uncertainty, it may ultimately not be the answer. We have argued that programmatic advertising is an auction market and in a fair auction, no intermediary should win 60% of the time, and it is reasonable for Magnite as the next largest player to be the biggest beneficiary of share shift from the obvious rule changes to make the auction more fair, like giving everyone equal access, equal information, and equal fee structures. A remedy decision from Judge Brinkema had been expected by now in a proceeding that wrapped up last fall. We think we are different than other analysts in really leaning into antitrust. Some other analysts argue that this kind of thing is uninvestable, and we say no, it must be part of the thesis at a minimum for auction value, and we argue for more than that.
It is important to note that any earnings lift from antitrust remedies is not in guidance or consensus, and so would be entirely incremental to the estimates. We are going to start with antitrust, and then we are going to talk about the fundamentals, which we also see to be constructive at Magnite. Nick, to start off with, just remind everyone, what did Google do to be found to have abused monopoly laws in this market?
Yeah. No, great question and great setup, Barton. I really appreciate this and all the rest of you that are joining in the throes of August here in summertime, especially staying in overtime in New York. Hopefully you had your shot of espresso to keep you interested and intrigued in ad tech here post 5:00 P.M. Yeah, I think from your question directly, what's Google done here to get the monopolistic ruling? This ad tech trial which Judge Brinkema headed, really focused on what they did in their network business and how they gained an advantage in tipping the scales in their favor, from an exchange perspective and an ad server perspective. Effectively, Google took over with their DoubleClick acquisition, and subsequent to that, they basically ran the table on ad serving in the open internet marketplace.
They did so by making the economics almost unattractive for anybody else to be able to come in and compete. As a result, they're really the only ad server in town for non-connected television. In doing so, the economics that they gained was really done so at the exchange side and the demand side, without the publishers having any choice or control over doing so.
They really did prefer their exchange by providing that timing of submitting a bid, information about the auction, clearing prices, the ability to bid in unique cases where the rest of the market was held to different standards, to really shift market share in their favor. It's estimated that their market share position's about 60% in that marketplace, as Barton reflected, even though we're a 6%, 7%, 8% share of total market, our fair share is 15%-ish , which corresponds with their 60% market share position. This really only touches the best top and mobile parts of our business per breakout, which is the DV+ side. Again, that's really the only piece that this would touch or impact. There is nothing related to this in our expectations, guidance, forecast, et cetera, conversations that we have with any of our sell-side or buy-side analysts to do so.
I know, and we really appreciate Barton's work on this. To date, he has not built in this to our expected numbers for 2026, 2027, 2028 and beyond. It's not something that we would stick our neck out there to describe. But we think there's a fair amount of market share that's up for grabs. What effectively, by giving the supply side or the exchange side an advantage from the ad server and kind of playing unfairly with the rest, they've been able to charge nearly a 50% take rate for servicing that inventory, while at the publisher level, making it seem like it's almost free because they don't really charge much for ad serving. By doing that's what they've done, to skew share.
We think in a fair competitive auction, as much as 30%, 40%, 50% of their share could be up for grabs and up to win. It depends on exactly what rules Judge Brinkema orders or puts in place to break the monopoly. But effectively, publishers have not been with that ad-serving position, they have not had a choice to move it from them to somebody else and do it at any reasonable economics for ad-serving and to break Google's stranglehold over how they transact and how much money flows to them from Google at whatever take rates they want. Another perfect example of that bargain is that Google, as an SSP, does not even have, or an exchange, does not even have contracts with publishers, meaning they do not have to negotiate take rates.
It is a take it or leave it. They either take Google's money or they do not.
If they choose not to, they are not sure if they would receive or cover it in any other way if they decided to exclude Google from participating. So they do not even have that choice. As a result, you would have no paper, no contracts where every other exchange that operates that has to access this inventory through Google's ad server has to basically make do with whatever Google decides to give you as an ad server, which is imperfect information, sometimes past the time that the auction is even completed and finished and a winning bid is returned, and you cannot do anything about it. So that is the monopoly is really you have no choice, no ability to negotiate fees, no alternatives to use in the marketplace, that that is truly where we believe these behavioral remedies should be aimed and really to treat things fairly.
There is a very simple, elegant solution by which to make this market perform fair. It is called Prebid. It is a bit of code. It is an industry standard. It is free, open source software that allows this to have Google's ad server pass the impression level or the ad request over to Prebid and then share it with Google and everybody else in the ecosystem at the same time, with the same data, with the same transparency. We believe that simple fix is truly what jump ball for amount of this volume to be able to transact openly and in a more fair manner. That is really the easiest description that we can use to identify what comes up for grabs.
There is a number of things on the demand side that they use to leverage and gain an advantage and how they use AdWords to buy or how they bundle this with other inventory that we also think is advantageous. But the cleanest, easiest example is simply sharing that impression, give it to us at the same time with the same amount of value, and for doing so, we think the market opens up.
Okay. That is a lot, Nick, but that is great. I want to just ask a very simple question. To what degree are they still doing these things that are abusive monopoly practices in your view?
In our view, we have not seen anything change from the behaviors or the patterns of which Google has acted and performed. In fact, even where optically you have things that have been volunteered by Google and they admittedly said, which for our civil trial would be very damaging and very beneficial in our case, which is, "Hey, we are doing away with Last Look." Last Look was a feature that two, three years ago they had in place that would let them steal any auction they wanted at the ad server level, regardless of who else competed, bid for it, won an auction for it. They could steal any impression they wanted for a penny more than what the winning bid was.
Why put everybody through the exercise of running an auction process to be able to steal it away for a penny at the end of the day? We even believed when that was taken away and advertised that they were killing Last Look, which said, "Hey, by the way, historically, we were really bad actors and we really did perform poorly." So there is a damage period of time there. There was actually a workaround because the minute they took that feature away, nothing changed.
Auction win rates, ability to compete in those auctions, nothing at all changed to our discernible patterns for what Google won and what we won and what their market share position was, which basically just spoke to there was a workaround put in place, and it was just optics by which they were hoping to appear to be good actors in the ecosystem versus truly doing something to open up a fair share of market.
Okay. To be clear, you're inferring that there was a workaround given that nothing changed. You don't have the workaround. You don't have the smoking gun that says this is the workaround that they did to put the lie to their offer of improving it.
They did numerous things. I think as part of this, the reason it's probably taken as long as it has is there's so many different complexities and wrinkles. There's not just one thing that you can do. So things like—
Yeah.
—universal price rules, which allowed them and them only to be able to pay beneath publisher price floors and bid on inventory and exclude everybody else from it. Or claim that nobody else was bidding on auctions if they send it to us late, which was very demonstrable evidence. I'll go into that in a second. But in those particular cases, moving things from an open market to a private market by showing that they were the only bidder, but they rigged the auction to do it. So evidence of malfeasance and bad behavior was when we actually ran tests of paying for a $1 banner ad, and we bid $5 and lost five times the market clearing price, $10 and lost $20 and lost the impression, and showed that to a publisher.
The publisher found that our bid was submitted after the auction had already closed, and we were asked to participate in the auction after it was already closed. So effectively, we were participating in something that never really existed but had the appearance of making the market look competitive. So when the publisher realized that we were willing to pay 20x the market clearing price and that bid wasn't making it through and wasn't leading to their realizing and capturing that revenue, they knew something was amiss. That was in the court records in discovery as we went through the trial.
Okay. You brought up Prebid, which is something that also came up in the trial where I think Judge Brinkema noted that, and the questioning seemed to indicate that she's certainly aware of what you just argued, that this is a way maybe to fix it. I think the markets have not really priced this into your stock because the markets have a really jaundiced view of antitrust, right?
Yep.
That's because there's a long history of antitrust and tech not doing anything. Most recently in Google was search, where we had Judge Mehta put out a remedy that was disappointing to many who wanted more, including the publishers. Just to reiterate, you think that this is much more fixable than search is just because of the structure of this market?
I think so. If you kind of dig in a little bit on what was found in search and what was found in this ad tech trial. In the search trial, it was found that Google exemplified monopolistic behavior. In this ad tech trial, you've got, are you or are you not demonstrating that behavior and how much of a stick can you carry around changing some edge behaviors? Right? Using one position to try to improve another part of the business without a direct influence is something that's guilt by association and proximity, as opposed to what was found in the ad tech trial, which was a guilty verdict. Guilty of monopoly in ad serving, guilty of a monopoly in exchange.
The fact that that was a much firmer ruling, at least from the beginning, we believe informs that there is a broader set or a more formative set of remedies that needs to be put in place to clear a guilty verdict versus guilty of monopolistic behaviors, which seems to be a lot looser in its formal teeth and formal description.
Okay. I just want to note one thing. One of the questions in here suggested a part of the preamble that I had may have gotten cut off and the question was, what was our valuation framework? We have a $40 price target. $20 is the fundamentals, $20 is antitrust, of which half is improved market share from remedies, and half is damages, basically settlement. All kind of net present value, assuming this happens sometime in the future. None of that is in our estimates for earnings for Magnite at this point. In terms of the remedies that this could apply to, you mentioned that this is relevant to DV+. I think one of the questions that you could put out there is that market still meaningful?
Given what is happening with zero-click and AI reducing traffic to websites, do you think that there is still an opportunity for you guys to get benefits in that market, or is Google basically killing that market so badly that there might not be any damages left to move in your direction?
Yeah. As much as the demise of the open web has been kind of published and predicted, the open web still is very sizable, carries a lot of advertising and performance dollars around it, and still plays a very important role in the ecosystem. That being said, is our overall traditional webpage views, open web views, and mobile and in desktop down? Yes. Does that trend continue to be down? Yes. But it is a massive market, right? So when somebody has 60% share, that is likely to redistribute, and we are sitting at 6% share, I would say that those shared characteristics are still massively attractive, even if the market is flattish to slightly down in open web. But there is other parts of that market that is not necessarily just an indication of how our overall DV+ business is doing.
That is a part of the open web, which is a challenging part of the DV+ business. But we have very other healthy places of our DV+ business, which includes mobile app, it includes commerce media. So there is a bunch of advertisers that are looking at very strong-performing, high-performance areas of DV+ that I think are going to be future continued growth drivers in our business. So that it is massively still attractive to where we are going.
Okay. Judge Brinkema, as we noted at the outset, she hasn't ruled yet. I think the proceeding kind of finished in the fall of last year, late fall, and I think people have been expecting a ruling by now. I think one of the questions is: Is there a point where she delays so much that the whole thing just becomes moot? Is there anything to say about that?
Yeah. I don't think this is something that can fade to black. Once you have a guilty verdict in place, you need to kind of finish and tie a bow on it, which is put in place remedies to fix what your actual ruling is. It would be a really bad look if the ruling was guilty and then nothing was done against it. That's something that I think would face appeals from both the DOJ and also something that would be important as part of the numerous civil lawsuits that have been filed against Google, where they've admitted doing very bad things over a long period of time. Even saying that they're no longer doing these bad acts anymore is very damning in civil litigation, and we're in the discovery period for that.
I would ascribe a tremendously low probability that nothing happens, combined in both Judge Brinkema's court, as well as what happens in civil litigation as part of this. The number of simple dollars, if people don't put dollars to work suing Google unless they feel like they have a very formidable case. The number of people, I could understand if there's one or two, but the amount and the volume of suits filed against them says there's a lot of parties that feel injured and feel very validated putting real shareholder money to work with a high ROI expectation against it.
Yeah. No, it certainly is clear. There's a number of SSPs and also publishers, and some of the largest ones, including you guys. In terms of the remedies, how quickly do you think remedies could be implemented, and how quickly before Magnite could see benefits?
Yeah, I think that's where the excitement is. I think, if there's a structural remedy, that's something that would be appealed and stayed on appeal for a period of time. Certainly Google has a chance to appeal anything, whether it's behavioral or structural. The judge has made it painstakingly clear that she's not going to necessarily wait for the behavioral remedies to be implemented until after appeals are over. Clearly if they win on appeal, then you can reverse some of the behavioral remedies that may be put in place quickly and immediately in order to open up the market. I think that things like running code through Prebid to share it externally, that could be done inside of a week. Right?
There also is the thought of how do you govern something like this or have a watchdog that can properly make sure that this rolls through, is implemented, doesn't have unintended consequences. I think prebid.org could also serve us as that body that represents publishers. Prebid is really based on thousands of publishers that have implemented that code as their operating system for their ad stack. I think you both have an org that represents publishers broadly, in open web, less so in CTV, also represents code that they use and have built their stacks on. Technically, and from a management overview perspective, I think it helps. I think to your point, I think Google has argued that in a period of time that the open web will be dead.
I think that almost hurts their argument and they've got a lot of, I think, future battles on antitrust that's coming ahead to them, relative to what some strategies may or may not be in AI or deals they may or may not want to do in AI. If this market truly is less strategic, less important to them, and it's the network business that's either been flattish or in decline without really spending much time against it, that Wall Street really doesn't appreciate as some of the other areas of the business. Is this something that is a give from them to, in the appearance of the DOJ, to clear a path for them to do what they'd want in some more attractive growth parts of their market?
Okay. All right. Just to reiterate, if other people want to put some questions in, there's an opportunity to type it into the Zoom interface here, and I'll see it. Let's switch to the fundamentals. CTV is really the standout now fundamentally, with contribution ex-TAC for CTV up 36% year-to-year in the second quarter. It's 51% of the mix, accelerating from the 30% growth of the first quarter. What's going on here? Why is this accelerating? Why is it growing so fast right now?
Yeah, investors in Magnite have been tortured for a period of time for growth to come and acceleration in CTV. We're at least getting mad if we haven't received something on the Google front just yet.
Right.
The laundry list of partners and customers and expansions that we've signed from a Netflix to a Warner to a Roku to Disney to those expansions and ESPN inventory coming in, those have all been signed, and investors rightfully so said, "Hey, where's the growth? Why isn't growth inflecting? Why aren't things happening faster?" We've always said it's a journey, right? People kind of turn the dial, and when they see that programmatic's working, they really start to accelerate that movement. The reason we're outpacing overall CTV market ad spend growth, which is growing low double digits, maybe low teens, is really programmatic and it's having its day. The market is moving more to programmatic, and Magnite is positioned much, much stronger in programmatic than even in regular, the traditional CTV market.
I'd venture to say, our share position in programmatic is probably even close to 2x, kind of our 20%, 25% market share position. You see that because we're the only programmatic partner by which to access Roku's inventory and Disney's inventory and Netflix's inventory. It puts us in a lot of the TV OEM inventory. The fact that we are the gatekeeper, the on-ramp to access that inventory, that's something that's critically important as the market embraces programmatic more and really plays to our strength. The other thing that's happening is, I think you probably have heard me in an earlier small group meeting say this, having inventory by itself does not matter. Inventory by itself has an expiration date, and in our case, that expiration date is inventory expiring. If you're not watching a show that's playing, that inventory's gone.
You need to have demand, and really what's happening in CTV is demand is finally coming in. Inventory is becoming not just top of funnel, it's going more mid-funnel, in some cases, performance. Not the large majority of it, but it's starting to move in that direction. You're starting to go from just enterprise and just large verticals now into mid-market, now into SMBs. It's performing, and as a result, a lot of the people bringing CTV demand into the marketplace is widening out. It's not just the land of two large DSPs and a few ankle biters as it was a couple of years ago, where it was just going to be Google and The Trade Desk. Now you have Amazon, you have Viant, you have Mountain, you've got a lot of other new entries into DSPs.
Google's DSP is continuing to grow, not just in their own inventory, but buying across other CTV channels and other inventory, bringing in new types of buyers. You have agencies that are standing up their own marketplaces that we're powering. You also have a lot of these publishers who view themselves large enough and scaled enough to create direct self-service opportunities and on-ramps to their inventory. We're the ones that are powering and creating those marketplaces. Said another way, we're not just tied to one or two DSP partners driving our overall results. We really have no blind spots to any demand as it enters into the CTV ecosystem.
As it wants to access CTV inventory, we are the easy on-ramp button to be able to get to the world's CTV industry supply, and that's what we've spent the best part of the last seven years building out and making sure we could handle and we provided the mediation layer, the ad server, and the primary sales channel for all of that inventory. Now that's starting to pay off as the market moves in that direction.
Okay. Now I think the core to your CTV offering is your technology branded as SpringServe, which includes an ad-serving platform. Competitively, I think the CTV ad-serving business has some other players. FreeWheel would be one, owned by Comcast. Publica, I think is another. You guys. Can you talk about respective market shares and what differentiates SpringServe relative to these others in this market.
It really goes back to my prior answer, which is programmatic. I think where FreeWheel really had its day was really with the large broadcasters as they moved from web-based TV into early days of connected TV, where the complexity and the ads run to different people for the same broadcast was very limited. Said another way, if you were running an ad or the ad server for a big broadcaster, everybody saw the same State Farm ad, everybody saw the same light beer ad, everybody saw the same Chase commercial, everybody saw the same BMW or Ford ad. It's very easy to run the same ad to everybody else. As a result, the core basic definition of an ad server is just stitching in a 30-second scene in between the rest of your content. It doesn't take rocket science.
Every publisher, every streamer has the capability to stitch some content in in between two other scenes. However, when you start to go to programmatic and you start to make that dynamic and you are starting to bring in other marketers, and you are doing that in real-time, the complexity explodes. Think about watching a game, an NBA Final that goes from 5 million viewers at the start of the game, drops down to 500,000 at halftime, it is a blowout, and then all of a sudden it goes to overtime and you have 45 million viewers. You have inventory, it goes to overtime. You never had any idea you would have inventory to sell.
In real-time, you are trying to run a million different types of ads to people of different types of audiences, whether it is gender audiences, income audiences, regional audiences, global audiences, doing that in real-time and monetizing every single penny for every single ad that you could possibly run. That explodes the model and breaks down what non-programmatic was really intended for. A FreeWheel, probably still of the broadcasters, has a 75%, 80%, 85% market share position. Outside of that audience, either for core ad serving or for programmatic ad serving, that is a layer on top of core ad serving. I would say we probably sit on the reverse of that. We are probably at 80% of the programmatic ad serving market today.
The real exception that we had, Publica, which was IAS owned and then owned by private equity, and now not as visible because they do not have to report their quarterly numbers. Their largest account was Samsung, and you probably just saw about a month and a half ago, we just announced that for the Samsung home screen. And the home screen for many TV OEMs is the largest ad unit that they have and the most valuable ad unit that they serve. We just won the entire home screen ad serving business and primarily SSP serving business for Samsung's hundreds of millions of TVs, and we have had that previously with others. I think even a Publica market, we have been able to take share away from somebody that was programmatic first, but yet not within a company strictly focused on the CTV market.
Okay. I want to segue into that. Can you talk a little bit about your position with Magnite, with SpringServe in the agency marketplaces and in these new platforms like Walmart and Vizio?
Yeah. Just so the audience understands this, we took SpringServe, which used to be just ad serving, collapsed that with our two CTV SSP exchange-like platforms, and made that all into one product. So you don't have separate user interfaces or different logins and different reporting and different analytics and different features. It's all one product now. Very sticky, and the crossover of customers is almost 100%, meaning when we first bought SpringServe, you may have had 25%-30% people using one product than the other. We went to 50, we went to 75. Now it's interchangeable. I don't think there's a single one that doesn't use one from the other, that it logs into an overall SpringServe platform. That is critically important in the market. Remind me the second part of your question, Barton.
Yeah. Just explain your position with the agency marketplaces and with Walmart and Vizio.
Yeah. In agency marketplaces, agencies are attempting to claw back some of the power that they ceded to DSPs in the open internet in the CTV marketplace. The reason is they still play a critical role with a number of publishers, meaning a limited number of broadcasters and streamers. They can still influence buying at scale across their marketers and brands. They can still exert overall volume in terms of what they negotiate from a spend perspective. But they don't have any tech to be able to access and deploy and reach those audiences and be able to connect into pipe to inventory that publishers offer. So those agency marketplaces are all built and powered by Magnite, and now they're hitting real scale. They can bring data into those marketplaces. They provide the match on the other side.
It gives transparency to brands for what they would like to buy. But they're routing that demand and that initial buying journey through themselves, as opposed to having them go to a DSP interface to be able to go to market. That transparency is they can pick and choose what they want to buy or what they don't want to buy. They can bring in their own data. They can have the match happen at the publisher, which is, by the way, where all the matches are happening in CTV because they don't want to share their identities outside of that. So there's ad serving fees to that.
There's execution fees to that, and effectively, at the end of the day, they choose their payment method, and they could use a DSP as their payment method, The Trade Desk or a Google or an Amazon, or they can use ClearLine, where they can bypass a DSP. The DSP's role that they play in that equation is much diminished. We're doing the execution, we're doing the audience, we're doing the matching. So all that execution work is ours, and the agency is doing the selling and the marketing of that inventory and negotiating the purchase. So you really have relegated a DSP in that equation to more of a payment method as opposed to the value provider or the on-ramp to be able to access, determine your audience, and select your inventory.
Okay. Now, I wanted you to talk a little bit about Your way to—
Well, you mentioned Walmart as well, and Vizio
Yeah. Go ahead and talk about that, please.
A little history lesson. Our relationship with Walmart started with working with Vizio prior to the Walmart acquisition. We were the ad server for Vizio. We were their preferred SSP selling inventory for them. We had a very, very deep and strong relationship. Walmart bought them, and then as Walmart has wanted to expand that to Walmart's branded TVs for Vizio's operating system, obviously our footprint grows commensurate with what Walmart does with that operating system and how they run their ad stack for Walmart on TVs. What's recently happened, call it a year ago, is Walmart used to use, as their DSP, they used to use a white label version of The Trade Desk. That was effectively their internal DSP, and in doing so, they only allowed their Walmart shopper data to be parked within The Trade Desk's DSP.
That was the only DSP that you could buy it from and have the benefit of Walmart data. Walmart, about a year ago, said, "We're going to break that, and we're going to make our data available outside of just one DSP. We don't want to force people's path to have to go through just one DSP." If you want to choose your own DSP, and there's been an emergence of other competitors and other peers in the marketplace, they want to be able to allow them to access their data on Vizio inventory or Walmart TV inventory, but do so through any DSP so that they could gain a benefit, which the benefit to them is a lift in CPMs. If their data's available on The Trade Desk, they may get a higher CPM, but now you can buy it through any DSP.
Walmart benefits from higher CPMs on anybody's inventory, so that's advantageous. Walmart took this a step further with what's recently transpired just before Cannes and then at Cannes. What Walmart has done since that point is said, "Look, now we would like to be able to do more." They bought a DSP. They're now going to use their own DSP that they're acquiring called Vibe.co. We're tightly integrated with that. They're a small and medium business focus, CTV primary DSP. They're going to move off of a third-party, white label version of The Trade Desk and move on to Vibe.co.
They're still integrated with us, but their desire is to now add other DSPs, so they can be the on-ramp, or they can work with other DSPs, and their data can be available beyond just the scope of Walmart inventory and their O&O, but be able to be a DSP that can buy across lots of different publisher inventory types and expand their value of their data beyond just their inventory. A very, very big opportunity. Again, this is announced, but nothing's transacted. They haven't even closed on Vibe.co, to my understanding, since they announced it less than 30 days ago. We have a lot of this, or a little over 30 days ago. We've got a lot of this growth with Walmart Connect coming in the future, as opposed to running through numbers at the moment.
Okay. I wanted to talk for a minute about agentic. You guys were talking a lot at Cannes about Magnite Orchestration, and I believe one of the opportunities there is increasing the addressable market, right? You are taking some of the inventory that might have transpired through a direct kind of manual process between a marketer and streaming television service provider, and putting that through the programmatic pipes, giving you a take rate opportunity, and also improving efficiency for the partners to the transaction. Can you explain a little bit about what you are doing there and what you think about that kind of characterization?
Yeah. Just to clarify slightly, you got most of that right. The little nuance between that is today we have a lot of spend going through programmatically, right? CTV primarily is where we are initially focused in our agentic work, as well as in DV+. If all we are doing is converting inventory that is being run programmatically to having that be done agentically, and we do not charge a different or an incremental fee for going to agentic versus going through programmatic, there is no ad spend difference to us whatsoever. It is the same revenue, just going through a different path. Arguably, there is a little bit less cost to doing it agentically than there is doing it programmatically with people setting up campaigns and refining and optimizing. But there is really not new benefit to TAM or growth or market share from that happening.
You could make the leap that says, maybe you bring in revenue faster, so there is a lift there. Maybe you take share from others. But there is a lot of if-thens tied to, is yours better than somebody else's? Do you gain share? Do you lose share? But if all you are doing is replacing that, there is not a whole lot of immediate upside. What I would say is the area specifically that you have mentioned that we are targeting is a lot of spend that is coming into agencies today is still traditional insertion order, to the tune of tens of billions of dollars still being spent in that traditional way. As that inventory, or as that buying demand comes in, what happens is there is a program goal that the marketer has. They give the agency those goals.
The agency then comes up with a media buying plan for what they would suggest should be bought to successfully run and achieve those campaign goals, but that is not connected to real inventory. Even the agency might be able to make that conversion as their own buyer agent, right? So that is what a buyer agent can do because that is only their workflow. We, however, give them access to real inventory, to real-time inventory, so they can actually run the creative, which our creative exists in that workflow as well.
They can actually run sample campaigns, analyze those results, refine what those are, rerun those tests, each of which generally takes one to two weeks at the agency level to come back with what those test results are, refine them, and they get the buy-in from the client that this is what should actually run as a full campaign. What today takes anywhere from four to eight weeks, we compress into 10 minutes because we're connected to real-time inventory and supply. What I would tell you is orchestration is we're agnostic whether it's our buyer agent, whether it's our seller agent. Our mediation agent is something that obviously we have yield management and core competencies and audience creation that we think we're better than anybody else.
Our orchestration basically means you can either use all of our agents together, and those three agents together create an optimal overall agentic workflow from everything from start to finish, or you can plug your own in and we'll optimize yours. Any one of our agencies, bring your own. If yours is better than ours or yours solves this problem, we'll connect it to real-time demand, we'll connect it to publisher inventory, and in real time, we'll be able to optimize all of those together, and we'll orchestrate that all of that works seamlessly together. That's the advantage that you work with Magnite.
Because we sit in between, because we're the largest scale player, because we see the most CTV inventory, and because we see the workflows on the buyer side and the publisher side and everything that happens with data in the middle, we are ideally suited to be the one that solve all three areas of agentic workflows, as opposed to sitting siloed on one side or the other, hoping that we can connect and hoping that there's a match somewhere in the middle. I think where we've sat is what programmatic truly is a massive shop of APIs that we are constantly, and have historically constantly worked on trying to refine and optimize and improve, both for a publisher as well as for a buyer.
The fact that we can now solve all of that agentically is really replicating and using AI to do what we've done programmatically and really improve velocity and reduce friction in the marketplace.
Yeah. But one of the takes is, those are transactions that would not have been subject to a take rate, but now could be, but still save money for the parties to the transaction.
100%. That new agentic IO business is all new TAM that now we are serving, and you are serving it with traditional publisher take rate. There is not a different fee for how those dollars come into the ecosystem or how they transact versus our typical rate card.
That could be roughly a 20%-25% kind of lift to the addressable market, if you think about it.
Absolutely.
Yeah. In the couple of minutes left, I know that we are at very kind of, nice kind of revenue trajectory, contribution ex-TAC trajectory for you guys. How should we think about margins? As you are growing revenue at this pace, what happens to margins?
So kind of where we started this year, we are seeing play out what we have all along said the operating model is capable of. We started this year with more modest growth expectations of call it greater than 11%. That would lead to EBITDA growth in the mid-teens and free cash flow growth above 20%. So now that growth has inflected, you have seen margin expand pretty dramatically, where flow-through to EBITDA from incremental revenue dollars is about 80%. Q2 showed exactly that. Our beat on the top line was $10 million. Beat on EBITDA, the bottom line was $8 million, so 80% flow-through absolutely played out the way we said. So there is real good evidence that the model works the way that it should, not just promised for a scenario that never happens. So we feel really good about where margins are.
So we are squarely now at a guide from we started the year at below 35% to now over 37%. You are seeing it flow through. There is no cap at where we go. Obviously, faster growth continues to fuel that margin expansion. So we have said 35%-40%, but that is not meant to put a cap on margins at 40%, especially with that flow-through of 80% of incremental margins coming through. You have got quite a bit of running room with this elevated rate of revenue growth if it is sustainable.
Okay. One final point. The company is basically de-levered, as I said at the outset here. So what do you want to do now with your free cash flow? And it is a meaningful stream of money right now. It could become a torrent if the antitrust thing swings your way, as we are arguing for. So how would you want to use your cash?
Yeah, honestly, I wish there was more quality CTV assets and different things for us to be able to add. Really, there is not much more than products and features that could accelerate a roadmap, things like a streamr.ai that we did most recently, but those are relatively small, very low double-digit million type deals. So, in the interim, until we kind of tease out whether or not there is a few of those or none of those out there, we have committed to at least 50% of our free cash flow will go towards share buybacks. We were more aggressive, and so we have a $200 million authorization over two years. We did get more aggressive with our balance sheet in Q1.
We did or sorry, in Q2, we did actually tap the revolver, and we did that in order to not wait to deploy cash flow that would come in later in the year when the stock was at lower levels, in the 13%-ish or so range. We'll continue to execute against that return of capital to shareholders through buybacks and continue to look for other ways to be able to drive high ROI relative to our investment profile. But that currently at the moment is our number one choice. But we'll keep looking at effective ways to deploy that capital.
Okay. All right. That's great. I guess, in the last kind of probably 20 seconds here, I know that your CFO is retiring. Just what can the company say about where you are in the process of the replacement search?
Yeah, no, I would say we're making good progress. It's probably a problem of good quality candidates to consider. Myself included, I've thrown a hat in the ring, so maybe there's a shot there. But I think we have good internal candidates and good external candidates we're considering, and I think that's probably made the process a little tougher to get through as we go through it. Then you also have kind of schedules typically in summer where either interviewers or interviewees may not necessarily be in the same place at the same time. So I think we feel really good about it. David's timeline is to stay on through the end of September, so I wouldn't foresee any reasons for extending this any further past the indicated original timeline.
Okay, great. Well, I guess that's basically it. Nick, thank you very, very much. Thank you, everyone, for joining us. I think that's a wrap.
Much appreciated, Barton, and everybody else that took time in August to spend time with us. Thank you.
Okay. Thank you.