Morning, I'm Laura Martin. I'm the senior media analyst, internet analyst at Needham & Company. I'd like to introduce our speaker right now. Michael Barrett is the CEO of Magnite, where he led the 2020 merger of Rubicon Project and Telaria and the subsequent acquisition of SpotX to create the world's largest independent sell-side advertising platform. A prominent leader in the ad tech industry, he previously served as CEO of Millennial Media and AdMeld and held senior executive roles at major technology and media firms including Yahoo, Google, and Fox Interactive Media. Welcome, Michael. Nice to see you.
Thank you, Laura. Pleasure being had.
Pleasure being here. I wanna start with a leadership question. I would say most people agree that generative AI is transformative, which creates uncertainty about the future and roles and job loss, so how does that change how you lead an organization in this kind of period of technological uncertainty?
Yeah, I do concur that it is a sea change. I think if you look at the history of Magnite and you go back to the Rubicon Project days.
Yeah
It hasn't actually been peacetime for quite some time.
Wartime CEOs the whole time.
Yeah. I think if you're in ad tech, you're a wartime CEO, period. Yeah, if you look at Rubicon, our team took over Rubicon when it was trading for less in cash. We were able to right that business, get it back to growth profile, and then that allowed us to be able to increase our exposure to streaming, which we thought was going to be a big sea change in terms of importance for the industry.
Yeah.
We were able to accomplish that through acquisitions and implementing the acquisitions that we purchased and turning the company into a cash flow machine and a growth engine. I think we've been through this before, so I think we have a track record of success as it relates to transformational opportunities, and I think AI is just another one of those. I think it's big. It's important. I think we're doing all the right things from a investment standpoint. We have talked about the launch of our agents and agencies like Publicis, MiQ using it to purchase from Disney. We are living it right now. Our crystal ball, I think, is as good as anyone's as to where it might be heading.
We aren't sitting back on our laurels thinking we got this all nailed, but I think we're doing the appropriate investing in it, testing, and we have the talent and the organizational structure to thrive in this coming agentic era.
Do you feel there's any demands on education or hand-holding or the human capital aspect of the organization with the uncertainty?
I mean, it'd be foolish to say that we roll out of bed and we're experts at it instantly. I think the learning curve that we've seen over the last two years has been just incredible, and I think if you look at the profile of the executives that we have at the company, you know, they're all in that kind of agentic profile in terms of they're agentic first. They've used these tools personally. You look at our engineering, you know, David Buonasera , our CTO, comes from SpringServe. That in and of itself was innovative in terms of its use of the cloud, they come with very diverse backgrounds that I think are very open aperture. They're willing to test and try anything.
I will honestly say we were quite frustrated for the early years of using the LLMs in terms of the output. Like, we were just like, "Why aren't we seeing what we're reading everywhere else, that fire all your engineers, it's right in the code itself?" I think what we realized that they weren't all that great early on. We were using them, but we were debugging half of the stuff that was being written in, so we were not seeing that. All of a sudden, with the latest models that have come out, December, January, February timeframe, it's just been a hockey stick in terms of productivity because they've gotten better, and they're only getting better. It is not from lack of use that we've been doing it. Maybe lack of best practice.
It takes a while, but I think we're starting to fire on all cylinders, and we feel really, really, really good about where we stand.
Okay, very interesting. Okay. You guys at POSSIBLE, I didn't see you 'cause I heard you were in Fiji for the day at a sales conference or something. I met with Sean-
No, it was a client conference.
Oh, it was a client conference.
We took our Australian and New Zealand broadcasters to Fiji. Someone had to jump on the grenade.
Right. For 20 hours each way.
It literally-
For a 20-hour meeting.
It literally was a 12-hour. It was lovely, the 12 hours.
It seems like you should have come to Miami and sent somebody else to Fiji for 20 hours each way.
Yeah. You try to get David Day to do anything that he doesn't wanna do.
David Day.
That's why he's retiring.
David Day. He's retiring 'cause you wanted to send him to Fiji.
That's exactly right.
Now it comes out-
Yep
For the day.
Exactly. Yeah.
I met with Sean. You guys announced a lot of new products.
We did.
At POSSIBLE. Let's sort of walk through those, and I'd love to have you rank which of them. We have agentic AI, you have your streaming AI integration, you have retail media upgrades, you have expansion in sports, live sports ad tech. Like, of these things you announced in the last literally six weeks, what is it you would rank as having the most impact that investors should care about the most to sort of least consequential?
Well, you know, I think we curate those announcements pretty well, so they're all important. You know, if we were to announce everything we did, you'd be getting two announcements a day. Those are the most meaningful. They all had a flavor of strategy to it. Some of those are more revenue instant, right? If you look at an announcement like a streamer, it brings with it revenue instantly. You know, you sign up a client, they start to use the product, and they start to buy. Where agentic is a little bit It's heavily strategic, very light on the revenue, right?
Yeah.
Test, test. Small amounts go through, but you're ready for when the big amounts come. Incredibly strategic there. I would say the one that straddles really strategic and revenue ready.
Yeah
Is Commerce Media.
Okay.
I think it's incredibly important to remember where Commerce Media was. We use that synonymously with Retail Media, so if that is more understanding for everyone, call it Retail Media, even though it's a broader category. I wouldn't consider United Airlines to be Retail Media.
Right.
That said, it's commerce. If you look at two years ago, the playbook in the early days of Retail Media/Commerce Media was go to a DSP partner, take your valuable data, put it in that DSP, and then tell all your advertisers that want access to that data to use that DSP. That was the first chapter. What you're looking at, you're seeing Commerce Media being kept closer to the commerce partner. If you're United Airlines, if you're Pinterest, if you're, you know, PayPal, you are saying, "Why am I taking my valuable data, putting it in a DSP and telling everyone to use that DSP when they wanna use other DSPs, other buying agents? I'll keep it close.
I'll work with an SSP to keep it safe. The valuable audience segments are gonna be created on the sell-side, not on the buy-side. That, to me, is a huge strategic shift because if you keep in mind that the valuations of some of the DSPs have been much higher than SSPs because people will say, "Well, they're more strategic. They do all the hard work." Now we're doing all the hard work, and so I think that's a big shift. It comes revenue ready. The buying kicks in right away. You democratize the ability to access that valuable data by using any DSP that you want or our ClearLine. And you, then you buy the data that sits within Magnite. It, it's very important.
My understanding of this is you don't charge for the data. You don't charge an up fee. It's just that you get higher CPMs when you attach data to these audience segments that you're selling, and therefore your take rate becomes more money because your CPM is higher. Do I have that right or no?
I think you have that exactly right for the instance that we're talking about, and that is incredibly valuable first-party data. You know, Walmart has great first-party data.
Yeah.
Disney has great first-party data.
Yeah.
It's hard for me to resell their data because they don't really want me in that. They want their data to be attached to the media that you wanna attach it to, and they are the ones that price that. We participate in the take rate on that.
Right.
The whole area of curation, and there's a whole host of other data that's out there that is not 1P in nature that we definitely participate in the economics of reselling of, data.
I just wanna make sure I understand this because Rajeev was just on stage. He's your direct competitor, he says he up charges for data. There he is. He up charges for data, which was the first time I'd heard that from an SSP. In fairness, I only talk to you as an SSP. That's not true, 'cause I talk to Index Exchange, I talk to private SSPs.
Oh, you gotta stop talking to them.
Yeah. I love that guy.
He's great.
I'm sorry.
Andy's a good guy.
Like, he's in competition with you, my favorite guy.
Andy's-
I love that guy.
Yeah.
Anyway.
Someone being a 10th the size of me in competition, he goes.
You're insulted. Okay. It's okay. I had to insult you over something. My friends, I'm happy to insult you over my friendships. Okay. Anyway, third-party DSPs, or sorry, the data that you're attaching. So curation was a big subject.
Yeah
With Rajeev. Curation, which I think is a hard idea for most people that aren't sort of immersed in ad tech, you're up charging for the third party, but not for the first party. That makes sense to me. You're benefiting from a higher, a same take rate, but on a higher CPM.
That's right.
That's how.
Yeah
For the first-party data.
Yeah.
That makes sense. Really, this retail media, this commerce media is really just an extension of your curation, right? It's not a new product, is it?
Well, it's a new, We're participating in it in a new way, right? Again, going back to the example, it was the belief that this is better done by DSPs.
Isn't curation for the last three years that the whole time?
It was not, keep in mind, some of the bigger commerce players were still doing the DSP.
I see.
Route.
Okay.
You're seeing a sea change.
This is like a segment.
Yes.
This is a new segment.
Yeah
That's really important.
Which carries with it in incredibly powerful dollars, you know?
Yeah.
These are hardworking media dollars that are generally more performic.
Yeah
At least more measurable.
Yeah.
You're seeing it going off network, right? You know, the initial part-
Going off network, what it means.
Well, if, let's use Amazon.
Okay.
Amazon, traditionally, you ran, if you wanted to use Amazon data.
Yeah
You would run it on amazon.com, and that would be the way the media worked in the, in the retail media world. Little by little, their advertisers, if you're Gillette.
Yeah
Hey, you do a lot of TV ads.
Yeah.
Why can't I have TV?" It's like, well, Amazon doesn't have TV. They have Prime Video.
Yeah
I usually run on Disney, I run on Netflix. Can I take that data and run it on a supply that isn't Amazon?
Yeah.
That was the early stages where we participated in commerce because it was coming off network.
Yes.
Amazon's owned and—
Yes
Operated—
I see.
Onto third parties.
Yeah
We work with.
Okay.
Now you're seeing that the whole audience segmentation and the audience value is shifting to being created on the sell side.
Right.
We're participating in it from the get-go.
Gotcha, gotcha. Okay. Which should bring more dollars.
Yeah.
Right?
No question.
Because it's not residual dollars that Amazon didn't spend.
No, no.
On its own platform.
It's wonderful for CTV publishers to be able to get these dollars that weren't, that were being trapped over here.
Yeah
And just—
Perfect.
Okay.
your clients benefit—
Yeah
You benefit.
No question.
A key theme, this stays on performance really. A key theme in advertising today is tying CTV ads to sales lift or incrementality. Yesterday we had a future of ad tech panel. It was all about incremental transparency. Incrementality is like sort of the future of ad tech. Let's talk about your recently expanded Roku deal, where Magnite will integrate Roku's unique first-party data with Best Buy, PayPal, and several other companies to improve your CPMs. Let's talk about Roku.
Yeah. Another example of a partner that is just becoming more and more programmatic first. You know, we are deeply embedded in Roku, run their exchange, and now they're in the curation business, your favorite word.
Favorite word?
this is—
Means nothing to anybody except me and you.
Essentially.
It's total inside baseball.
But essentially-
I like the word, which my friend uses, sell-side decisioning, much more descriptive word than curation, which means nothing to anybody except us. Look at him. He just hates it.
You and Andy, you just have to stop with this bromance. Like
We, bromance. Total bromance. Oh, my God. Sell-side decisioning is a better term for this.
Yes.
Go ahead.
Roku.
Roku, again, it's an overlay of curation, Commerce Media, and sell-side decisioning.
Okay.
Ultimately, what Roku wanna do was bring valuable Commerce Media Retail data.
Yep
Decorate their inventory with it.
Yeah. Okay.
It's probably you could do that yourself if it was just one company.
Yeah
They start with three, and it'll be many, many more.
Okay.
That's when they turn to someone like a Magnite.
Okay
Who has that capability to help them. Do we have a relationship with the commerce media people? We have the tools to be able to do it for them.
Yeah.
We have the relationship with Roku, so—
Okay
It's a perfect match there.
Okay. Is that because your tech stack?
Yeah, because it's, we do it for a living.
Yeah. Okay. Do you get the sense people are specializing? Like the tech aspect in a generative AI world, 'cause I sort of feel like agentic lets people not specialize, 'cause agents do so much of the work. From what you're saying with Roku, it's like they're letting you specialize 'cause you do this for a living.
It's just, I think what you're seeing is, in the early days of being an SSP, it was kind of one size fits all.
Yes.
Especially as we've, you know, eclipsed the more than 50% of revenue in streaming.
Yeah
That is much more of a world of, I'll take one of those, two of those, and one of those. You just kind of bundle whatever you want together.
Okay
That was a point that we tried to make in our earnings, was about how prevalent SpringServe is in everything we do. It's just not a programmatic first ad server used by most of the programmatic first folks. It is this technology is interwoven into all pieces of our stack when it comes to streaming. It makes it incredibly valuable for our proposition. SpringServe and ClearLine would be a part of curation piece of it.
You made a lot of brilliant acquisitions. SpringServe, hands down, was the best one you made.
We joke with the SpotX people, who we love. We paid $1 billion for SpotX, and worth every penny. With that, we got the call option for SpringServe at $30 million, and people now look at the deal and say.
It was worth $1 billion.
You got SpringServe for only $1 billion?
Yeah.
No one likes that story at SpotX, and no one likes that story at SpringServe.
By the way, when Michael Eisner bought ABC, he thought he was buying broadcasting.
Yes
He had a little asset called ESPN.
Bingo.
And it was worth—
That's exactly right.
The entire purchase price
There you go.
ABC. Now they're trying to get Jimmy Fallon, you know, fired. Anyway, Jimmy Kimmel anyways. Okay. Anyway, so that's the Roku deal. Where is AI already driving measurable ROI today? By the way, you're hiring a lot of people. I'm keeping track of that and not loving it. Tell me where you're getting ROI from generative AI, both on the internal and external front.
Yeah, we look at it in three buckets, right? The first bucket is operationally, how can AI make us—
Yeah
Run faster, more agile, more cost efficient? We've made really positive inroads there. Early days, one of the biggest areas we looked at was we have a lot of contractors because ad ops is a kinda crappy, dirty, messy business.
Yeah.
We, you know, like every company, we carry tech debt, and we're just like, you know, okay, for every resource that we could apply internally to fix a problem from a workflow standpoint, those same resources can be applied to fixing a client, and we can make more money. You always, as a tech company, kinda say, "We'll hire some contractors to get around that.
Yeah.
We've been able to release almost all of our contractors.
By building agentic tools to make this stuff easier, smarter, faster.
Okay.
Arguably, we should've done it a while ago, but it's so much easier now and faster to do agentically. Operationally speaking, we're looking at hundreds of things that can be done better if we use AI.
AI. Okay.
It's resulting in some of the headcount that was projected to be hired this year and the offshore resources to be eliminated. The other way we look at it is, you know, agentic development from our engineers. As I said before, we started early on using the tools that everyone uses, and we're not blown away by the results that we were seeing. Lately, literally talking the last quarter and a half.
Yeah.
It's just a hockey stick in terms of how—
Okay
Sophisticated they've become and how maybe sophisticated we've become using them.
Yeah.
It's really I'll give you an example. A not insignificant cost we incur every day is load balancing in the cloud, and AWS charges for that service. It's a pretty penny on a daily basis. The idea of building our own load balancer seemed like just a bridge too far. Using the agentic tools, we built it in 2.5 months, and it's in production and gonna save us thousands of dollars a day. Exactly. We can bring it, put it on-prem, and do it ourselves.
Yeah
These are real examples that are occurring every day. Like, there's almost astonishment. Like, "Oh, my God, that guy disappeared for a weekend, came back, and boom.
We have a load balancer
Look at this." Yeah, there are just gonna be more and more and more. We just scratched the surface there.
One of the CEOs I talked to at POSSIBLE Private, he was saying that he pays people bounties for things like this. Like, if a guy goes and does something that, you know, either he pitches it or a engineer walks in and says, "I made this. I think it'd make the business better," and they test it out and they use it, he'll give them $50 grand or $100 grand. Do you guys think in terms of that if somebody brings something really high value or just part of the job?
Yeah, I would think generally speaking we say it's part of the job, but we do have incentive programs for the engineering group, and it's pegged on certain milestones. That could definitely feed into the incentive program for engineers.
I absolutely feel like there's some capabilities here that could be really astronomically valuable, but people are like, "Why do I do this? What's in it for me?" type of thing.
Yeah, very selfish company.
Maybe you don't have.
Yeah. Our folks are selfless.
Selfless? Yeah.
Yeah.
They just do it for you?
Yes, exactly.
Yeah, leader?
Yeah.
Okay, got it.
You find that with those Canadian ad tech companies.
I am not talking about my bromance here. We have left the bromance silo.
We have moved on.
I have other friends.
We have moved on.
That you might not like that I'm not going to name.
Do you consider Rajeev a friend?
Yes, I do.
Okay.
Yes, I do.
I do, too. I do, so.
I didn't see him at POSSIBLE. Okay, one of my key takeaways from—
Lastly.
Yeah
Obviously the agent stuff that we've talked about.
Okay
Are going to want to use agentic-first tools, both publishers and buyers. It's our job to be able to make sure they're using ours. If they're using theirs, make sure it talks to ours in a way that is seamless and productive.
Let me ask you something. If two years from now, which feels like forever, but two years from now, how much do you think agents will have infiltrated the big ad agency holding companies?
100%.
Really?
Yeah. Now, let's caveat that by saying, what do those agents do? The idea of having one dashboard instead of 10 dashboards.
Okay
Be able to talk to that dashboard, that agent in a natural language capacity.
That Buys, Google and—
Yeah, exactly.
Meta and—
Yeah
Open Internet
they still may have to log in to, through a, you know, MCP, but they'll have to talk to those other dashboards, but the person that is the end user, the trader, will be freed up because they don't any longer have to go into PubMatic, go into, you know, Magnite, go into Facebook.
Meta
Yeah, exactly.
YouTube
Yeah. Now are those agents sophisticated enough to do real-time bidding, just talking to seller agents? Absolutely not. They're going to need the rails and the infrastructure that exists today to be able to execute programmatic buys. It will absolutely make their lives easier. How much of it displaces, how much gets lost in it, I think that's the battle between now and the next 24 months in terms of figuring that one out.
You think it's sorted out in 24 months?
I think if they're, if it's 100% use and they're pervasive, yeah, you better have figured it out or else you're gonna get cut out of it.
One of the things that another public company CEO is saying is that he's saying that there's rails and not Rajeev, who's in the room. He's saying there's rails for programmatic, and the agentic rails are gonna be different rails.
I wouldn't bet on that one.
Okay.
It I would say that we better be dabbling in that, and we better be prepared.
Separate rails.
That said, I think that the hundreds of millions of dollars that our companies have invested into a trusted transaction layer that works.
Milliseconds
Yeah, it works in under 100 milliseconds, and it's if you bought Disney inventory, you feel comfortable. It's almost like NASDAQ. You, if you buy a stock on NASDAQ, it's all electronic. It's all done. You feel comfortable that you actually did buy a share of Disney.
Right, it did it in a transaction.
You bought it at a right price, and you're going to get the share.
Yeah.
The person that sold the share is gonna get settlement from you.
Yeah.
That definitely To think that you're gonna blow all that up in 24 months and relearn a whole new system I think is a little.
Without the trust factor.
Fanciful. Yes.
Okay.
Exactly.
Yeah.
Will there be fewer? Will there be fewer folks that are doing that?
Fewer what?
Fewer exchanges of record? Will more volume be on the ones that survive?
Okay
Completely believe that there will be a continued shrinking of players. I think the players that exist will be bigger, stronger, and more important.
I wanna push on that because we're seeing a real resurgence of DSPs, that the barriers to entry for DSPs with agentic has really been lowered. We're seeing sort of this flowering of new DSPs, which doesn't feel like consolidation to me.
Oh, heck, you can build anything agentically, right? Try to steal clients away from Amazon and Trade Desk and Viant, they have commercial relationships. It's like, you know, there'll be facile buyer agents that marketers wanna work with.
Okay.
The question is, does that displace a DSP completely? Is that a dashboard that still works with the innards of a DSP? I do think that right now, the phase that we're in in agentic, the buy side is under a lot more pressure, because— People can more easily envision in a one-to-one world where you're just doing a buyer agent to one publisher. Th at can seemingly be quite simple.
Yeah.
You're almost, like, taking direct sold or programmatic guaranteed and j ust making it more easily transacted. We haven't seen the world yet where one buyer agent is trying to talk to 10,000 seller agents in real time, bidding and trying to, you know, win. I find that to think that that blows up the whole existing rails and infrastructures and protocols that the industry's been working on for 15 years, I think is a bit fanciful.
Even five years.
Yeah.
You don't see that. Okay. Very interesting. Okay, cool. One of the things that came out of POSSIBLE for me as a takeaway, as an investment analyst, was that there's been this separation between value capture and value creation because it's so easy and cheap to create sort of generative AI layers, even content creation. But pe ople aren't charging it. They're just throwing it into their offering to try to sort of create competitive advantage for their products. Do you see this happening, that these are gonna become like table stakes, these AI integrations and upgrades you're making?
I think it's a lot like any product improvement or innovation that you do, and it is table stakes to a certain degree.
Okay
That you have to keep up with what the clients want. You have to be a step ahead of what they need, and you have to be constantly innovating and differentiating yourself in the market that way. Some will clearly be able to be charged for because it's a brand-new, greenfield opportunity. The stuff that's just table stakes, it's part of being able to keep your existing business model in place.
Mm-hmm. Okay. All right. You know I hate that answer as Wall Street, but that is the answer. That's the answer. By the way, a lot of people are giving away AI stuff 'cause it's just not that expensive to do.
Yeah, it's becoming, if you wanna continue as a entity.
Yeah
You better be able to meet the market needs.
Be able to do it.
Yeah.
Yeah. No, that's fair. Okay. In the midst of supply path optimization, that's what we're in the middle of, the open internet ad tech industry fully optimizes supply path. Doesn't that hurt your take rates, too? Like, we're in this supply path optimization, so isn't there going to be pressure on take rates in the open internet, which hurts everybody that's a player?
Well, I think open internet, sometimes streaming gets lumped into that.
I think that's true.
Yeah, let's separate the two 'cause I think.
Okay
They're both quite different. open internet, web-based, browser-based, I think you've already seen a supply path at its finest, right? There are a handful of serious players that are left. There was a time when honestly there were, could've been over 200 SSPs.
Yeah.
Now you're really down to, you know, five or six that matter, and then some, like us, matter so much more than a PubMatic, it's just ridiculous.
Stop 'Cause he's sitting in the front row.
I did that on purpose.
He's doing that on purpose.
I did that 100% on purpose.
You're so mean. You're so mean.
We've already seen that. Where will I take a hit on take rates for, in exchange for volume? When we talk about these agency marketplaces that we do, if someone wants to shift a couple of hundred million in spend our way, of course we'll look at doing preferable treatments on take rates.
Okay.
It's been amazingly stable for a very long period of time, you know, you'll do the math on take rates from, you know, the amount of revenue we produced last year and its total spend and you'll see that it's been incredibly flat. I don't think we're looking at a complete collapse in terms of the economics of the business because I think the biggest supply path has really already occurred. I think in streaming it's entirely different. We've always talked about that as a winner-take-most business. There's never really the need to operate with the number of SSPs that you did with in the open web, and I think that more and more Magnite is so well-positioned as a technology partner.
Yeah
That you're also talking about technology that would have to be ripped out. We feel very good, and again, it's over 50% of our revenue now. We feel very good about the position from an SPO standpoint in streaming.
I think let's slow that down 'cause it's something we've been writing about a lot, but not everybody may know, not know as much, which is this shift from sort of, I would call it vendor of SSPs to these sort of integrated tech stack developers for Netflix and Disney. Why don't you slow that down and give a little bit of the storytelling, and also how the business model shifts and, you know, what you think of long-term, lifetime value for that kind of relationship versus the traditional, SSP relationships you've had with these publishers.
Yeah. It's a really important point. I think that, early on in the Telaria days.
Yeah
Hulu was looking to go into programmatic and was looking for a partner. They chose Telaria, a company that we wind up merging with. This is, you know, they have 10 years of experience at this, and the initial spec that Hulu wanted had nothing to do with bringing third-party demand from DSPs. It was being able to build out programmatic capabilities on the Hulu ad server. It was more of a statement of work. We literally built technology for Disney that, you know, over a period of time, like six months later, we could repurpose and introduce it to other people in the ecosystem.
We've been at this for quite some time. Netflix is a good example of someone who wanted to really lean into programmatic, but only saw the need for a technology partner to build out programmatic infrastructure for them. That's what we do there. You know, you look at Roku, they used to be first party sold, you had to use their DSP, then they opened it up and became a almost, you know, bring one, bring all DSP.
Yeah.
That precipitated the need for working with an SSP that could do that for them. Also, you know, powering their exchange. I think there's flavor. I think the importance is everyone's slightly different. The good news is it doesn't come at, "Oh, gosh, now we gotta drop everything and rebuild something for this guy." We've been at this for so long that the suite of products that we have, and their interoperability and their ability to be compartmentalized so that I'll just take that one and that one has really set us apart in that area. I think that, you know, more and more of the new product, the new client announcements that we've made have really had a backbone of that to it.
Very little of it in that space is, "Hey, I just need you to bring me undifferentiated DSP demand." It's much more of, "Hey, I'm building my own ad server. I don't want anyone to touch it, but I don't wanna do programmatic. Can you do programmatic for me, but I wanna call it my own?" That's fine. It's a very different business.
How does the business model change? Super easy on the SSP side, take rate. You know.
Yeah
We take 8% or 12% of ad spending on our platform. Tell me how this business model is different when you're building the ad tech stack.
You have statement of works that have become few and far between because we've built so much. You have seat fees so that folks who want a certain level of service from you and a certain level of demand, they pay a fixed monthly, and you have take rates, which is a percentage of media spend. What we're seeing, as we've talked about, you know, the publisher sold programmatic, which used to be prevalent, and it was the only way that broadcasters would do it. Starting to morph much more so into Magnite, sold programmatic which carries with it a much higher take rate. You're starting to see that in our growth figures, where, you know, 30% growth in CTV in Q1, some of that is aided by product shift in terms of take rate.
Mm-hmm. Let's move to live sports on that, because live sports is really helping drive that.
Yes.
Talk about what's happening with live sports, h ow much more is coming into the programmatic ecosystem, and where you see that going.
Yeah, it's super exciting. I mean, the first stage is live sports has to be streamed, right? We've done that. I mean, how annoying has it been trying to follow NBA basketball games in there? We gotta fix that's more on the content owner side. Phase I, live sports comes to streaming.
You know, check.
Phase II, you have to stop just doing your linear feed on streaming. If all you do is linear feed on streaming, there's no dynamic ad insertion.
Right.
There's no program.
It's just the ads from the linear.
Exactly. That's going by the wayside because folks are realizing they're accruing a younger audience in streaming.
Yep.
A more valuable audience in streaming.
Yeah.
Why not take advantage of it with a different set of ads?
Have a new ad unit and selling a new ad unit.
Exactly. Now you gotta get over the hurdle of, boy, DSPs really aren't built for live sports. If you gave a normal DSP a $1 million campaign to spend over a month, they would spend it during the middle of the day to later evening, one-thirtieth, one-thirtieth, one-thirtieth, and then end the campaign at the end of the month. Live sports comes along, it's on a Saturday afternoon, and it's a huge spike in traffic, and most DSPs are like, "What is that?
Fraud.
I don't wanna touch it. It's fraud. We've done a ton of work in terms of helping educate, building LSA, Live Stream Acceleration. All this, we know, because of the ad server, we know more about the programming than the DSP, so we help educate the DSP.
It's live sports. That's why you're seeing it
It's all safe. You wanna spend it. It's super exciting. We said that March Madness for us was up over 80% year-over-year. You're going into FIFA.
Yeah.
They'll be shown a lot on Tubi. That's all programmatic. Now you're going into the fall where all these sports that have now reached their third year of being on streaming, they're all dynamic ad insertion capable. ESPN just recently, I mean, in the last couple quarters, moved onto the Disney stack. Now they can be programmatic.
Wow.
It's super early, but it's.
Great timelines.
Incredibly exciting. Yeah.
Questions from the audience for Michael. Any questions from the audience? Okay. Okay. We'll keep going. Is it your point of view that in the generative AI future, walled gardens take share from the open internet, or the open internet has more tools to compete with walled gardens and they start to pull some of the ad revenue from walled gardens back to the open internet?
Is it even possible for the walled gardens to take more share?
Yes.
So maybe gen-
Have you seen Facebook's numbers? They're like 30%, 33% revenue growth, highest ever. They just keep accelerating their revenue. Same with the search engine, too.
One of the advantages they've always had is that they are very easy to buy.
Yep.
They're outcomes oriented.
Yes.
They, allegedly have all the measurement to prove that.
Right
It worked great. Open internet has been a lot harder to buy, and I think one of the things that could turn that around is the agentic age, right? It becomes so much more easy to plan a campaign, to make it more performant, to make it outcome-based, and I think that, we are probably at that point where it can't possibly be any more incursion from walled gardens onto the open.
Be, yeah, be careful of that. It couldn't get worse. It can always get worse.
It, on the margins it could get worse, I think there's a starting to look at blue sky.
Okay
From here on out.
You think positive for the open internet?
Yeah. Yeah, I mean, you just, going back to the whole idea of dashboards, right? I mean, if you're able to use one dashboard that is smart enough to say from an incrementality standpoint—
Incremental reach.
Why—
Yeah
Why are you spending another $100 here?
Yeah.
You should spend it here.
Over there. Mm-hmm
Because there's more customers there and you haven't reached them and your frequency, your, all that.
I mean, one of the things that Wall Street believes is that unique data and unique supply, let's call that content, have pricing power. We're unsure anything else has pricing power. It feels like SSPs, and specifically Magnite, has both, 'cause you have exclusive CTV inventory, which is unique supply, and you have all this data that's increasingly being attached to you, not only the commerce data, but all the other curation data sets. It feels to me like you guys have increasing pricing power, but are you actually using that pricing power, or is everything?
Well, it's all part, again, it's part of the secret sauce of how you are effective as an SSP because in addition to bringing in demand, we're helping publishers price inventory.
Yeah.
Those trillions of impressions that we process every day, we know the price, we know what cleared, we know the losing bid, we know what time of day, and we know what region, what ad unit size, et cetera, et cetera. That is, through machine learning, always being pumped into our mediation layer in terms of being able to make publishers smarter a bout being able to get the best possible price. I think that there's a big opening i f you start to look at buyer agents.
Yeah
They're only as smart as the data they have.
That's true.
They can be completely super smart in terms of buying what they bought, so if they have 30 advertisers and they had 30 campaigns, and they bought off of 60 publishers, that's good data. It ain't great data because you didn't know about all the other bids that happened on those publishers and so if you're starting to look at the data that you can aggregate from an exchange like ours, and the market analysis of it. I think there's a, these agents have to be trained somehow on.
Yeah
On valuable data.
Right.
I think there's a big opportunity to directly monetize data.
Wall Street loves this. Do you have the rights to sell the data you have to LLMs?
In aggregate, we do.
Oh, you do.
My example would be far less than that. It's not selling it to, you know, Anthropic or to OpenAI.
Right.
It would be if you are a advertiser and you have your own agent—
Okay
It's, you wanna make it smarter.
Why would you do that? Why wouldn't some kid drop out of Harvard and do this for the whole industry?
They're only as good as the data they have. If they don't have the—
Well, if the ad agencies can buy it, why can't the kid that drops out of Harvard buy it?
They only see what they see. Like, they only see what they see from their campaigns.
Oh, yeah.
We're, well, we have relationships with tens of thousands of publishers.
Right
Tens of thousands of buyers.
Right
Trillions of impressions on every given day.
Right.
Not one of those publishers or one of those buyers sees the total big picture that we see.
Right. Right.
That is the big picture that can help educate how to behave, how to decision, how to optimize, how to make more money, whether you're selling, make it more efficient buys, whether you're buying.
Right.
I think that that becomes proprietary first-party Magnite data.
Right. The question is can you monetize it separately? I thought I heard you say, "Yes, we can." I immediately thought—
I think in a future world, yeah, you can. Yeah.
Okay. You have the rights to do that.
Yeah. Oh, I mean, Yes, we do have the rights to that.
Okay. That's interesting.
We have to run our business. You know, in other words, that intelligence that we have fuels the decisioning.
The underlying business.
That we make. Exactly, yeah.
You think you have the rights to sell it in aggregate to a third party also.
I think under many occasions, yes. Yeah.
Okay. That's interesting. Okay. Any other questions before we call it? Okay. I will call it there. Thank you very much.