As artificial intelligence reshapes advertising, much of the industry’s attention has focused on creative tools, AI agents and automation.
But for one of the world’s largest independent sell-side platforms (SSPs), the bigger question is how AI is changing the future of the sell-side itself.
In a conversation with tipsheet yesterday, Michael Barrett, CEO of Magnite, discussed how AI is reshaping the company’s long-term vision for the sell-side platform and its role in the future of advertising.
Topics covered include:
- Magnite’s role as AI reshapes advertising
- Why the company sees orchestration as the next evolution of the SSP
- Every buyer and publisher having AI agents
- AI’s potential to create a “golden age” for connected TV
- The path to protocol-based and agentic advertising
- Why one-to-many agentic buying represents the industry’s tipping point
- How conversational media could fit into the programmatic ecosystem
- The strategic importance of sell-side decisioning
- What Walmart’s partnership says about the future of retail media
- How Magnite is measuring success in the AI era
Scroll down for the interview, which has been lightly edited for clarity.
Magnite’s AI vision
tipsheet: When you think about AI reshaping advertising over the next five years, what role do you want Magnite to play?
Michael Barrett: We’d love to be positioned as the trusted infrastructure layer where AI-driven advertising is executed across the open internet.
Stepping back a bit, we fundamentally believe that every publisher will have an agent and every buyer will have an agent. The key for us to stay relevant on the sell-side is to remain the trusted layer between them.
We recently announced our orchestration layer, and we believe that while these agents will be incredibly powerful—and certainly capable of communicating with one another—you’re still dealing with tens of thousands of global media companies and tens of thousands of buyers. An orchestration layer that enables discovery, evaluation and transactions across premium omnichannel inventory while respecting publishers’ rules and delivering trusted outcomes at scale becomes essential.
The SSP absolutely evolves. I think there will be far fewer of them, and they’ll need to be highly sophisticated technically to help publishers and buyers navigate this transition. They’ll need scale, proven execution and deep publisher relationships.
The SSPs that built their businesses around inventory arbitrage, undifferentiated demand and shallow publisher relationships will struggle. We’re entering a new era.
We’ve already been living that reality in streaming. Most of our streaming publishers don’t use every Magnite product. They select the capabilities they need. Some primarily use our ad server. Others use different combinations of products.
Whether you call it an ad server, an SSP or something else, the important point is that publishers increasingly need modular, sophisticated technology. That’s exactly where AI is taking the industry, and we feel very good about the investments we’ve made.
Beyond the traditional SSP
tipsheet: Magnite has traditionally been known as an SSP. Recent announcements around infrastructure, orchestration and AI suggest the company is evolving beyond that description. How should people think about Magnite today?
Michael Barrett: We’ve wrestled with these labels throughout our history—exchange, SSP, mediation layer and others. I’m actually not terribly concerned about what people call us.
I’m completely comfortable with the SSP label. We work for publishers, we’re paid by publishers, and our responsibility is to create a transparent, open and fair transaction layer on their behalf.
What has changed is that our streaming business looks nothing like the traditional Rubicon Project model built around open web header bidding. It’s much more modular, much more technically sophisticated and much more customized for each publisher.
That’s the direction AI is pulling the entire SSP category.
We’ve created a strong moat there. Companies that relied on the old playbook of undifferentiated demand—your father’s SSP—are going to have a difficult time.
AI and connected TV
tipsheet: How does the AI vision you’ve described show up first in connected TV? Where do you see AI creating the biggest changes—planning, buying, optimization, measurement or creative?
Michael Barrett: All of the above.
AI is going to usher in a golden age for television. Five years from now, television will essentially mean connected TV because consumers have already decided they want to consume entertainment, sports and news through streaming.
AI affects the market in two major ways.
First, outcome-based advertising. You saw it throughout this year’s upfronts. If agencies are increasingly compensated on outcomes, they’re no longer going to buy Nielsen ratings and CPMs alone. Media companies have to become partners in delivering outcomes.
That simply isn’t possible without AI at the center of the system.
The second impact is that AI dramatically expands the advertiser base for streaming.
Even if streaming captured every dollar that once went to linear television—which it won’t—there still wouldn’t be enough inventory to absorb all of that spending. The economics require entirely new advertisers to enter the market.
AI removes many of the traditional barriers, including creative costs, campaign complexity and measurement challenges.
We’re already seeing that happen with our streaming products. Businesses that previously couldn’t advertise within streaming media are now entering the market. They’re signing up with credit cards, launching campaigns and becoming increasingly sophisticated.
AI is at the heart of all of that.
I think its biggest impact will ultimately be on streaming because that’s where the economics make the greatest difference.
Orchestration, agents and ad infrastructure
tipsheet: Along the lines of AI-enabled mediation, agentic execution through Magnite’s evolving infrastructure and now Magnite Orchestration—what’s the common thread? When you tie those pieces together, what’s the long-term vision?
Michael Barrett: The orchestration layer becomes the next level up. It orchestrates all of these capabilities.
It’s not just about agents discovering inventory. The orchestration layer also handles settlement, policy enforcement, pricing, yield management, identity, payments and compliance. Those responsibilities can’t simply be left to agent-to-agent interactions. There have to be common rules, common standards and trusted infrastructure.
I think about it like the New York Stock Exchange. Years ago, trades happened in a trading pit with brokers passing paper tickets. Today it’s electronic, but you still need a trusted exchange to ensure that when someone buys a share of General Motors, the seller actually owns it, the buyer receives it and everyone gets paid.
The infrastructure we’ve built around trust and transparency becomes the rails upon which AI operates.
ClearLine and the buy-side
tipsheet: Where does the buy side fit into that strategy? I’m thinking specifically about ClearLine.
Michael Barrett: The buy side is critically important to Magnite because we wouldn’t have publisher relationships if we couldn’t bring the demand those publishers need. We have to be trusted by both sides of the marketplace.
Originally, ClearLine was a self-service platform that allowed midsized agencies without DSP relationships to submit insertion orders that were automatically converted into programmatic campaigns. Today, most of those agencies already have those capabilities.
Now ClearLine primarily serves two purposes.
First, it’s a buying tool that gives advertisers access exclusively to Magnite’s supply. It’s particularly useful for fee-sensitive spend that otherwise wouldn’t enter the programmatic ecosystem. Since our economics are driven by the sell-side, we don’t need to maximize economics on the buying side. We can simply make it easier for those dollars to flow into programmatic.
Second, ClearLine serves as an entry point for many large agencies accessing premium private marketplaces on Magnite. Strategically, that’s actually the more important role.
People often think of ClearLine as a low-cost DSP alternative. It certainly can serve that purpose, but it’s much more strategically important because it’s embedded across many of the workflows we provide.
Protocol-based advertising
tipsheet: Magnite has been an early supporter of AdCP, and we’re seeing broader discussion around protocol-based advertising, including ARTF. Where do you think the industry is on that adoption curve?
Michael Barrett: We’re still in the discovery phase.
Both buyers and sellers want proof points. They want to know the technology works—even if it’s only a $100 transaction. Once they’ve successfully completed one transaction, they’re in the agentic game.
What we haven’t seen is any pressure from our partners to throw away today’s infrastructure and rebuild everything from scratch. Publishers trust our rails. Buyers trust our transparency. Everyone knows transactions settle correctly and people get paid.
So the question becomes: why not take the best of agentic and layer it onto the infrastructure we already have instead of starting over? That’s the conversation we’re hearing throughout the market.
tipsheet: When do you think protocol-based buying reaches meaningful scale?
Michael Barrett: That was probably my favorite question to ask everyone at Cannes because every meeting eventually turned into a discussion about agentic advertising.
The range of answers was enormous. Some people believe there will be essentially no meaningful DSP adoption by 2027. Others think the industry could reach $600 million to $700 million in spend.
Somewhere between those two extremes lies the answer.
Even at the high end, that’s still relatively modest within the context of the overall programmatic market, but it’s clearly moving in the right direction.
Think about how much time people spend moving from dashboard to dashboard today.
Agentic interfaces eliminate much of that friction.
We’re already seeing it internally. Tasks that once required large numbers of contractors are now automated. We haven’t needed to replace those people because AI is doing much of that work.
It’s hard to believe the broader industry won’t eventually experience those same productivity gains.
When agentic buying reaches scale
tipsheet: Magnite has its seller agent. When does that become a meaningful part of the industry?
Michael Barrett: We’re certainly preparing for it. We have the infrastructure in place and the protocols in place.
What we’re seeing today—and I think much of the industry is seeing the same thing—is primarily one-to-one transactions.
A buyer works directly with a seller. An insertion order that previously required manual work is now automated through an agentic workflow.
That’s already meaningful. These are large businesses on both sides of the transaction.
One of the hidden benefits of AI is that it expands programmatic’s total addressable market. Many direct insertion order deals that historically sat outside programmatic can now move into it because agents make the workflow dramatically easier. An RFP can be generated automatically, inventory surfaced and evaluated, and the transaction completed.
The real tipping point comes when we move from one-to-one to one-to-many.
When a buyer’s agent can simultaneously discover, negotiate and transact across many sellers, that’s when the industry will know the technology has truly scaled.
Leadership changes at Magnite
tipsheet: Magnite has seen several leadership changes recently. What’s driving that evolution?
Michael Barrett: I actually think the bigger story is how stable our leadership team has been over the past decade.
We came together through a series of acquisitions—Rubicon Project, Telaria, SpotX and SpringServe—and leadership teams built through acquisitions don’t usually stay together this long.
Today, most employees have only ever worked at a company called Magnite, not one of those predecessor businesses.
As with any company, though, you periodically look at your organizational structure. Against the backdrop of AI and this next phase of the industry, we’re creating opportunities for leaders like Katie Evans and Sean Buckley to take on additional responsibility. We think that’s the right move for Magnite and for our customers.
At the same time, some executives decide they still want the opportunity to run a company themselves or pursue something new.
We couldn’t think more highly of the people who’ve moved on. They’ve been incredibly important to Magnite, and they’ll remain part of our broader network.
Conversational media
tipsheet: Do you see a role for Magnite in conversational media such as ChatGPT ads and other AI-native advertising experiences?
Michael Barrett: So, putting aside specific companies, what’s interesting is that many of these new AI companies are introducing third-party advertising from the very beginning.
Historically, companies tried to build their own advertising businesses first. They wanted to recreate the Meta or Google model by selling directly. Only later, if necessary, would they open up to third-party demand.
Now we’re seeing companies begin with third-party demand instead — that’s encouraging for the programmatic ecosystem because it recognizes how difficult it is to build an advertising business from scratch when you’re competing with companies like Meta and Google.
From Magnite’s perspective, we typically become involved after the initial phase.
A company might start by working directly with one DSP, then two or three. Eventually they want to operate globally across hundreds of demand partners. That’s where an orchestration layer becomes valuable.
You’ve seen that progression with companies like Pinterest and Roku, which initially pursued more closed approaches before embracing broader third-party demand.
I wouldn’t be surprised if conversational AI platforms follow a similar path as generative interfaces mature.
Walmart, retail media and sell-side decisioning
tipsheet: Turning to your recent Walmart partnership, what’s its significance as you think about the future of advertising—and Magnite’s role in it?
Michael Barrett: There were a couple of announcements (April 28, May 28), but the biggest one we discussed at Cannes was the evolution of Walmart’s thinking around its first-party data.
The first inning of retail media was essentially: take your valuable data, give it to a single DSP and require advertisers to use that platform if they wanted access to it. The data was valuable enough that advertisers were willing to accept that constraint.
What’s changing is the realization that there’s no reason to limit demand to a single DSP.
Instead, you can move that data out of the DSP, place it with a trusted sell-side partner, move it closer to premium inventory—like Vizio, which Walmart now owns—and allow multiple DSPs to activate against it.
Advertisers can bring whichever DSP they prefer. Obviously, we’re excited about the Walmart partnership itself. But I think the more important story is what it says about sell-side decisioning.
Keeping data closer to the supply is becoming increasingly valuable.
All of the major streaming platforms have incredibly rich first-party data. Increasingly, campaigns are being activated using that data where it resides—on the supply side rather than the demand side.
I think the public markets are gradually recognizing that SSPs aren’t simply dumb pipes bringing undifferentiated demand. They play a much more strategic role than that.
As we move toward agentic advertising and orchestration layers, the importance of sell-side decisioning only increases.
Sell-side decisioning beyond retail media
tipsheet: Can that same sell-side intelligence extend beyond connected TV? Is there an omnichannel opportunity for sell-side decisioning?
Michael Barrett: Absolutely. If you look at curation tools—and there are plenty of different names for them today—one of the biggest differentiators for Magnite is our omnichannel approach.
Buyers increasingly want to reach consumers wherever they are across programmatic channels. They don’t want to manage six or seven specialized exchanges to access different inventory types.
If they can execute a single campaign across streaming and digital display through one platform, that’s much simpler. That’s what buyers are increasingly asking for.
There’s no question that combining those channels creates a much stronger offering than treating them as separate silos.
The growing value of first-party data
tipsheet: It feels like first-party data and customer graphs are becoming more valuable than ever. We see that reflected in the Walmart partnership. How can publishers participate in that value?
Michael Barrett: Some publishers are large enough—companies like Netflix, Disney and Roku—that they can directly combine their data with retailers like Walmart and create powerful audience offerings.
For most publishers, though, the benefit comes through higher CPMs.
When campaigns are activated using valuable first-party data, that inventory becomes more valuable. If it’s sold through bidding environments, prices can increase even further. Publishers end up monetizing inventory that previously wasn’t fully valued because the right data is now attached to it.
The Walmart acquisition of Vizio is a great example of where this is headed. Most retail media advertisers are television advertisers at heart. Retail media networks have their own owned-and-operated properties, but brands like Gillette ultimately want to advertise on Netflix, Disney, Paramount and other premium streaming services.
We’re beginning to see retail media expand beyond owned-and-operated environments into the broader premium web.
That’s a significant opportunity for publishers.
Looking toward 2030
tipsheet: As a public company, I realize you may be limited at what you can say, but if we were having this conversation again in 2030, what would you hope people would say Magnite understood before the rest of the industry?
Michael Barrett: As a public company, I can’t help but think about metrics.
There’s a lot of discussion around AI today, but many companies are talking about AI while their costs continue rising and their revenue growth slows.
If we’re still talking about AI in 2030 but our growth is flat and our headcount has ballooned, then we’ve missed the point. Success has to show up in the numbers. Whether it’s 50% growth with 20% margins or some other combination, AI should improve both growth and operating leverage.
The real test is whether we’re applying AI internally to become more efficient while also positioning ourselves to participate in the industry’s growth.
Until that happens, AI is just a talking point. It’s not reality.