A side-by-side look at two SSPs reporting Q4 2025 results within 24 hours of each other exposes a striking paradox: the company talking most loudly about AI is the one losing ground on the scoreboard.
The two most prominent independent sell-side platforms in programmatic advertising released their fourth-quarter 2025 earnings within a day of each other - Magnite on February 25, PubMatic on February 26. Both operate in the same market, serve overlapping publishers and buyers, and have spent the last two years making the same strategic bet on connected television. But placing their results side by side produces a picture that is considerably more complicated than the bullish narratives each management team presented to investors.
At its most blunt, the comparison raises an uncomfortable question for the programmatic advertising market: if agentic AI is the transformational force that PubMatic describes, why is the company investing most aggressively in that narrative also the one with declining revenues, a stock down 51% over the past year, and a balance sheet one-third the size of its rival's? And why is Magnite - which posted genuine reported revenue growth, told analysts AI will simply "make things work better," and spent its technology budget on an acquisition aimed at small business CTV buyers - trading near the top of the two-company comparison on almost every financial metric that matters?
Start with what cannot be adjusted away. Magnite's Q4 2025 revenue was $205.4 million, up 6% year-over-year on a clean, reported basis. PubMatic's Q4 2025 revenue was $80.0 million, down 6% year-over-year. No exclusions required for Magnite. PubMatic's comparable growth figure - the 18% it cites as the genuine underlying rate - requires removing the impact of a legacy DSP partner that changed its auction methodology in mid-2024 and stripping out political advertising that was unusually heavy in Q4 2024 and essentially absent in Q4 2025.
Both adjustments are legitimate. Neither is fabricated. But the asymmetry matters: Magnite does not need to reach for an adjusted metric to show growth. Its reported number is the growth number. For the full year, the gap widens further. Magnite generated $714.0 million in 2025, up 7% from $668.2 million in 2024. PubMatic's full-year revenue was $282.9 million, down 3% from $291.3 million in 2024.
On profitability, Magnite's Q4 adjusted EBITDA reached $83.8 million, a 42.9% margin against contribution ex-TAC. PubMatic's Q4 adjusted EBITDA was $27.8 million, representing a 35% margin against total revenue - its own best quarterly margin in recent memory, and a genuinely impressive operational result. But the absolute figures are not comparable. Magnite's quarterly EBITDA is three times PubMatic's. For the full year, Magnite posted $232.1 million in adjusted EBITDA against PubMatic's $61.6 million.
The cash positions tell the same story in a different register. PubMatic ended 2025 with $145.5 million in cash and zero debt - a genuinely clean balance sheet, a product of the company's decision to build and own its infrastructure rather than acquire. Magnite ended with $553.4 million in cash and approximately $565.5 million in debt, a leverage ratio that rounds to net zero but reflects the significant debt load accumulated through the 2021 acquisitions of SpotX and SpringServe. PubMatic's capital structure is leaner. Magnite's capital deployment created a CTV ad server position that now processes nearly half its quarterly revenue.
The stock market's verdict over the past year has not been kind to either company - but it has been worse to PubMatic. PUBM shares closed at $6.58 on February 26, following a 7.52% gain on the day of earnings, and remain roughly 51% below their 52-week high of $14.80. Magnite's MGNI shares hit a 52-week high of $26.65 in August 2025 before falling back to around $16-17 by the time of their Q4 report, a decline of approximately 35% from the peak. Both stocks have disappointed. Magnite's absolute decline is steeper in dollar terms, but PubMatic's proportional drawdown from peak is greater, and PUBM entered 2026 trading near its 52-week low of $6.15.
The one area where PubMatic presents a headline metric that appears superior to Magnite's is CTV growth rate. PubMatic's CTV revenue grew more than 50% year-over-year in 2025, excluding political advertising, and has sustained a four-year compound annual growth rate above 50%. Magnite's CTV contribution ex-TAC grew 17% for the full year 2025, and 32% in Q4 specifically when excluding political.
The framing favors PubMatic if growth rates are the primary measure. It is less favorable once scale enters the equation. Magnite's Q4 CTV contribution ex-TAC alone was $93.6 million - a figure that approaches half of PubMatic's entire quarterly revenue. For the full year, Magnite's CTV contribution ex-TAC reached $304.2 million, more than PubMatic's total 2025 revenue. Magnite's CEO Michael Barrett stated on the February 25 earnings call that "making streaming the majority of our business - that is a defining moment for Magnite." CTV crossed 50% of Magnite's total business in Q1 2026. PubMatic's omnichannel video segment - which includes CTV alongside desktop and mobile video - represented 39% of total revenue in 2025.
The structural reason for this gap traces back to 2021. When Magnite paid for SpotX and SpringServe, it acquired a CTV ad server position embedded inside premium streaming publishers. Ad server relationships are stickier than SSP relationships, more deeply integrated into publisher workflow, and harder to displace. An independent analysis from Jounce Media published in March 2025 found Magnite covers 99% of the CTV supply market with a 24-point lead over its nearest competitor. PubMatic's CTV growth is real and impressive in rate terms. In structural depth - the kind that creates durable revenue rather than incremental share - Magnite's position inside the major streaming platforms is qualitatively different.
PubMatic's partnership count - 28 of the top 30 global streamers - is a strong number. CEO Rajeev Goel said on the call: "We partner now with 28 of the top 30 global streamers, including Roku, Samsung TV Plus, DirecTV, Fox Sports, Tubi, VIZIO, and more." But a partnership with a streamer as an SSP and an ad server relationship inside that same streamer are different things. The former means competing for available demand. The latter means managing the publisher's entire monetization stack.
PubMatic's Q4 call carried an underappreciated disclosure that cuts in two directions at once. CFO Steve Pantelick stated: "Display revenues in the fourth quarter returned to year-over-year growth in the mid-single-digit %. Excluding the legacy DSP referenced earlier, display revenues grew over 20% in the fourth quarter, significantly outpacing the market rate of growth."
Goel reinforced this during the Q&A: "Hard to see any bleed over from AI pressures in that regard. What it speaks to is that when you think about all the things that we've been doing, executing against our strategic priorities, all of those efforts are starting to benefit across all formats and channels. In effect, it's lifting all boats."
That is an honest and arguably reassuring observation. The display recovery suggests that PubMatic's operational improvements - adding 50 new DSP partners, growing Activate buying activity threefold, expanding the buyer-focused sales team by 20% - are generating broad-based platform gains rather than format-specific tailwinds. But the display recovery also implicitly undermines the urgency of the agentic AI narrative. If display is growing 20% when you remove the single-DSP distortion, the core open-web advertising business is not structurally broken. The question then becomes: how much of the agentic story is genuine incremental opportunity, and how much is a compelling frame for a business that needs a growth narrative during a difficult reporting period?
Magnite's display equivalent - its DV+ segment covering desktop and mobile web - posted a 1% decline in Q4 contribution ex-TAC, or 4% growth excluding political. Barrett framed this as structurally positive: "We observed accelerated budget reallocation from DV+ into CTV across agencies, DSPs, and brands," and described CTV revenue as "more protectable and sustainable" than DV+. The framing is different from PubMatic's but rests on the same underlying dynamic. Budget is migrating from open-web display into streaming. Both companies are positioned to benefit from that migration. Magnite is simply further along in the transition because it made the CTV investment earlier and at greater scale.
PubMatic's display recovery narrative carries a complication the company did not address directly during its February 2026 earnings call. While Steve Pantelick cited mid-single-digit year-over-year display growth in Q4 2025 - and 20% growth excluding the legacy DSP contract - the open web display market that PubMatic operates in is not the same market as the one capturing the majority of advertiser budgets. That market belongs, overwhelmingly, to Meta and Google.
Meta generated $58.1 billion in advertising revenue in Q4 2025 alone, a 24% year-over-year increase driven by AI improvements across ranking, creative tools, and campaign automation. Full-year 2025 advertising revenue reached $196.2 billion. To place that in context: PubMatic's entire annual revenue of $282.9 million represents approximately 0.14% of what Meta generates in advertising in a single year. Instagram Reels watch time in the United States grew more than 30% year-over-year, and Reels now runs at an annual revenue rate exceeding $50 billion on its own. These are not peripheral players in the display and visual advertising space - they are the display and visual advertising space, for most marketers with performance goals.
Google compounds the structural challenge from a different direction. Performance Max campaigns, which Google has steadily positioned as its primary automated buying product, optimize across Search, YouTube, Display, Discover, Gmail, Maps, and Waze simultaneously. Demand Gen campaigns - Google's dedicated visual advertising product covering YouTube, Discover, Gmail, and now the Google Display Network since March 2025 - give advertisers a more focused alternative for upper-funnel and visual inventory within Google's own ecosystem. Together, Performance Max and Demand Gen constitute Google's answer to Meta's Advantage+ suite: AI-optimized, automated, and confined to inventory that Google owns or directly controls.
The structural tension for PubMatic and for the open web programmatic ecosystem generally is not subtle. Google's network advertising revenues - the segment that funds third-party publishers through AdSense, AdMob, and Google Ad Manager - declined 1% year-over-year to $7.4 billion in Q2 2025, while Google's owned properties grew 12% in the same period. A subsequent analysis found that 90% of Google's advertising revenue now flows to its own properties, the first time that threshold has been crossed in over a decade. Advertisers are spending more with Google than ever before, but an increasingly small portion of that spending reaches the open web publishers where SSPs like PubMatic operate.
The pattern is not coincidental. Google's September 2025 Power Pack strategy explicitly frames AI Max for Search, Performance Max, and Demand Gen as a unified system - one where the AI handles targeting, bidding, and creative optimization across Google's surface area. During the September 2025 announcement, Google noted Performance Max campaigns had received more than 90 quality improvements in the preceding year, driving conversion increases exceeding 10%. The pitch to advertisers is efficiency and reach, with minimal need to engage external platforms or manage publisher relationships directly. That is a direct structural challenge to any SSP whose value proposition rests on access to open-web publisher inventory.
Meta's pitch is more concentrated but no less formidable. Advantage+ sales campaigns have generated average 22% return on ad spend improvements according to Meta's own reporting. Ad impressions across Meta's family of apps increased 18% in Q4 2025 while the average price per ad rose 6%. The automated campaign structure - which became the default for sales, leads, and app objectives during 2025 - operates entirely within Meta's closed environment. Instagram, Facebook, Threads, WhatsApp, and Messenger form a self-contained ecosystem where advertisers can reach 3.35 billion daily active people without touching open-web programmatic infrastructure at all. Research published in August 2025 found that 88% of the $45 billion consumer mobile app advertising market flows to Google and Meta, leaving 12% for the rest of the ecosystem combined.
What makes this competitive dynamic particularly difficult for PubMatic is that Meta and Google are not bidding against it in auctions. They are not competitors in the traditional programmatic sense. They represent an alternative to programmatic entirely - a parallel system where the inventory, the data, the targeting, the measurement, and the optimization all sit within a single company's infrastructure. An advertiser moving budget from open-web display into Meta's Advantage+ or Google's Performance Max does not reduce competition in the SSP marketplace; it removes that budget from the SSP marketplace altogether.
Pantelick's comment that it was "hard to see any bleed over from AI pressures" on display demand may accurately describe what PubMatic observed in Q4 2025. The more precise question is whether display budgets that might have flowed to open-web publishers are instead being captured upstream - by platforms that have already persuaded advertisers they need not participate in programmatic at all. Teads, another programmatic player dependent on publisher traffic, disclosed 10-15% pageview declines across its publisher network in Q3 2025, directly attributing the decline to AI summaries and discovery changes. Declining publisher traffic means declining publisher advertising inventory, which means a shrinking addressable market for any SSP.
PubMatic's AgenticOS positioning attempts to reframe this structural challenge. If AI agents become buyers, and if those agents use open standards like the Ad Context Protocol rather than platform-specific APIs, then the walled gardens face a different kind of pressure - one where standardized machine-readable access to supply becomes an advantage. Rajeev Goel's argument that PubMatic's "real-time execution layer cannot be replicated by vibe-coded software" is partly a claim about technical infrastructure and partly a claim about what happens when automated buyers need open, auditable, third-party supply rather than inventory that only a platform can verify on its own terms.
Whether that argument lands with advertisers - and their AI agents - remains to be seen. For now, Meta and Google are not waiting for the programmatic ecosystem to set the terms of competition. They are generating hundreds of billions in advertising revenue annually, growing at double-digit rates, and offering performance metrics that make it difficult for open-web display to compete on a pure return-on-investment basis. PubMatic's display recovery in Q4 2025 is a real data point. But it exists within a structural context where the platforms dominating visual and display advertising are not SSP partners or DSP customers - they are entities that have built their own closed loops, and whose growth increasingly comes at the expense of the open web ecosystem that PubMatic depends on.
The most philosophically revealing difference between the two Q4 calls was not in the financial results. It was in how each CEO described artificial intelligence.
Goel opened PubMatic's prepared remarks with a declaration: "Advertising is entering a new phase, one defined by AI-driven autonomous systems operating in real time. We sit at the center of a highly competitive millisecond-level auction environment where value is determined by measurable outcomes such as yield, performance, and efficiency. PubMatic is enabling AI adoption across the open internet. Our proprietary data, scaled infrastructure, and thousands of deep integrations across buyers and publishers form a real-time execution layer that cannot be replicated by vibe-coded software."
He continued: "By 2028, I expect 25% of all digital advertising to be executed autonomously via agentic AI, and by 2030, I expect that to jump to 50%. As an early AI leader, this unlocks transformative growth for PubMatic long before our peers."
Barrett took a materially different tone on Magnite's call. He dismissed suggestions that AI would disintermediate SSP infrastructure: "What AI agents are going to do is make it work better. It's going to alleviate menial tasks from the traders, the planners, the ops people, and it's going to put more working dollars to play." Magnite did run its first agent-to-agent CTV advertising campaign during Q4, with Scope3 as the buyer agent on behalf of MiQ, executing across LG and Warner Bros. Discovery inventory - similar in structure to PubMatic's December Butler/Till campaign. But Barrett did not present it as a transformational moment. He presented it as one capability among many.
Magnite's September 2025 acquisition of streamr.ai is the clearest expression of how it is deploying AI commercially: AI-powered tools for creative generation and campaign setup, aimed at small and medium businesses who previously lacked the production capacity to run CTV ads. That is a demand-side expansion play - broadening who can afford to advertise on streaming television - rather than a protocol-level infrastructure bet. It is less intellectually exciting than AgenticOS. It is also an acquisition that has an immediately legible commercial logic: more advertisers buying CTV inventory through Magnite.
The irony crystallizes around one data point. PubMatic has transacted more than 250 agentic deals on its platform since December 2025 - and management acknowledged on the call that meaningful revenue from these transactions is a 2026 and beyond story. Analyst Steve Roman from Oppenheimer asked directly: "The over 250 deals metric is encouraging, but can you help us understand whether these deals are already driving meaningful revenue, or is that more of a this year and next year story?" Pantelick's response: "We're out of the gate, and from our perspective, we are absolutely the leader among our peers. As we've done with many other innovations in our company, we build out the foundation, and then we scale it over a number of quarters and years."
That is an honest answer. It is also, functionally, a statement that agentic AI is not yet in the revenue line in any material way. Magnite generated $232.1 million in adjusted EBITDA in 2025 from a business it describes as "making CTV the majority" - a transition executed through acquisitions and platform integration, with AI as an efficiency layer rather than a primary growth driver. PubMatic generated $61.6 million in adjusted EBITDA from a business that is loudly positioning AI agentic as its defining advantage. The company talking most about the future is generating less cash from the present.
Both companies have built direct-buying platforms that attempt to reduce DSP dependence and connect buyers more directly to supply. The mechanics differ and reveal something about each company's strategic hypothesis.
PubMatic's Activate platform allows advertisers and agents to buy directly in the SSP, bypassing DSP intermediation entirely. Buying activity on Activate grew more than threefold in 2025 over 2024. The IPG Kinesso campaign using Activate delivered 72% more clicks, 11% more impressions, and 20% lower CPMs compared to the DSP-intermediated baseline, according to company disclosures. Goel explained the strategic logic on the call: "With Activate, buyers can now buy directly in our SSP, which really simplifies the end-to-end workflow and agentic communication. This is really critical because we are not in a position where we have to wait for standards to emerge so that sell-side tech and buy-side tech can communicate in a standardized protocol. Because we have Activate, which is direct buying in our SSP, we're free to innovate beyond any standards, and so we can move a lot more quickly."
Magnite's equivalent is ClearLine, a self-service direct buying tool that Barrett described as gaining "significant traction with agency marketplaces" in Q4. Magnite expanded ClearLine's capabilities in October 2025, unifying curation and activation. The difference is emphasis and framing: PubMatic positions Activate as the transaction layer that makes agentic AI commercially viable without waiting for industry protocol adoption by Google or The Trade Desk. Magnite positions ClearLine as an efficiency tool within the existing programmatic stack.
The significance is that PubMatic does not need dominant DSPs to adopt the Ad Context Protocol - which Google, Amazon, and The Trade Desk have notably avoided - to generate agentic revenue. Activate creates the direct path. The risk is that if DSPs build their own competing direct-buying infrastructure, or if open agentic standards do eventually emerge with major-platform participation, Activate's differentiation diminishes. For now, it is a genuine competitive advantage in the direction PubMatic is moving.
Both companies filed antitrust lawsuits against Google in 2025 and both are awaiting Judge Brinkema's remedies decision - widely expected sometime in Q1 2026. Neither has included any potential upside in financial guidance. Both have, however, quantified what market share recapture would be worth.
Magnite has stated publicly that every 1% of market share gained from a more competitive ad exchange market could represent approximately $100 million in incremental contribution ex-TAC. PubMatic's Pantelick estimated on the call that each 1% of market share would add $50 million to $75 million in revenue at 80% to 90% incremental margins. Goel added context: "We estimate Google's a 60% market share player. Each 1% of market share would add $50 million to $75 million in very high margin revenue to our platform."
The per-percentage-point value is higher for Magnite in absolute terms, reflecting its larger existing base. What the numbers share is the underlying assumption: if regulatory action opens the open-web ad exchange market to genuine competition, both companies benefit enormously. The European Commission's €2.95 billion fine against Google and the parallel U.S. remedies process represent the most significant potential external catalyst for both companies' revenue trajectories. Goel was measured: "Many people are expecting it to be more so behavioral remedies, rather than structural remedies, where the primary structural remedy is Google divestiture of ADX. We don't know anything special. We're kinda waiting and seeing."
The stock market's current valuation of the two companies implies something about how investors are weighing these competing narratives. PUBM at $6.58 gives PubMatic a market capitalization of roughly $300 million to $350 million against $282.9 million in 2025 revenue - approximately 1x revenue. MGNI at approximately $16 to $17 per share gives Magnite a market capitalization of around $2.4 billion against $714.0 million in 2025 revenue - roughly 3x revenue. Both are trading at depressed multiples relative to their growth profiles and certainly relative to The Trade Desk, which posted $2.9 billion in 2025 revenue with 18% growth and trades at far richer multiples.
The valuation gap between Magnite and PubMatic - approximately 3x revenue versus 1x revenue - reflects the market's assessment of which company has better resolved the key structural question: does it have a defensible position in the part of the market that is actually growing? Magnite's CTV ad server position at scale provides an answer to that question that investors find more legible. PubMatic's answer - that agentic AI will create a new revenue category on top of a recovering core business - requires more trust in a narrative that is, by management's own admission, still building its foundation.
The honest takeaway is that Magnite and PubMatic are not equivalent competitors at different stages of the same journey. They are companies that made different bets at a critical juncture - Magnite's 2021 acquisitions versus PubMatic's decision to own and operate its own infrastructure organically - and are now living with the consequences. Magnite's bet has produced a CTV position that is, by most measurable indicators, structurally superior. It has also produced $565 million in debt and a share price down significantly from its peak. PubMatic's bet has produced a pristine balance sheet, impressive impression-processing efficiency, and a company that is genuinely early in what could be a significant agentic AI transition. It has also produced revenues that are shrinking on a reported basis and a stock near a 52-week low.
The question Magnite's results implicitly pose to PubMatic investors is not whether agentic AI is a real opportunity. It probably is. The question is whether PubMatic's specific structural advantages - NVIDIA-accelerated infrastructure, nearly 2,000 publisher integrations, Activate direct-buying, and 250-plus early agentic deals - are sufficient to generate the kind of durable revenue at scale that Magnite has generated from a more conventional infrastructure play in CTV. Magnite is growing 6% to 7% on clean reported numbers. PubMatic is growing 18% on adjusted numbers that require two exclusions to reach. One company is already in the position it wants to be in. The other is describing the position it intends to reach.
For the marketing community, both companies' trajectories confirm the same directional reality: programmatic infrastructure is consolidating around CTV and mobile-app environments, open-web display is stabilizing but not leading, and whoever builds the most efficient path between advertisers and streaming inventory will capture a disproportionate share of where media budgets are going. Magnite has built that path at scale, through acquisition and ad server depth. PubMatic is building it through technology, protocol leadership, and a direct-buying platform that genuinely reduces friction. Which approach generates more value for publishers, advertisers, and shareholders over the next three to five years is the open question - and the 24 hours between their earnings calls did not close it.
Who: PubMatic (Nasdaq: PUBM), led by CEO Rajeev Goel and CFO Steve Pantelick, and Magnite (Nasdaq: MGNI), led by CEO Michael Barrett and CFO David Day. Both are independent sell-side advertising platforms competing for publisher relationships, buyer demand, and CTV inventory monetization contracts.
What: A comparison of Q4 and full-year 2025 financial results, AI strategy, CTV positioning, display performance, direct-buying platforms, Google antitrust exposure, and stock market valuation. Magnite posted Q4 revenue of $205.4 million (up 6%) and full-year revenue of $714.0 million (up 7%) with adjusted EBITDA of $232.1 million. PubMatic posted Q4 revenue of $80.0 million (down 6% reported, up 18% excluding legacy DSP and political) and full-year revenue of $282.9 million (down 3%) with adjusted EBITDA of $61.6 million. The central tension: PubMatic is making the louder AI claims with the smaller financial footprint. Magnite is growing faster on reported metrics while describing AI as an efficiency tool rather than a strategic transformation.
When: Magnite reported on February 25, 2026. PubMatic reported on February 26, 2026. Both cover the period ended December 31, 2025.
Where: Both companies operate globally across connected television, mobile app, and open-web display advertising environments. Magnite is headquartered in New York. PubMatic is headquartered in Redwood City, California.
Why: The comparison matters for the marketing community because the two companies' divergent strategies are a real-time test of competing hypotheses about where value will be created in programmatic advertising infrastructure. If agentic AI generates meaningful incremental revenue at the platform level within the 2026-2028 window, PubMatic's early positioning could prove prescient. If CTV ad server depth and sheer inventory scale remain the dominant value drivers, Magnite's 2021 acquisitions will have been the more important strategic decision. Both outcomes are possible. The earnings calls make clear which company has already won the argument in the present - and which is making the case that the future belongs to it.
