PubMatic, the Nasdaq-listed company that sells advertising space on behalf of websites, app developers and streaming services, on November 9, 2026 reported GAAP net income of $5.5 million for the third quarter, against a net loss of $6.5 million a year earlier. Revenue for the three months to September 30 rose 17% to $79.6 million from $68.0 million, above the $75 million to $77 million range the company had given in August. Adjusted EBITDA, the profit measure PubMatic uses to guide investors, rose 133% to $26.0 million, a margin of 33% against 16% in the third quarter of 2025. The board added $150 million to the share repurchase program and extended it to the end of 2028. For the fourth quarter, PubMatic expects revenue of $91 million to $93 million, about 15% growth at the midpoint, with political advertising making up roughly 7% to 8% of that total. The results were published through Business Wire after the market closed, and the quarterly report on Form 10-Q was filed with the Securities and Exchange Commission the same day.

In Short

PubMatic, a company whose software sells ad space on websites, apps and streaming TV, made a profit of $5.5 million between July and September 2026 after losing money in the same months of 2025, as its revenue grew 17% to $79.6 million. That matters for advertisers and publishers that buy and sell outside the large walled gardens, because PubMatic says buyers are now making more of their decisions on the seller's platform, using its audience data and AI agents rather than only their own buying tools. The company expects fourth-quarter revenue of $91 million to $93 million, about 15% above a year earlier, with election ads making up 7% to 8% of it. Part of the improvement is not operational: a decision to depreciate servers over five years instead of three added $1.5 million to profit, and the percentages it gives for its three fastest-growing channels add up to more than the combined total it reports.

A quarter that cleared its own forecast

The quarter landed above the range PubMatic set three months earlier. In August, when the company reported second-quarter revenue of $78.6 million, up 11%, it guided third-quarter revenue to $75 million to $77 million and adjusted EBITDA to $17.0 million to $19.0 million. Actual revenue of $79.577 million exceeded the top of that range by about $2.6 million, or 3%. Adjusted EBITDA of $25.956 million exceeded the top of its range by about $7.0 million, or 37%, PPC Land's comparison shows. The gap on profit was far wider than the gap on sales, which says more about costs than about demand.

Wall Street had expected less. According to Investing.com, analysts' consensus for revenue was $76.1 million, and the consensus for earnings per share was a loss of $0.02. PubMatic reported non-GAAP earnings of $0.25 per diluted share and GAAP earnings of $0.11.

"We delivered an outstanding third quarter, with revenue growth accelerating for the third consecutive quarter and adjusted EBITDA more than doubling year-over-year," Rajeev Goel, co-founder and chief executive of PubMatic, said in the release. "Our investments in advertiser performance are translating into faster, more profitable growth and a more diversified business."

The phrase "accelerating for the third consecutive quarter" needs a careful reading. It describes the rate of change, not three quarters of growth. Revenue in the fourth quarter of 2025 was $80.0 million, 6% below a year earlier. In the first quarter of 2026, revenue of $62.6 million was down 2%. The second quarter brought the return to growth at 11%, and the third quarter reached 17%. Each figure was better than the one before, which is what PubMatic means. Two of the three quarters in its run of "acceleration" were, however, a smaller decline and then a first increase. PPC Land described the second quarter at the time as revenue gains after three quarters of decline; the third quarter of 2025 itself had produced revenue of $68 million, down 5% from $71.8 million.

Over nine months, the picture is more modest. Revenue for January to September 2026 was $220.7 million, up 9% from $202.9 million. PubMatic's chief financial officer, Steve Pantelick, put the year in two parts on the call: growth of 5% in the first half, "which absorbed the legacy DSP change", and roughly 16% in the second half, as implied by the fourth-quarter outlook. PPC Land's calculation from the quarterly figures confirms both numbers: first-half revenue of $141.2 million was 4.6% above $134.9 million a year earlier, and third-quarter revenue plus the $92 million guidance midpoint gives $171.6 million, 15.9% above the $148.0 million of the second half of 2025.

"Our third quarter is the clearest evidence yet that the transformation of our business is working," Pantelick said in the release. "And this is not a one-quarter event."

From gross profit to net income

The income statement explains why profit moved so much faster than revenue. Revenue rose by $11.6 million. Cost of revenue fell by $1.4 million. Operating expenses fell by $4.1 million. Put together, operating income swung by $17.2 million, from a loss of $8.4 million in the third quarter of 2025 to a profit of $8.7 million, an operating margin of 11%.

Gross margin and the five-year server

Gross profit rose 31% to $55.6 million, and gross margin reached 70% against 63% a year earlier, according to the 10-Q. That seven-point improvement has two sources, and Pantelick separated them on the call. "Our revenue growth and productivity improvements drove gross margin up by 5 percentage points year-over-year," he said. "We also benefited from an additional 2 percentage points from a change in useful life of our server infrastructure."

The change in useful life is set out in the notes to the financial statements. In July 2026, PubMatic completed an assessment of its network hardware and extended the estimated useful life from three years to five years, effective July 1, 2026. The reasons given in the 10-Q are "investments in software that increased efficiencies in how it operates as well as advances in technology of the Company's network hardware." For the third quarter, the change reduced depreciation expense by $2.1 million and increased net income by $1.5 million, or $0.03 per basic and diluted share.

That $0.03 is a meaningful share of a $0.11 result. Without the change, and with everything else equal, GAAP net income would have been about $4.0 million, PPC Land's arithmetic shows. Adjusted EBITDA is not affected, because depreciation is added back in that measure.

Pantelick framed the effect as temporary. "From just an analytical perspective to understand the impact on gross margin, over time, over the next couple of quarters, the period of time, you can make sort of an adjustment, I will call it $1.5 million-$2 million, in terms of that impact," he said, according to the Investing.com transcript. "That will dissipate as these servers are end of life." The 10-Q makes a related forecast for the whole year: PubMatic expects cost of revenue to be lower in 2026 than in 2025 in absolute dollars, "driven by depreciation from updating the estimated useful life of certain hardware network from three to five years."

Cost of revenue fell $1.4 million, or 6%, to $24.0 million. The filing lists the moving parts: a $1.9 million decrease in depreciation and amortization, mostly from the useful-life change, and $0.3 million less in facilities costs, partly offset by $1.3 million more in data center costs. Data center spending rose while servers lasted longer on paper. Over nine months, data center costs were up $4.7 million.

Operating expenses fell 8%

Total operating expenses were $46.9 million, 8% lower than the $51.0 million of a year earlier. Every line fell:

  • Technology and development dropped 18% to $7.9 million, mainly from $2.9 million lower personnel costs, partly offset by $1.0 million less capitalized internal-use software (less spending capitalized means more of it shows up as an expense).
  • Sales and marketing dropped 9% to $23.4 million, mainly from $2.2 million lower personnel costs.
  • General and administrative was almost flat at $15.5 million, a decrease of $0.1 million that the 10-Q calls "not material."

The segment table in the 10-Q gives a cleaner view. Employee compensation, which covers payroll, share-based compensation, bonuses and benefits across the company, fell 13% to $43.6 million from $49.8 million. "Other segment items", which include data center operations, professional services, facilities and marketing, rose 17% to $18.8 million from $16.1 million. Labor became cheaper; infrastructure and outside services became more expensive.

Stock-based compensation fell 21% to $7.5 million from $9.5 million. Over nine months, it was $24.4 million against $29.0 million.

Pantelick said some of the savings were timing. "In Q3, operating expenses were 8% lower year-over-year, primarily from productivity gains, with some hiring and go-to-market spend shifting into the fourth quarter," he said. The earnings presentation adds two facts that the release does not: the sales team grew 5% year over year "while total headcount down YOY." Neither the presentation nor the 10-Q gives a headcount number.

For the full year, the 10-Q expects technology and development expenses to be flat in absolute dollars against 2025, sales and marketing to rise "primarily due to additional headcount investments", and general and administrative to rise, "primarily due to professional services and employee compensation."

A 44% tax rate

Income before taxes was $9.9 million. The provision for income taxes was $4.4 million, an effective rate of 44% against the 21% federal statutory rate. The 10-Q attributes the gap to "tax expense from nondeductible stock-based compensation and Section 162(m) limitation on the tax deductibility of officers' compensation, partially offset by the tax benefits from research tax credits and deductions for equity awards." For the nine months, the effective rate was negative 147%: PubMatic recorded a $4.9 million tax charge on a $3.3 million pre-tax loss.

The method changed this year as well. PubMatic normally estimates an annual effective tax rate and applies it to each quarter. For the nine months ended September 30, 2026, it concluded that "the historical method would not provide a reliable estimate" and used a discrete effective tax rate method instead. The Internal Revenue Service is examining the company's 2023 tax year; the 10-Q says the audit "is in its preliminary stages and no material adjustments have been proposed to date." Income taxes paid in cash over nine months fell to $3.1 million from $8.3 million.

Other income moved in PubMatic's favor too. Interest income was $1.1 million, slightly lower than a year earlier, while other income was a gain of $0.1 million against a loss of $0.9 million, a swing the 10-Q attributes to foreign currency.

Adjusted EBITDA and the non-GAAP bridge

PubMatic defines adjusted EBITDA as net income adjusted for stock-based compensation, depreciation and amortization, litigation-related expenses, interest income and taxes. Starting from net income of $5.545 million, the reconciliation adds back $7.540 million of stock-based compensation, $8.428 million of depreciation and amortization, $1.155 million of litigation-related expenses and $4.379 million of tax, and deducts $1.091 million of interest income. The result is $25.956 million.

The litigation line is new this cycle. It was zero in the first two quarters of 2025, first appeared at $0.5 million in the third quarter of 2025 and reached $1.2 million in the third quarter of 2026, according to the quarterly reconciliation in the earnings presentation. The release describes it as "external legal fees and other expenses, net of insurance recoveries, associated with pending litigation that arose outside of the ordinary course of business," relating to "a discrete matter." The filings do not name the matter. The only proceeding described in the legal proceedings section of the 10-Q is the antitrust suit PubMatic filed against Google, which PubMatic filed on September 8, 2025, the same quarter in which the line first appeared.

Non-GAAP net income was $12.6 million, or $0.25 per diluted share, against $1.6 million, or $0.03, a year earlier. It adds back stock-based compensation and litigation expenses and deducts a tax adjustment of $1.6 million.

One small inconsistency runs through the documents. The release, the 10-Q and the fact sheet give third-quarter stock-based compensation of $7.540 million, adjusted EBITDA of $25.956 million and non-GAAP net income of $12.591 million. The quarterly reconciliation tables in the earnings presentation give $7.538 million, $25.954 million and $12.589 million. The differences are $2,000 on each line and change nothing in the rounded figures, but the presentation and the SEC filing do not match to the thousand.

PubMatic's own counting runs long. "This was our 42nd consecutive quarter of positive adjusted EBITDA," Pantelick said. On a GAAP basis, the third quarter was the company's first quarterly profit since the fourth quarter of 2025, when it earned $6.7 million. The two quarters in between produced net losses of $12.5 million and $1.2 million, according to the presentation's quarterly table.

Where the revenue came from

PubMatic reports revenue on a net basis. According to the 10-Q, revenue "represents gross billings to buyers, net of amounts we pay publishers and rebates associated with SPO agreements with buyers." The $79.6 million is the company's share, not the advertising spend that passed through its platform. That is also why the balance sheet carries accounts receivable of $381.0 million and accounts payable of $385.1 million, of which $352.3 million is owed to publishers: both are recorded at gross amounts, and both look large next to a quarterly revenue figure measured net.

The company does not break revenue into channels in its financial statements. It has one reporting segment. The channel figures come from the release, the prepared remarks, the fact sheet and the call, and almost all of them are rounded or approximate.

Three channels, two totals

The headline claim is that CTV, mobile app and "emerging" revenue represented approximately 64% of total revenue, up from approximately 60% in the second quarter. Emerging revenue, per a footnote to the release, "includes Activate, Commerce Media, Connect and AI solutions." The fact sheet lists the same group as AgenticOS, Activate, Commerce Media and Connect.

The individual shares are given as well:

  • CTV: revenue grew more than 30% year over year and represented approximately 21% of total revenue. In the Americas, CTV grew more than 40%.
  • Mobile app: revenue grew nearly 40% and represented approximately 27% of the total.
  • Emerging revenue: more than doubled and represented approximately 18%, "an all-time high", according to Pantelick.

Added together, 21%, 27% and 18% make 66%, not 64%. Rounding alone would not usually produce a two-point gap across three numbers. The more likely explanation, which the documents do not confirm, is that some revenue sits in more than one category: a CTV deal bought directly through Activate, for instance, would be both CTV and emerging revenue. PubMatic does not say whether the categories overlap, and none of the documents reconciles the three shares to the combined figure.

In dollars, the approximate shares translate into about $16.7 million of CTV revenue, about $21.5 million from mobile app and about $14.3 million of emerging revenue in the quarter, on PPC Land's arithmetic from the rounded percentages.

The fact sheet adds a broader grouping: "Combined revenue from mobile and omnichannel video represented 83% of total revenue in Q3 2026." It does not define omnichannel video.

The other third of the business

According to Pantelick, the high-value group "grew more than 50% year-over-year." That claim, combined with the 64% share, says something about the rest of the company that none of the documents states.

If the group made up about 64% of $79.6 million, it produced about $50.9 million in the third quarter. Growth of 50% would mean it produced about $34.0 million a year earlier, or half of the $68.0 million total. The remainder of the business, which includes desktop and mobile web display, would then have fallen from about $34.0 million to about $28.6 million, a decline of roughly 16%. Allowing for the rounding in "approximately 64%", the decline falls between about 15% and 16%. If the group grew by more than 50%, as Pantelick said, the decline in the remainder was larger. This is PPC Land's calculation from rounded company figures, not a number PubMatic published.

The calculation matters for publishers whose inventory sits mostly on the open web. PubMatic's growth is concentrated in logged-in apps and streaming television. "Importantly, much of this revenue is generated in logged-in mobile app and CTV environments, where audience signals are strongest," Pantelick said.

Display, as a format, still grew. "Total display revenues also grew strongly at 10% year-over-year, primarily driven by mobile app display," Pantelick said. Display in PubMatic's usage spans devices, so the 10% figure includes in-app banners and interstitials as well as web inventory.

A footnote at the bottom of the third slide of the earnings presentation reads: "CTV and Display revenues exclude legacy DSP." It is printed twice, in small dark type at the foot of the page, and is not attached to any figure on that slide, which shows only total revenue growth, net income margin, the 64% share and the adjusted EBITDA margin. Neither the release nor the 10-Q says whether the 30% CTV growth and 10% display growth quoted elsewhere exclude the large demand-side platform that changed how it bought from PubMatic in 2025.

Political money

Political advertising contributed approximately $5.6 million in the third quarter, against $1.1 million a year earlier, according to the 10-Q, which specifies that the spend was in the United States. Pantelick said the figure was "in line with our expectations." That is about 7% of quarterly revenue.

Without political spending in either year, third-quarter revenue would have grown about 10.6% instead of 17%, PPC Land's calculation shows ($74.0 million against $66.9 million). The prepared remarks and the presentation both list "revenue growth excluding political advertising" among the non-GAAP measures used on the call, but neither gives that figure for the third quarter.

For the fourth quarter, Pantelick assumed political spending at "roughly 7%-8% of total revenue." Applied to the $91 million to $93 million range, that is about $6.4 million to $7.4 million, on PPC Land's arithmetic. He also built in caution. "In prior election cycles, there is this period of, I would say, pause and pacing as advertisers reset the game plan for the fourth quarter seasonality, and we are sort of right in that area now," he said. "As I put together the outlook, based upon prior history, I decided that it is appropriate to be prudent in the outlook."

He compared this cycle with the last presidential election. "Obviously, election spending now has largely disappeared, but when I take a look at what we have been able to accomplish, for example, in the third quarter and the fourth quarter, that aggregate amount is very similar to what we were able to do in the presidential cycle," he said. He did not give the presidential-cycle figure.

Political money also helped Activate, PubMatic's direct-buying product. "It was certainly propelled in part by political, but it reflects the underlying value creation," Pantelick said of Activate's growth. His answer on CTV points the same way. Asked by Ethan Widell of B. Riley why CTV grew 40% in one region and 30% globally, Pantelick corrected the framing: "We said actually over 30% globally. What we're really seeing is across the globe, significant CTV expansion, and it was led certainly by the Americas, with the assist of political."

Magnite, PubMatic's closest listed competitor, reported its quarter five days earlier, with net income up 70% to $34.0 million and its full-year political expectation lifted above $40 million, against $15 million previously projected. The comparison shows the scale gap: Magnite's quarterly revenue of $204.2 million was more than two and a half times PubMatic's.

Asia Pacific grew 54%

The 10-Q splits revenue by the billing address of the publisher. That is the only revenue breakdown in the audited-format statements, and it tells a different story from the channel commentary.

Region Q3 2026 Q3 2025 Change Share of Q3 2026
United States $41.3m $37.6m +10% 52%
EMEA $24.9m $21.3m +17% 31%
APAC $11.8m $7.7m +54% 15%
Rest of world $1.5m $1.4m +8% 2%
Total $79.6m $68.0m +17% 100%

Changes and shares are PPC Land's calculations from the 10-Q figures.

Asia Pacific added $4.2 million of revenue, more than the United States added ($3.7 million), from a base one fifth the size. Over nine months, APAC revenue rose 50% to $29.0 million, EMEA rose 10% to $70.9 million, and the United States rose just 1% to $116.3 million. The United States represented 53% of nine-month revenue; no other single country exceeded 10%.

The US figure for the third quarter includes the political spend. If all of the $5.6 million of political revenue was billed through US publishers - the filing does not say so, but the spend was in the United States and the geographic split follows the publisher's billing address - then US revenue excluding political would have fallen about 2%, from $36.5 million to $35.7 million. This is a PPC Land calculation resting on that assumption.

The company's activity in the region shows up in the fact sheet, which lists RunnTV in India, a new office in Melbourne and increased investment in Hong Kong "to Accelerate Buy-side Business Growth." The largest agentic case study in the release, a luxury watch campaign run by the independent agency The Media Shop, spanned six Asia Pacific markets. The documents do not attribute the APAC growth to any specific customer.

For Europe, EMEA revenue of $24.9 million grew 17%, in line with the company as a whole, and held its share at 31%. The fourth-quarter adjusted EBITDA outlook "assumes a negative foreign currency exchange impact predominantly from Euro and Pound Sterling," according to the release. The 10-Q explains the exposure: buyer contracts are mostly in US dollars, while publisher contracts are in US dollars, euros, pounds and Australian dollars, and expenses are mostly in dollars, Indian rupees, pounds and euros. A hypothetical 10% move in the dollar against the pound would have changed the nine-month operating loss by $1.9 million; against the rupee, by $1.6 million. The fact sheet's case study list includes an agent-to-agent activation in Germany with the agency BCN. None of the documents gives revenue for any European country.

Verticals

Across PubMatic's top 10 advertising verticals, spending rose 17% year over year, according to the prepared remarks. "Growth was led by Shopping, Health & Fitness, Hobbies & Interests which in aggregate increased over 30% in the quarter," Pantelick said. "This helped offset softness in Business, and Food & Drink." The company did not give the size of any vertical.

The legacy DSP, five quarters on

The first half of 2026 carried a problem that started in 2025. PubMatic had warned in August 2025 that its third-quarter 2025 guidance of $61 million to $66 million included the impact of a major DSP partner. The company has never named the buyer in its results materials. In the first quarter of 2026, reported revenue fell 2%, while growth excluding the legacy DSP was 13%.

On the November 9 call, Jason Kreyer of Craig-Hallum asked about what he called "the DSP2 event," which he said the company had now anniversaried. Goel answered first. "The relationship is very strong and very healthy," he said. "That DSP continues to buy from us at significant volumes. So we feel very good about four quarters, I guess, five quarters later now, having worked through that change with the DSP."

Pantelick set expectations low. "I've shared that we anticipated that stability would be the prevailing theme, and that's exactly what's happened. Very stable quarter to quarter. The current outlook assumes continued stability. Just for framing it, we're not anticipating significant growth, but at the end of the day, it's an important legacy DSP, and we're appropriately investing behind it."

The 10-Q shows how much the business still depends on a handful of buyers. As of September 30, 2026, three buyers accounted for 29%, 14% and 14% of accounts receivable, a combined 57%. At the end of 2025, the three largest accounted for 24%, 16% and 14%, or 54%. No publisher represented more than 10% of revenue. The filing repeats the standing risk: "We depend upon a limited number of large DSPs for a large percentage of impressions purchased and our business results, including revenues, may be impacted by changes in their pricing strategies, bidding algorithms or go-to market efforts."

Buyers moving closer to the supply

Much of what PubMatic told investors concerned buyers, not publishers. The company built its business as a supply-side platform, the software publishers use to auction their inventory to demand-side platforms. A growing part of its pitch is that buyers should come to it directly.

Activate grew more than eight times

Activate is PubMatic's product for buyers to execute deals on its platform without a separate DSP in the middle. According to the 10-Q, it "allows buyers to execute direct deals on our platform across our publisher inventory." Buying activity on Activate grew more than eight times over the third quarter of 2025, the release says, "as buyers and publishers prioritized performance, control and transparency."

Goel called it an "exciting inflection point." Pantelick listed the sources: "In addition to political advertising, we're seeing significant growth on Activate from commerce buyers, independent agencies and holding companies."

The measure is "buying activity," not revenue. PubMatic does not disclose Activate revenue or the base from which it grew eightfold, and the presentation's disclaimer notes that operating metrics such as "buyer activity, campaign counts and impression data, are based on internal data, are not audited, and may be calculated differently from similarly titled metrics used by others."

Asked by Rob Coolbrith of Evercore where Activate had grown fastest, Goel pointed to independent agencies. "They are much quicker to make decisions. They are also, I think, looking to level the playing field from a technology perspective relative to the large holding companies," he said. "I think it is important to keep in mind or keep in context that over the last 10 years, the share of total budgets that the large agency holdcos manage has reduced significantly. The independent agencies, and actually direct to brand, are gaining." He did not give a figure for that shift.

Mid-market DSPs up 40%

Spend through what PubMatic calls mid-market DSPs grew more than 40% year over year. Pantelick later defined the group on the call as "the DSPs 11 to 20," meaning the buyers ranked eleventh to twentieth by spend on PubMatic. "Those don't go away as we turn the calendar," he added.

The company integrates new demand-side platforms at a steady rate and says it has automated the work. "We released an AI agent that onboards and integrates new DSPs, right? Last year, we integrated, I think it was 50 DSPs, and this year it will be a little bit lower than that, but a similar number," Goel said. "With this AI agent, we have been able to go from two to three calendar weeks to integrate a DSP to one to two days. Actually, the DSP can do all of the work on their side."

Supply path optimization at 55%

Supply path optimization, the practice by which buyers consolidate spending on fewer, direct routes to publisher inventory, represented approximately 55% of total activity on the platform in the third quarter. The 10-Q says PubMatic has invested in SPO technology and partnerships "for six years" and calls it "a major growth driver." SPO agreements carry a cost: PubMatic pays rebates to buyers under them, which are netted against revenue.

Sell-side targeting is the related figure. "Nearly half of the activity on our platform now uses it for sell-side targeting through curated deals, publisher auction packages and data partnerships," Goel said of PubMatic's data. The presentation phrases it as "proprietary intelligence used in nearly half of activity on PubMatic."

Publishers and retention

On the supply side, PubMatic served approximately 2,040 publishers and app developers on September 30, 2026, against approximately 1,980 a year earlier, according to the 10-Q, which counts groups of related accounts as a single "master" publisher. Net dollar-based retention, which compares revenue from the same publishers in consecutive twelve-month periods and excludes new publishers, was 103% for the twelve months to September 30, 2026, against 98% a year earlier. A rate above 100% means existing publishers, on balance, generated more revenue than the year before.

The case for sell-side decisioning

Goel spent most of his prepared remarks on an argument about where advertising decisions will be made. PubMatic's position is that buyers will increasingly decide which impressions to bid on inside the exchange, where the inventory and the data sit, rather than inside their own buying platforms. The release puts it as a shift "from sell-side targeting to sell-side decisioning, enabling buyers to act on richer intelligence closer to the inventory and first party insights."

Three forces

Goel named three forces behind that shift. The first is volume. "The scale of audience and inventory is significantly higher in the SSP due to traffic shaping - the process by which DSPs limit the impression volume passed to them by SSPs. This inhibits advertiser reach and performance," he said. Traffic shaping exists because no buyer can evaluate every bid request an exchange receives; the filtering decides, before any bid is considered, which impressions a DSP ever sees.

The second is data. "As cross-site identity diminishes and privacy regulations increase, first-party publisher data, contextual signals and direct publisher relationships are becoming far more valuable to advertisers," he said.

The third is automation. "Over the past two decades, buying media on the open internet has grown increasingly fragmented and complex, with layers of technology between advertisers and publishers. AI agents are cutting through that complexity, allowing buyers and publishers to transact directly."

The company describes what it brings to those decisions in broad terms. "It combines publishers' first-party data, contextual and intent signals, proprietary bidstream and auction data that exists only on PubMatic, along with direct relationships with thousands of publishers," Goel said. The 10-Q supplies the scale. PubMatic coordinates "over a hundred billion real-time auctions and nearly a trillion bids globally on a daily basis," each auctioned impression consists of 1,005 independent data parameters, and the processing "must occur in less than half a second." In September 2026, the platform processed approximately 888 billion ad impressions a day.

Two routes into the auction

PubMatic offers two ways for buyers to make decisions on its infrastructure. "This happens in two ways - containerized bidding using third party DSP models and direct buying in our SSP using PubMatic models in Activate," Goel said. "In both scenarios, advertisers get access to all of our inventory and intelligence - before traffic shaping limits what buyers can see. And in both scenarios, the decision happens on PubMatic, powered by our sell side intelligence."

The containerized route runs through Decision Fabric, which also hosts PubMatic's own models and which PubMatic launched on June 1, 2026 to let DSPs, curators and algorithm companies run their own models inside its auction. PubMatic said at the time that partner models respond in under 10 milliseconds and that it processes 2.7 trillion bid requests a day, with inPowered as the first live partner. The approach is not unique to PubMatic: in April, Index Exchange hosted the first DSP to run its bidder inside an exchange, with Bedrock's container execution window under five milliseconds.

According to the release, Decision Fabric "applies real-time signals to decisions across the impression journey, including audience qualification, traffic shaping and optimization." Partners "can deploy their proprietary models within PubMatic's infrastructure, enabling buyers to access sell-side intelligence through their existing workflows."

Goel listed what a partner gains by running inside PubMatic's systems. "Running within the flow of our inventory and intelligence, partners can also see supply before traffic shaping, gain more time within each auction to run sophisticated performance models and eliminate their own infrastructure costs." He added a point about PubMatic's own advantage: "Whether decisioning is made with a partner's model or via our own, our models learn in real time, making our proprietary intelligence smarter with every transaction."

That second sentence is worth noting for any DSP considering the arrangement. A partner whose model runs inside PubMatic's infrastructure makes decisions that PubMatic's own models learn from. The documents do not describe what data, if any, flows back to the partner, or what PubMatic retains.

Adoption, by Goel's account, is uneven. "Here, I would say we are even earlier than we are with Activate," he said of Decision Fabric. "What we are finding is that the smaller independent companies are not only taking advantage of it, but we have some very large partners, very large DSPs that are also quite interested in things like containerized bidding, because they see the performance advantage from an inventory and from a user access perspective." No large DSP was named, and Goel described their position as interest rather than use.

What the case studies show

PubMatic published five campaign results in its materials to support the argument. None names the advertiser except where an agency did, and the presentation's disclaimer states that "case study results reflect the specific campaigns described and are not necessarily indicative of future results."

  • A major online retailer, via a leading agency holding company, through Activate. "PubMatic unlocked new audiences and inventory that the retailer's existing buying platform couldn't access," Goel said. The campaign "delivered 10 million additional impressions to ideal audiences at a 32% more efficient CPM - and the client's brand awareness and brand favorability increased by nearly 7%." The presentation dates the case study to October 2026. Neither the retailer, the holding company, the baseline nor the method for measuring awareness is disclosed.
  • A national quick-service restaurant brand, using a PubMatic model on Decision Fabric through its own DSP. Average daily reach of its target audience rose 50% "while staying within its daily frequency caps," which the release says led to a 76% increase in spend delivery.
  • A national insurance provider, on Decision Fabric. Average daily reach rose nearly 14%, leading to a 41% increase in spend delivery, according to the release.
  • Goodway Group, for a regional grocery client, on AgenticOS. The campaign reached more than 1 million unique users "verified against the retailer's own first-party data," delivered 2 million impressions above plan and "put 20% more of the budget into working media." The campaign was built from a single prompt in PubMatic Assistant, and "the deal was created 75 seconds later."
  • The Media Shop, for a luxury watch brand, on AgenticOS. Campaign setup time fell by approximately 80% across six Asia Pacific markets, with 9% more impressions than planned and viewability up 17%.

Two more appeared in the Q&A. Goel cited the agency Klever, whose client Rouge Care "saw 5x return on ad spend, which was double their target of 2.5x," and Level Agency, where AgenticOS "performed better against their incumbent buying platform, incumbent DSP, with 53% more efficient CPMs and 2x more impressions per dollar."

Two of the Decision Fabric results measure spend delivery, which is the share of a budget a campaign manages to spend. For an exchange paid on transacted volume, a campaign that spends 76% more of its budget is also a campaign that generates more revenue for the exchange. The case studies report the advertiser outcomes on reach and delivery; only the retailer example reports a brand metric, and only Klever reports a return on spend.

AgenticOS after ten months

AgenticOS is PubMatic's system for AI agents that plan, negotiate, set up, report on and optimize campaigns. PubMatic launched it on January 5, 2026 with live campaigns running, when the company cited early-testing claims of setup time down 87% and named WPP Media, Butler/Till and MiQ among early participants. In August, PubMatic added a five-step governance layer to block agents from unapproved inventory, reporting more than 80 autonomous campaigns and more than 4,000 direct deals at that point.

The counts

The release updates both numbers. "Since launching in January 2026, PubMatic has delivered measurable performance gains for customers across more than 120 fully autonomous, end-to-end agentic campaigns globally," it says. "This is up from over 80 campaigns a quarter ago, and includes all five global agency holding companies." The number of "AI-powered deals created on PubMatic to date" passed 14,000, up from over 4,000 a quarter earlier.

The two numbers describe different things. The 120 campaigns are buyer-side campaigns run end to end by agents. The 14,000 deals, according to Goel, were created by publishers: "To date, publishers have created 14,000 AI-powered deals on our platform which automates the manual work of packaging and selling each audience they want to monetize." A deal created is not necessarily a deal that transacts, and none of the documents says how many of the 14,000 deals carried spend, or how much.

The fact sheet adds that "AgenticOS includes 20+ AI agents and workflows available for media buyers and publishers" and that "agentic campaigns on PubMatic deliver over 80% time savings in campaign setup." It lists among the quarter's proof points an agent-to-agent activation in Germany with BCN and what PubMatic calls Southeast Asia's first agentic luxury campaign, with The Media Shop.

Set against a platform with 2,040 publishers and buyers ranging from the largest holding companies to dozens of DSPs, 120 campaigns over ten months remains a small base. PubMatic did not say what share of revenue AgenticOS campaigns represent; it groups AI solutions inside emerging revenue. "Emerging revenue streams more than doubled year-over-year and reached an all-time high of approximately 18% of total revenue, driven by increased adoption of our new AI products, including AgenticOS," Pantelick said.

The 10-Q, the document filed with the SEC under the certification of the chief executive and the chief financial officer, does not mention AgenticOS, agentic campaigns or OpenAI at all. Its description of value-added products names Connect, Activate and OpenWrap.

The Campaign Insights Agent

The release introduces one new agent. PubMatic's Campaign Insights Agent "extends AI-powered intelligence into campaign management, analyzing campaigns daily to flag pacing, delivery, targeting and performance issues." Traders can "investigate findings through natural-language queries and evaluate recommended actions, connecting campaign intelligence with buying decisions while retaining control over execution." The release does not give a launch date, pricing or availability by market.

How much a human still does

Dawson Frisinger of Lake Street Capital Markets asked how much of a campaign a human still handles. Goel answered that it depends on the buyer. "The way that we think about the agentic opportunity is that eventually there will be AI agents that are running autonomously across all aspects of the media planning and buying process. That's discovery, planning, activation, measurement, reporting," he said.

He described a range of clients. Some ask an agent to create a plan, review it, then review again before approving execution. "Other folks, while they might start there, pretty quickly they get to, 'Okay, great. I can see the AI is doing a terrific job.' Actually, the goal is that AI should make less errors than humans. The AI doesn't need to be perfect. Of course, there's plenty of volume of human errors in media trading operations globally." At the other end, he said, some clients "let the AI go everywhere from planning to activation to measurement, and they might have their own AI agent do a periodic check to make sure that our AI platform is executing correctly."

Half of buying by 2030

Barton Crockett of Rosenblatt Securities asked about what he called Goel's "very bold predictions of 25% agentic in 2028 and maybe 50% by 2030." Goel restated only the second. "So I had said that by 2030, I expect 50% of buying in our ecosystem would be agentic," he said. He did not repeat the 2028 figure.

He identified the obstacle as organizational, not technical. "I would say right now, it's more of a change management gating the pace of change across the ecosystem. We're seeing rapid uptake with independent agencies. I think the agency holdcos are still a little bit getting their arms around what does agentic buying mean for them and how do agents plug into their technology stacks." He expects that to change. "I think 2027 is set to be a pretty exciting year from the perspective of agentic execution."

The holding company point sits awkwardly next to the release's statement that the campaign count "includes all five global agency holding companies." Both can be true: each holding company may have run at least one agentic campaign while none has moved a meaningful share of its buying.

Advertising to agents

Frisinger's second question concerned agentic shopping. Goel said the industry is still working out whether advertisers will need to reach AI agents rather than people. "I think that's a conversation that's happening in our industry, in our ecosystem, and we're talking actually to DSPs to find out, how are you thinking about or how are you seeing your clients think about advertising to AI agents? If that becomes a meaningful market, then I think we have a key role to play in terms of connecting the consumer signal all the way through to the DSP."

He also raised the question of who makes the purchase decision when software does the shopping. "If it's no longer a consumer, we could see power aggregated behind a variety of different AI agents. It could be AI agents for commerce that own the shopping destination, like, let's say, an Amazon agent, but it could be an entirely third-party agent that doesn't own commerce and then is going to find the right opportunity on behalf of the client."

OpenAI and ChatGPT ads

The most widely reported item in PubMatic's quarter was not in its financial statements. "Excitingly, we recently partnered with OpenAI, expanding into a new growth category," Goel said in his prepared remarks. "ChatGPT Ads reach consumers with strong signals and clear intent, and we expect it to become a regular part of advertisers' media plans. It builds on our leadership in agentic advertising, and as the category grows, there's meaningful opportunity for us to grow with it."

PubMatic's own materials do not describe the terms. The fact sheet lists the partnership under "customer momentum" with the headline "OpenAI Signs Deals with Magnite, PubMatic to Fill Unused ChatGPT Ad Space" and links to a report by The Information, not to a PubMatic or OpenAI announcement. The release does not mention OpenAI, and the 10-Q does not either. No start date, revenue share, inventory volume or market list appears in any of the documents.

Shweta Khajuria of Wolfe Research asked about "the OpenAI-type partnerships. Even if they are not contributing, what does it mean?" Goel did not say whether the partnership contributed revenue in the third quarter. "I think some buyers are buying that out of search budgets today, and others are buying them out of display budgets. It has a high degree of relevancy for us," he said. "As the category grows and more AI platforms build advertising businesses, I think there's meaningful opportunity for us to grow with it."

The category is growing quickly. In September, OpenAI's David Dugan said ChatGPT ads had passed a $1 billion annualized run rate, with ads listed in 63 countries by September 24, up from a single US market in early April. For PubMatic, the opportunity is an exchange role selling inventory that OpenAI does not fill directly. How much of that inventory exists, and at what price it clears, is not stated.

CTV, live sports and the PlayStation

Connected TV grew more than 30% globally and more than 40% in the Americas, and reached approximately 21% of revenue. PubMatic says it partners "with 28 out of the top 30 global streaming companies," according to the fact sheet, which does not list them.

The main CTV announcement of the period came from Sony. "Sony Pictures Entertainment named PubMatic its preferred programmatic advertising partner for LIVE TV on PS5, connecting premium inventory and authenticated audience insights with brands and agencies worldwide," the release says. Goel described the inventory as "premium video to logged-in viewers inside the console experience." The release does not give the number of channels, markets or users, or say whether the arrangement is exclusive.

Live sports streaming is the other strand. PubMatic launched a live sports marketplace in July 2025, citing at the time a more than threefold increase in live sports activity in the first half of 2025. The third-quarter release says the marketplace now carries "verified, event-level content intelligence, enabling advertisers to programmatically target specific sports, leagues, teams and individual games as they air across available streaming inventory." Goel added that buyers can target "by sport, league, team, game window and geography, down to the ZIP code where inventory permits."

Strategus, whose "CTV expertise" the release cites, named PubMatic its preferred live sports programmatic partner. Goel gave one example: "During the FIFA World Cup, Strategus, MayoSeitz Media and PubMatic targeted Citadel Credit Union's priority markets - bringing a global event to a local Pennsylvania audience." No performance figures were given for that campaign.

The fact sheet lists further CTV items without detail: RunnTV in India, Smadex extending mobile buying to CTV "Scaling Monthly Spend by 80x," and a partner perspective from Philo on pause ads.

Mobile app and the mediation layer

Mobile app revenue grew nearly 40% to approximately 27% of the total, the largest of the three high-value channels. "The biggest drivers were our relationships with the top mediation platforms and continued expansion of our global app publisher base, supported by ongoing product innovation and double-digit growth in mobile app display," Pantelick said.

Goel explained the reliance on mediation platforms, the software layer app developers use to choose which ad network or exchange fills each slot. "We have our own software development kit, or SDK, for about 10 years now. We access over 90% of inventory available via SDKs through our partnerships with companies like AppLovin, Google AdMob, Unity, and so forth, that have the leading mediation platforms," he said in answer to Jason Kreyer.

He also tried to widen the image of the channel. Advertisers, Goel said, "are understanding that it's not just teenagers that are playing games. There's actually a very broad set of mobile app usage." "Users tend to be logged in many of the environments that we're driving monetization within. There's a lot of rich data available around context and around attention that drives better targeting and better performance for advertisers." PubMatic is "today attracting brand dollars, but going deeper into attracting performance advertising budgets," he said.

Pantelick connected the growth to a 2025 decision. "One of the areas that we told the investing public that we would be investing more aggressively last year was to bring on new DSP buyers who specialize in performance advertising, and mobile app is an ideal channel for them."

The dependence runs both ways. Barton Crockett asked whether AppLovin, which has signalled ambitions in CTV, would become a competitor. "AppLovin's a terrific partner for us in the mobile app space, significant and fast-growing part of our business, and we tap into their mediation layer to access inventory with our SDK," Goel said. "Obviously, they own Wurl, so they have some CTV assets." He described a convergence instead of a contest: "There's a convergence of performance buying between mobile app and CTV. In fact, this was one of my key takeaways when I was in Cannes, where we saw that there's a number of buyers that are in mobile app and are looking to buy in CTV, and vice versa." He did not say whether PubMatic regards AppLovin as a competitor.

The fact sheet's mobile examples include TimeTree, for which PubMatic's OpenWrap SDK "delivers a 17% CPM uplift," and Opera Ads. OpenWrap is PubMatic's header bidding product, through which publishers let several exchanges compete for an impression at once.

Safari and iOS 27

Rob Coolbrith asked about "this WebKit in iOS 27," a reference to Apple's browser changes. Apple's iOS 27 began blocking The Trade Desk's ad-serving domains in Safari after the operating system shipped on September 14, with ID5, Audigent, LiveRamp and Permutive also on the list.

Goel's answer was short on exposure. "Safari is a small part of our business. It is a low single-digit percentage of revenue, so our direct exposure is limited," he said. "Structurally, two-thirds of our revenue already comes from CTV, mobile app, and emerging revenue like commerce media and AI solutions, which do not depend on browser identifiers. Changes like this, I think, tend to reinforce the shift towards the areas where we are already quite strong." He did not say whether PubMatic's own domains are affected. The forward-looking statements in the release list "the lack of similar or better alternatives to the use of third-party cookies, mobile device IDs or other tracking technologies if such uses are restricted" among the risks to the business.

Fewer impressions, more revenue

PubMatic owns and operates its servers rather than renting public cloud capacity, and the quarter's clearest operating change came from that hardware. Monetized impressions rose 15% year over year while gross impressions processed fell 8%, according to the release. "Last quarter, I discussed reallocating compute toward the products and impressions that create the most value for our customers and for PubMatic. In Q3 we delivered on that," Pantelick said. "Because we own and operate our infrastructure, we have control to make these shifts quickly, unlocking incremental revenue while also generating cost savings."

In plain terms, the platform handled fewer ad opportunities and sold more of them. The 10-Q describes the selection process: PubMatic adds or removes impressions "based on an assessment of the projected value of the impressions, which is influenced by the type of publisher and its related consumers, as well as the potential volume of monetizable impressions and ad format types, such as digital video." Publishers whose inventory is judged less valuable receive less of PubMatic's processing capacity.

The same logic runs through PubMatic's supply policy, which charges publishers for impressions sent above their limits; PubMatic cut that excess inventory fee by 80% in April. The 10-Q does not mention the fees or quantify any revenue from them.

Cost per impression: lower over twelve months, higher in the quarter

The release states that "cost of revenue per million impressions processed decreased 15% on a trailing twelve month period, as compared to the prior period." The 10-Q, discussing the third quarter alone, says the opposite moved: "Overall, our cost of revenue per million impressions processed for the three months ended September 30, 2026 increased by approximately 2% compared to the three months ended September 30, 2025."

Both statements can be correct at once. Over twelve months, PubMatic processed more impressions at lower cost per unit. In the third quarter, it deliberately processed 8% fewer impressions while cost of revenue fell only 6%, so the cost of each million processed rose slightly. PPC Land's arithmetic from those two percentages gives an increase of about 2.6%, consistent with the filing's "approximately 2%." The release cites only the twelve-month figure.

The metric that improved in the quarter is the one PubMatic now emphasizes: revenue relative to capacity. "A metric that I had shared earlier in the year that we are going to keep an eye on and report back to you is, how are we doing in terms of growing our monetized impressions relative to our overall capacity?" Pantelick said in answer to Justin Patterson of KeyBanc Capital Markets.

A move to GPUs and a higher capex range

PubMatic raised its capital spending outlook for 2026 to $24 million to $28 million, "up several million dollars from our prior range," Pantelick said. The documents do not state the prior range. The money goes into what he called "GPU-centric infrastructure," which "underpins Activate, AgenticOS, sell-side intelligence and containerization."

The quarter's cash flow statement shows the spending starting. Purchases of property and equipment were $9.9 million in the third quarter, more than double the $4.7 million of a year earlier, and $12.9 million over nine months. Capitalized software development costs were $3.3 million in the quarter and $13.5 million over nine months. The documents do not define which lines the capex outlook covers. If it refers to purchases of property and equipment alone, the fourth quarter would require $11.1 million to $15.1 million, more than the third quarter's $9.9 million, PPC Land's calculation shows.

Pantelick described the investment test. "Our perspective is we are going to continue to put investment behind those because that is where we see all the growth and our ROI expectations is to deliver 12-month return paybacks or lower," he said. "I think the key position that we have shifted over the last couple of quarters is to focus less on pure capacity expansion and more on high-value GPU-centric infrastructure to deliver against these capabilities."

The longer server life and the GPU shift interact. PubMatic now depreciates network hardware over five years. New GPU equipment bought in 2026 will therefore carry a lower annual depreciation charge than it would have under the old three-year assumption, which supports reported gross margin while the equipment ages.

Fewer staff, more sellers, a new CRO and a departing CFO

PubMatic's plan for 2027 is to move money from operations into sales. "One of our biggest opportunities is to invest in our buyer-focused sales team to bring more buyers onto our platform. Importantly, we're funding that investment through productivity gains and efficiencies across our business primarily driven by AI adoption," Goel said.

The presentation gives the evidence so far: a sales team 5% larger than a year earlier, total headcount lower. The fourth quarter brings more hiring. "In Q4 we're ramping up hiring, primarily in our buyer-focused sales team, to expand across high-value channels and formats and go deeper with existing advertisers and publishers. We expect these investments to drive growth in 2027 and beyond," Pantelick said.

Megan Ramm joined in August as global chief revenue officer. Goel said she brings "deep experience scaling revenue organizations at Uber, Snap and Google." Her offer letter, dated June 24, 2026, was filed as an exhibit to a Form 8-K on August 12, 2026, according to the exhibit index of the 10-Q. Goel described her brief: "to put more people in market to cover agency holdcos, to more aggressively go after independent agencies, and also to build relationships with brands alongside the DSP relationship building that we've been doing."

The other leadership change was announced in August. Pantelick, chief financial officer for fifteen years, plans to retire; he will remain chief financial officer into the first quarter of 2027 and then serve as a senior adviser through July 1, 2027. The November 9 call was therefore one of his last as CFO. The documents for the quarter do not name a successor. The Form 8-K and the 10-Q were both signed by Pantelick.

Cash, buybacks and the balance sheet

PubMatic ended September with $120.3 million in cash and cash equivalents and $24.1 million in marketable securities, a total of $144.5 million, and no debt. At the end of 2025, it held $145.5 million, all in cash. The total barely changed over nine months in which the company spent $45.5 million on its own shares.

Operating cash flow was $30.8 million in the third quarter, 5% below the $32.4 million of a year earlier. Free cash flow, which PubMatic calculates as operating cash flow minus purchases of property and equipment and capitalized software, was $17.6 million, down 23% from $22.8 million, mainly because equipment purchases doubled. The free cash flow margin was 22% against 34%. Over nine months, free cash flow rose to $42.0 million from $39.4 million, a margin of 19% in both years.

Working capital explains much of the nine-month operating cash flow of $68.3 million. Accounts payable rose $40.6 million and accounts receivable rose $22.8 million. A year earlier, the same lines had moved the other way, with receivables down $62.2 million and payables down $30.0 million. Because PubMatic collects from buyers and pays publishers at gross amounts, the timing of those two flows can move quarterly cash generation by tens of millions of dollars. The 10-Q notes that advertisers "have historically spent relatively more in the fourth quarter."

"Since the beginning of 2021 through Q3 2026, we have generated nearly $480 million in net cash from operations and more than $264 million in free cash flow," Pantelick said.

$198 million for buybacks

On November 9, the board authorized an additional $150 million of repurchases and extended the program from December 31, 2026 to December 31, 2028. With $48.4 million left under the previous authorization at the end of September, the program now has approximately $198.4 million available, according to the 10-Q.

The program's history, set out in the 10-Q, shows how it has grown. The board authorized $75 million in February 2023, added $100 million in February 2024, another $100 million in May 2025, and now $150 million, for a total of $425 million. Through September 30, 2026, PubMatic had spent $226.6 million to buy back 16.5 million Class A shares, an average of about $13.73 a share.

During the third quarter, the company bought 969,828 shares for about $15.2 million. The monthly detail in the 10-Q shows rising prices:

Month Shares bought Average price
July 2026 291,030 $13.19
August 2026 428,698 $16.45
September 2026 250,100 $17.10

Over the first nine months of 2026, PubMatic bought 4,088,651 shares for $45.5 million, an average of about $11.13 a share, PPC Land's calculation shows. Earlier in the year, the share price was far lower: restricted stock units granted in the first nine months carried a weighted-average grant date fair value of $7.45, and options were granted at a weighted-average exercise price of $6.29, according to the equity tables in the 10-Q.

As of November 2, 2026, PubMatic had 37,630,887 Class A shares and 8,207,864 Class B shares outstanding, 45.8 million in total. At the $19.03 closing price on November 9 reported by Investing.com, that implies a market value of about $872 million, PPC Land's calculation shows. The $198.4 million authorization equals about 23% of that value. The release notes that the program "does not obligate us to purchase any amount of Class A common stock." The 10-Q adds a new risk factor stating that the company "cannot guarantee that our share repurchase program will be fully consummated or that it will enhance long-term stockholder value."

The repurchases have shrunk the equity base. Stockholders' equity fell to $241.8 million at the end of September from $262.6 million at the end of 2025, as treasury stock rose to $239.2 million at cost.

Other uses of cash

PubMatic made two small investments in private companies this year. The cash flow statement shows a $3.5 million purchase of non-marketable equity investments in 2026, carried at Level 3 fair value; the 10-Q does not name the company. After the quarter closed, on October 28, 2026, PubMatic entered into a Simple Agreement for Future Equity with Apostra Inc., a Delaware company, making an initial investment of $2.0 million for the right to acquire shares in a future financing. The 10-Q does not describe Apostra's business. The presentation lists "potential M&A" among uses of cash.

PubMatic's $110 million revolving credit facility, originally arranged with Silicon Valley Bank and now held by First Citizens Bank, remains undrawn and matures on October 17, 2027. The applicable interest rate on September 30 was 8.00%. The 10-Q also records $8.3 million of long-term income tax liabilities related to uncertain tax positions.

Guidance: two halves and a cautious fourth quarter

PubMatic's outlook for the fourth quarter is revenue of $91 million to $93 million and adjusted EBITDA of $31 million to $33 million. For 2026, it expects revenue of $312 million to $314 million and adjusted EBITDA of $79 million to $81 million. The presentation adds margins: 34% to 35% for the fourth quarter and 25% to 26% for the year.

Measure Q4 2026 outlook Q4 2025 actual FY 2026 outlook FY 2025 actual
Revenue $91m-$93m $80.0m $312m-$314m $282.9m
Growth 14%-16% - 10%-11% -
Adjusted EBITDA $31m-$33m $27.8m $79m-$81m $61.6m
Adjusted EBITDA margin 34%-35% 35% 25%-26% 22%

Fourth-quarter 2025 and full-year 2025 figures are from the presentation's quarterly reconciliation; the full-year 2025 totals and the 22% margin are PPC Land's sums of the four quarters.

Two observations follow from the table. First, the full-year outlook is consistent with the fourth-quarter outlook: nine-month revenue of $220.7 million plus $91 million to $93 million gives $311.7 million to $313.7 million, and nine-month adjusted EBITDA of $48.2 million plus $31 million to $33 million gives $79.2 million to $81.2 million.

Second, the fourth-quarter margin does not expand. Revenue is expected to grow about 15%, yet the adjusted EBITDA margin of 34% to 35% is roughly the same as the 35% of the fourth quarter of 2025, and well below the 44% of the fourth quarter of 2024, when revenue was $85.5 million. Three factors in the documents explain the flat margin: the hiring that was deferred from the third quarter into the fourth, the negative currency effect from the euro and the pound assumed in the outlook, and Pantelick's decision to be "prudent" about post-election spending. The release frames the year in terms of "adjusted EBITDA margin expansion", which holds for 2026 as a whole (25% to 26% against about 22%), not for the fourth quarter on its own.

The company also said October started well. "The strong momentum we built throughout the third quarter continued into October as revenue grew double digits year-over-year led by monetized impression growth," Pantelick said. No October figure was given.

Pantelick declined to discuss 2027 in detail. "As you know from prior years, we'll come back in February and talk about 2027," he said, before listing what he called the trends already visible: the mix shift toward high-value channels, Activate and AgenticOS, commerce media partners, mid-tier DSPs and OpenAI. "Finally, underpinning all this momentum that we have is the structural shift," he said. "Targeting clearly is moving to sell-side, as is decisioning."

The outlook, the release says, "assumes that general market conditions do not significantly deteriorate as it relates to current macroeconomic and geopolitical conditions." The 10-Q lists the conditions in question, including "the ongoing conflict in the Middle East and resulting disruption to international trade and energy markets," tariffs, interest rates and persistent inflation, and adds: "To date, we have not observed material impacts in our business or outlook."

The Google remedies

Pantelick closed his prepared remarks with a line on the US government's ad tech case against Google. "Looking ahead, we're also preparing for the potential benefits of the remedies in the Google ad tech case as they emerge."

The remedies are recent. On September 2, Judge Leonie Brinkema rejected the Justice Department's request to force a sale of Google's AdX exchange and accepted behavioral remedies instead. The unsealed opinion set out a six-year worldwide decree with an independent monitor, including real-time AdX bids through Prebid. A count by the consultant Scott Messer found that all 19 conduct provisions Google itself had proposed appear in the ruling, 16 as written and three modified, with six more added by the court. Goel had already given his view of the timing: in September, he said Google needs up to 15 months to open AdX to Prebid, citing a Google-side engineering estimate.

On the call, Shweta Khajuria asked how much PubMatic expects to benefit in 2027. Pantelick answered that the fourth-quarter outlook assumes nothing from the remedies. "Absolutely, it's going to be a positive for us and others. As a reminder to everybody, there isn't this huge amount of incremental investment that's required to take advantage of it. What's required is the implementation of those remedies, and it's clear there's some big upsides, as has been shared previously, just the advent of incremental brand new demand on our platform. It will be a net positive. We're just waiting for all the pieces to fall into place."

PubMatic is also a plaintiff. Its own suit in the Eastern District of Virginia, which seeks injunctive relief and damages for what it calls monopolistic conduct in the publisher ad server and ad exchange markets, remains in its early stages, according to the 10-Q: "Given the nature of the case, including that the proceedings are in their early stages, we are unable to predict the ultimate outcome of the case." The fact sheet links to a PubMatic blog post titled "Understanding Google's Ad Tech Remedies: What the Market Should Know."

How the shares reacted

PubMatic's shares closed the regular session on November 9 at $19.03, up 6.85% from $17.81, according to Investing.com, which put the 52-week range at $6.15 to $19.48 and said the stock had risen 157% over the previous year. In after-hours trading following the results, the shares fell 3.93% to $18.28. On November 11, Investing.com's quote panel for PubMatic showed the shares at $17.90, up $0.94, or 5.54%, on the day, which implies a previous close of $16.96 - below the results-day close.

The share price history matters for the buyback. The company bought stock at an average of $13.19 in July and $17.10 in September. The new $150 million authorization arrives with the shares close to the top of their 52-week range.

What the documents do not say

Several numbers that would help assess the quarter are missing from all eight sources:

  • Revenue by channel in dollars. CTV, mobile app and emerging revenue are given only as rounded percentages, and the three published shares sum to 66% against a stated combined 64%.
  • Activate and AgenticOS revenue. Activate's growth is measured in "buying activity"; AgenticOS is measured in campaigns and deals created.
  • Third-quarter growth excluding political spending. The non-GAAP measure is listed in the safe harbor statement, but no figure for the quarter is given.
  • The OpenAI terms. No start date, revenue model or inventory volume, and no mention in the release or the 10-Q.
  • The identity of the legacy DSP and its current share of revenue.
  • The prior capex range and which cash flow lines the new $24 million to $28 million covers.
  • Headcount. The presentation says total headcount fell and the sales team grew 5%, without numbers.
  • The advertisers in the case studies, apart from the agencies Goodway Group, The Media Shop, Klever and Level Agency and the Klever client Rouge Care.

The presentation itself warns that its operating metrics "are based on internal data, are not audited, and may be calculated differently from similarly titled metrics used by others." That covers the 8x Activate figure, the 40% mid-market DSP figure, the 55% SPO share, the 15% monetized impression growth and the campaign counts.

The transcript quoted in this article was published by Investing.com, which states that its page "was generated with the support of AI and reviewed by an editor." Quotations from the prepared remarks match PubMatic's own published text.

Why this matters for marketers

For buyers working outside Google, Meta and Amazon, PubMatic's quarter offers a data point on a question that has dominated programmatic advertising this year: will decisions about which impressions to buy move from the DSP to the exchange? PubMatic says they are moving. The evidence it offers is a mix of audited figures (revenue, margins, cash) and unaudited activity metrics (Activate buying activity, campaign counts, the share of activity using its data). The audited figures show a business growing again, mostly in apps and streaming television, with costs falling. The unaudited figures show growth in the products that would carry the sell-side argument, from bases that are not disclosed.

For publishers, the infrastructure numbers carry a direct message. PubMatic processed 8% fewer impressions in the third quarter and monetized 15% more. The company decides which inventory gets its capacity on the basis of projected value, and its growth is concentrated in logged-in app and CTV environments. On PPC Land's arithmetic from the company's own rounded figures, the part of PubMatic's revenue outside CTV, mobile app and emerging products shrank by roughly 15% or more year over year. Open-web publishers selling display inventory through PubMatic are working with a partner whose investment, and whose processing capacity, is being directed mainly elsewhere.

For agencies, the call contained an unusually direct statement from an exchange chief executive about who is adopting agentic buying. Independent agencies are moving first, Goel said, while the holding companies "are still a little bit getting their arms around" what it means. That matches the pattern of PubMatic's own case studies, which name agencies such as Goodway Group, The Media Shop, Klever and Level Agency - the first two described by PubMatic as independent - and leave the holding company example anonymous.

For the industry's structure, the PubMatic and Magnite results now point the same way: independent sell-side platforms with exposure to streaming television, political budgets and app inventory grew in the third quarter. PPC Land noted in March that PubMatic's 2025 revenue fell 3% while Magnite's rose 7%. The third quarter of 2026 narrowed that gap in growth rates, though not in size.

Timeline

Summary

Who: PubMatic, a Nasdaq-listed advertising technology company that sells ad inventory for about 2,040 publishers and app developers, its chief executive Rajeev Goel and chief financial officer Steve Pantelick, and the analysts from Craig-Hallum, Wolfe Research, Evercore, B. Riley, Rosenblatt Securities, KeyBanc Capital Markets and Lake Street Capital Markets who questioned them on the call.

What: PubMatic reported third-quarter 2026 revenue of $79.6 million, up 17%, GAAP net income of $5.5 million against a $6.5 million loss a year earlier, and adjusted EBITDA of $26.0 million, up 133%. CTV, mobile app and emerging products made up approximately 64% of revenue; political spending contributed about $5.6 million; a change in server depreciation added $1.5 million to net income. The board added $150 million to the share repurchase program, and the company guided fourth-quarter revenue to $91 million to $93 million.

When: The results, the Form 8-K and the Form 10-Q were published on November 9, 2026, covering the three months ended September 30, 2026. The earnings call took place the same day at 4:30 p.m. Eastern Time.

Where: PubMatic, which describes itself as having no headquarters and lists Redwood City, California, in its releases, reported through Business Wire and the SEC. Revenue grew 10% in the United States, 17% in EMEA and 54% in Asia Pacific.

Why: PubMatic attributes the growth to buyers moving decisions onto its platform through Activate, Decision Fabric and AgenticOS, to logged-in mobile app and CTV inventory, and to political spending, and the profit increase to lower personnel costs, more revenue per unit of infrastructure and a longer depreciation period for servers. The results matter to advertisers and publishers because they test whether independent sell-side platforms can take a larger share of programmatic decisions as identity signals weaken and as the Google ad tech remedies take effect.