PubMatic's 17% Quarter Has a Footnote: Five Points of Real Leverage, Two of Depreciation
Revenue accelerated for a third straight quarter and EBITDA doubled — but the margin story, the Q4 guide and an agentic future measured in campaigns all invite scrutiny
PUBM · Earnings Call · 2026-11-09
The Quarter That Was Supposed to Be Loud
For six straight quarters PubMatic's story has been a promise: the sell side gets smarter, agentic buying arrives, and the business reaccelerates once a legacy DSP headwind is lapped. On November 9 management finally delivered the arithmetic. Revenue grew 17% year-over-year, the third consecutive quarter of acceleration, and adjusted EBITDA more than doubled to $26 million at a 33% margin against $11.2 million and 16% a year ago. “Revenue grew 17% year-over-year and adjusted EBITDA more than doubled with a margin of 33%, both well ahead of expectations.” — Rajeev Goel, Co-Founder and Chief Executive Officer (CEO) · 2026-11-09 The shape of the year is unusual and worth stating plainly: “5% growth in the first half, which absorbed the legacy DSP change and a rapid reacceleration to approximately 16% in the second half” — Steven Pantelick, Chief Financial Officer (CFO) · 2026-11-09. But note the second derivative. Q3 printed 17%, and the Q4 guide of $91–93 million implies 15% at the midpoint. The acceleration story is therefore largely a Q3 event; what Q4 buys is a much higher exit run-rate heading into 2027, on top of a $312–314 million full year. The tape already suspected something. Shares are up 27.3% over the last 90 trading days, though they sit 11.6% below the October 1 high of $19.18 and remain 42% lower since the December 2020 reference and 76% below the 2021 peak of $69.92. This is a recovering name with a still-discounted multiple, not a momentum darling.The Margin Has a Footnote
Gross margin of nearly 70% versus 63% is the headline that makes the operating-leverage thesis feel real. Management was unusually transparent that 5 of those points came from operations and 2 from a change in the assumed life of server infrastructure — three years to five. “you can make sort of an adjustment of, I'll call it, $1.5 million, $2 million in terms of that impact, and that will dissipate as these servers are end of life” — Steven Pantelick, Chief Financial Officer (CFO) · 2026-11-09. The $1.5–2 million per quarter is real cash-free earnings help and it expires. Second wrinkle: operating expenses were 8% lower year-over-year, with hiring and go-to-market spend explicitly pushed into Q4. Q3's cost base is therefore not the run-rate. Strip both effects and the underlying improvement is smaller than the optics — though still directionally genuine, because monetized impressions rose 15% while gross impressions processed fell 8%. That is real capacity reallocation, not just accounting. Gross margin has now climbed four of the last five quarters after a multi-year slide, and operating margin has gone from deeply negative in 2025 to marginally positive. The gap between the adjusted and GAAP pictures is the honest measure of how much operating leverage is still aspirational.Sell-Side Decisioning: The Phrase Changed, the Bet Didn't
The vocabulary shift this quarter is subtle but telling. For several quarters PubMatic sold "sell-side targeting." Now it sells sell side decisioning, and the mechanism is traffic shaping. The argument: DSPs cap how many impressions they will even look at, so buyers who bid inside the SSP see more inventory and better signal fidelity before the funnel narrows. That framing is far more actionable for an agency than "we have good data." What follows from it is measurable. Activate, the direct-buy interface, grew more than 8x year-over-year — the third straight leg up in a sequence that started around 4x in late 2025. Mid-market DSP spend grew over 40%. CTV accelerated past 30% globally and 40% in the Americas at 21% of revenue; mobile app grew nearly 40% at roughly 27%. Together with emerging streams, the high-value channels crossed roughly two-thirds of revenue and grew over 50%. Independent agencies are the fastest adopters, which is exactly what one would expect from firms trying to buy technology leverage against holding companies.Agentic: From Slide to 120 Campaigns — and a New Door
PubMatic's most-quoted company keyword is now AI agent, and its second is traffic shaping. That ordering tells you where the narrative has moved. Fully agentic campaign count rose to more than 120 from over 80 last quarter — a respectable slope, but a number that remains economically irrelevant to a $313 million revenue base. The CEO's own read on the gating factor has changed tone:That is a more honest framing than the August call, when the same 25%-by-2028 / 50%-by-2030 forecast was paired with “the seeds are planted and the grass is starting to grow” — Rajeev Goel, Co-Founder and CEO · 2026-08-06. The forecast is unchanged; the explanation for why it is not arriving faster has matured from metaphor to org-chart reality. What is genuinely new is ChatGPT ads. In May, Rajeev Goel was speculating about OpenAI's "$100 billion ambition around advertising" and describing an ecosystem-wide effort. “it's going to be an ecosystem-wide effort, I think, for them to get to that level. And I think we do have an opportunity to partner with companies like that.” — Rajeev Goel, Co-Founder and CEO · 2026-05-08 Six months later the speculation is a partnership, and management explicitly argues the category comes out of both search and display budgets. It contributes nothing to guidance today, but it is a free option on a category that is showing up across the reporter set — Yelp, Airbnb and QuinStreet all spent their calls on agents and shopping agents. PubMatic is riding a wave here, not creating one.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.