PUBM · Research
PubMatic Earns Cash After Stock Pay but Two Founders Control the Vote Until 2030
A small ad auction software company with positive owner cash, shrinking share count, and a founder vote lock that buybacks tighten until it expires in December 2030.
Spreadsheet & IC Deliverables: Download the formula-linked PUBM 5-tab Excel (.xlsx) workbook and printable 1-page tearsheet.
Download PUBM .xlsx →- Owner cash after capex, capitalized software and all stock pay
- Plus 5.3% of revenue
- Class B share of all votes, held mostly by the Goel brothers
- 68.9%
- Class B stock converts to one vote
- Dec 11, 2030
1. Our call: Hold off, because the cash is real but the founders and one rival hold the keys
Our call: Hold off, conviction 3 of 5. PubMatic, Inc. runs a sell-side advertising platform: publishers and app developers plug its software into their pages, apps and connected TV channels, and buyers bid for each impression through it [1]. It is a Delaware company based in Redwood City, California, so this is a U.S. pick [1].
It is small, with 2025 revenue of $282.9M, so we would normally pass it by [1]. We chose it for the control setup.
Co-founders Amar and Rajeev Goel hold most of the 10-vote Class B stock, and the Class B carried about 68.9% of all votes on the Jul 30, 2026 count, our calculation [2][3]. That control ends on a fixed date, Dec 11, 2030, when every Class B share converts to Class A [9].
We measure owner cash as operating cash flow, less capital spending, less capitalized software, less all stock pay. For the six months to Jun 30, 2026 that was plus $7.5M, 5.3% of revenue, our calculation [2].
Unlike most of the names we cover, PubMatic earns cash for owners after paying staff in stock [2].
2. A software toll on ad auctions, with Google on both sides
| Metric | Value |
|---|---|
| What it sells | OpenWrap header bidding, Activate direct deals, Connect data and the AgenticOS buying tools, priced as a share of ad spend [1][4] |
| Revenue mix | About 60% from connected TV, mobile app and emerging revenue in the second quarter of 2026 [4] |
| Stock pay | $16.8M in six months, 11.9% of revenue, our calculation [2] |
| Buyer concentration | Three buyers were 24%, 16% and 14% of receivables at Dec 31, 2025 [1] |
| API quality | 3 of 5, our scoring: open Prebid and OpenWrap integrations, and outside models can now run inside its auction, but most value flows through a few large buying platforms [1][4] |
| Toll-gate score | 1 of 4 pass, 2 partial, 1 fail (section 2b) |
2b. PubMatic sits in the auction path, but buyers set the price
A gate is valuable when its owner holds a record others must check and prices what passes through. PubMatic sees every bid its publishers receive through its software, and it processed nearly 92 trillion impressions in the second quarter of 2026, up 18% [4].
| Test | Result | Evidence |
|---|---|---|
| Owns the record | Partial | It holds bid stream data across about 2,000 publishers and app developers, but publishers keep their own audience records [2][4]. |
| Owns the connection | Pass | OpenWrap places PubMatic in the publisher's page or app, and supply path deals were over 55% of platform activity in the second quarter [1][4]. |
| Prices close to outcomes | Fail | Revenue is a share of ad spend that clears, so pricing follows buyer budgets and bidding rules, not publisher results [1]. |
| R&D on the gate vs features | Partial | Technology and development cost $33.8M in 2025 and capitalized software $20.5M, together 19.2% of revenue, our calculation, spent largely on the auction itself [1]. |
The verdict is one pass. The connection is real, because a publisher has to rewire its pages to leave [1].
But the 10-K names Google and The Trade Desk as two of its largest buying relationships, both can end on 30 days' notice, and changes at two large buyers cut revenue in 2024 and 2025 [1].
PubMatic sued Google on Sep 8, 2025 in the Eastern District of Virginia, seeking damages for monopoly conduct in publisher ad serving and exchanges [2]. The 10-K says a material portion of revenue comes through Google and its DV360 buying platform, and warns of retaliation [1].
3. Growth came back in the second quarter
The 2025 decline had three causes the company names: $14.2M of 2024 U.S. election spend that did not recur, and platform changes at one large buyer in the second half of 2024 and another in the second half of 2025 [1]. First quarter 2026 revenue of $62.4M was still soft, though above guidance [8].
The second quarter turned. Mobile app revenue grew over 40% and made up about 25% of revenue, connected TV grew 13% and made up about 20%, and emerging revenue roughly doubled to about 15% [4].
Ad spend from mid-market buying platforms grew over 25%, which reduces reliance on the two largest buyers [4].
4. The quarter on Nov 9
Third quarter guidance is revenue of $75M to $77M and adjusted EBITDA of $17.0M to $19.0M, with a currency drag noted [4]. The midpoint is below the second quarter, which included strong mobile app demand [4].
The quarter will show whether double-digit growth holds, as the chief financial officer said he expects for the year [4]. It will also show buybacks for the period and any change in the Google case, both of which matter for owners [2].
5. Profit is thin on GAAP and wide on the company's measure
Second quarter 2026 GAAP net loss was $1.2M, against $5.2M a year earlier, and adjusted EBITDA was $19.6M, a 25% margin [4]. For the first half, the GAAP net loss was $13.7M and adjusted EBITDA $22.2M [2].
The gap is mostly depreciation of servers and stock pay: $20.0M and $16.8M in the half [2]. Gross margin rose to 67% in the second quarter from 63% because revenue grew faster than data center costs [2].
In 2025 the company lost $14.5M on a GAAP basis while reporting adjusted EBITDA of $61.6M [1]. Stock pay was $38.4M, 13.6% of revenue, our calculation, which is moderate for software [1].
6. Owner cash is positive and rising
Owner cash is our calculation from the cash flow statements [1][2]. It was 9.0% of revenue in 2023, minus 1.0% in 2024, 2.8% in 2025 and 5.3% in the first half of 2026 [1][2].
Working capital moves this number a lot, because PubMatic collects from buyers and pays publishers. In the first half of 2026 receivables used $25.0M and payables supplied $36.6M [2].
A full year is the fairer test, and 2026 is on track to beat 2025 [2].
The company's own free cash flow was $13.7M in the second quarter, up 47%, but that figure leaves out stock pay [4].
7. Capital: no debt and steady buybacks
Cash, equivalents and marketable securities were $137.5M at Jun 30, 2026, with no debt [4]. Through Jun 30, 2026 the company had spent $211.4M buying back 15.5 million Class A shares, with $63.6M left under the 2023 program [4].
It spent $46.5M on buybacks in 2025 and $30.5M in the first half of 2026 [1][2]. In the second quarter alone it retired 2.1 million shares, which it put at 4.2% of fully diluted shares [4].
Unlike many names we cover, the count is falling. Class A shares outstanding fell to 37,303,647 on Jul 30, 2026 from 39,142,185 on Feb 19, 2026, a drop of about 4.7%, our calculation [2][1].
8. The main risks are Google, buyer concentration and leadership turnover
Google. The company is suing a buyer that also competes with it, and the 10-K lists retaliation risks: changed access, worse commercial terms or lower priority [1].
Google can end its buyer agreement on 30 days' notice [1].
Concentration. Three buyers held 24%, 16% and 14% of receivables at the end of 2025, and two large buyers' platform changes already cut revenue [1].
No publisher was more than 10% of revenue [1].
Leadership. Chief growth officer Paulina Klimenko and Americas revenue chief Kyle Dozeman left in 2026, and chief financial officer Steve Pantelick, in the role since 2011, plans to retire, staying into the first quarter of 2027 and then advising through Jul 1, 2027 [8][5].
Data rules. Privacy laws and limits on cookies and mobile identifiers can cut the value of each impression, the 10-K warns [1].
Of 1,030 employees at the end of 2025, 708 were outside the U.S. [3].
9. A new revenue chief to sell agent-driven buying
On Aug 6, 2026 the company named Megan Ramm global chief revenue officer from Aug 10, 2026 [7]. She led global sales at Uber Advertising and earlier worked at Snap and Google [7].
Her offer letter sets a $0.5M salary, a $0.5M target bonus, a $3.45M new-hire stock grant and a $2.3M annual grant from 2027, all vesting quarterly over four years [7]. The search was run by Heidrick and Struggles after the April departures [8].
Rajeev Goel calls the second quarter an inflection point and says agent-driven buying raises the value of performance advertising [4]. The company says more than 80 autonomous campaigns had run since January 2026, including for all five global agency holding companies [4].
10. Superintelligence tools are the pitch, and the auction is the business
PubMatic now markets itself around AgenticOS, which lets software agents plan and buy campaigns, and Decision Fabric, which lets outside buyers run their own models inside its auction [4]. Over 4,000 deals have run through these tools [4].
We treat these as features on top of the auction rather than a new business. Emerging revenue, which includes the new tools, was about 15% of second quarter revenue [4].
If superintelligence agents buy directly from publishers, the party that runs the auction could gain, but the large buying platforms are building the same agents, our reading [1][4].
11. Two brothers control the vote until 2030
Vote control. Class B stock carries 10 votes a share and Class A one, and they vote together [3].
At Apr 1, 2026 there were 38,411,753 Class A and 8,262,574 Class B shares [3].
Amar Goel, co-founder, chairman and part-time chief innovation officer, held 36.4% of total voting power through trusts including the Birchwood Trust [3]. Rajeev Goel, co-founder and chief executive, held 29.3% through family trusts and options [3].
Together that is 65.7%, and directors and officers as a group held 67.8% [3].
Buybacks raise their share. Because only Class A is repurchased, the Class B share of votes rose to about 68.9% on Jul 30, 2026 from about 67.9% on Feb 19, 2026, our calculation [2][1].
Class B would keep a majority unless Class A more than doubled [2].
The 2030 date. All Class B converts to Class A on Dec 11, 2030 [9]
12. The questions we would ask management, and the calls we are tracking
Question one: how much of 2026 revenue still comes through Google, and has it changed since the lawsuit? [1]
Question two: will buybacks continue while the founders' share of votes rises? [2][3]
Question three: will the board add performance conditions to chief executive equity? [3]
Question four: who will succeed the chief financial officer, and when? [5]
Question five: what does the board plan for governance before the 2030 conversion? That answer sets our conviction.
Prediction one: third quarter 2026 revenue is at or above $75M (80%). [4]
Prediction two: 2026 owner cash as we define it is positive (70%). [2]
Prediction three: Class A shares outstanding on the next 10-K cover are below 37.3M (60%).
Bottom line: cash for owners, but the keys sit with the founders and Google
PubMatic earns real cash for owners after stock pay, has no debt, and is shrinking its share count while growth returns. But two founders control the vote until 2030, buybacks raise their share, and a rival it is suing still drives much of its revenue.
We would turn positive on a full year of double-digit growth with no sign of Google pulling back.
Stock Profiles & Comparable Companies
PUBM Stock Profile →More Published Research
View All →Mobility Global Owns the CARFAX Record and the Debt That Paid for the Spin
CARFAX grew 8% and no customer tops 10% of revenue, but the spin left $2.0B of notes and S&P Global's finance chief on the board.
Cerence Looks Fixed but a Samsung Patent Check Did Most of the Work
An in-car voice software supplier whose nine month growth falls to about 4% without one settlement, with no activist, a slipping pay vote and an ex-Intel board tie.