One bidder. 2.35 times more revenue per publisher than the next-largest SSP in the auction. 20.5% of total site revenue, on average, from a single buyer. Across thousands of publisher sites and roughly 70 billion impressions a month. Median 17.6%. The next-largest single SSP on the same sites averages 8.7%.

That is the headline number from Playwire's 2026 State of Publisher Ad Revenue report, read by approximately fourteen people outside the publisher-side world, and not covered by a single one of the ad tech trade publications that exist specifically to cover this industry.

The bidder is Amazon.

Duh, of course it is.

One dollar in every five. From one buyer.

With no contract, no SLA, no published eligibility criteria, no appeals process, and (this is the funny part) no acknowledgment from anyone in trade press that this might be, you know, a story.

AdExchanger didn't notice.

Digiday didn't notice.

AdWeek, MediaPost, the entire conference circuit, the LinkedIn thought-leader brigade, the analyst desks at Forrester and Gartner.

All of them sat through 2025 and the first half of 2026 watching this happen and managed to file approximately zero stories about it.

That isn't a demand partner. That's the kind of customer concentration a public company's CFO has to disclose in a 10-K risk factor. The only reason publishers don't think of it that way is that the ad tech industry has spent fifteen years training everyone to call it "programmatic demand," which sounds like plumbing and not like a single counterparty quietly controlling a fifth of the P&L.

Now the part where it gets actually interesting.

Sometime in the past 12 to 18 months, a lot of publishers stopped getting that 20.5%. Some lost access to TAM, the version of APS for bigger publishers. Some lost UAM, the version for mid-sized ones. Some got quietly throttled, which is the polite industry term for "the dashboard stops looking the way it used to and nobody at Amazon will return your email."

There has been no press release. No eligibility changelog. No public list of who got cut, when, or why. Amazon has not commented. APS does not publish a directory of its current members and never has. The closest thing to an official acknowledgment is a vague LinkedIn post from an Amazon partner manager that reads like it was lawyered for six weeks before being approved.

Publishers found out the same way publishers always find out. The revenue line went the wrong direction. The head of yield spent a weekend on Slack trying to figure out which integration broke. Eventually someone said "oh, we lost TAM," the way you'd say "oh, we lost a tooth," and then everyone went back to pretending this was normal.

In a functional industry, somebody would have written the story. The trades would have called Amazon for comment. Someone would have run a "what publishers are saying" piece. Someone would have gotten a leaked internal memo. None of that happened. The story has lived almost entirely in private Slack channels, one quietly devastating data point in a Playwire report that approximately fourteen people outside the publisher-side world have actually read, and a handful of LinkedIn posts that get fifty likes and no follow-up.

There are reasons for the silence and none of them are flattering to anyone involved.

Amazon doesn't comment because APS has always been invite-only and Amazon prefers to operate the way Amazon always operates, which is by saying nothing and letting the lawyers handle it.

The SSPs benefiting from the substitution don't say anything because the substitution is making them money, and you do not call attention to the thing making you money. Magnite, PubMatic, and Index Exchange all now sit on Amazon's "Certified Supply Exchange" preferred-partner list, which is exactly the kind of name a marketing team comes up with when the legal team won't let them say "we are now Amazon's chosen pipes." All three have spent 2025 and 2026 deepening Amazon integrations and very loudly not mentioning it on earnings calls.

Michael Barrett, Magnite's CEO, did a 40-minute interview about Magnite's CTV dominance and did not say the word Amazon, APS, TAM, UAM, or Fire TV once. Not one time. In forty minutes. About Magnite's CTV business. Which has a formal partnership with Amazon, announced via press release. That is not an accident. That is a comms strategy. Hi, Michael.

And then there is Lori Goode, Index Exchange's CMO, who spent five years inside Amazon Advertising before taking the Index job.

Goode is not Michael Barrett. Goode talked to us. On the record. On camera. And what she said is the entire architectural shift this report is built around, delivered in plain English, by a former Amazon Advertising executive now running marketing at one of the three founding Certified Supply Exchange SSPs.

What Goode told us:

Read those three quotes together. Decisioning is moving sell-side. It is happening before the bid reaches the DSP. It is designed to work across every DSP. And Index has built a programmable layer where other companies' AI and algorithms can run on top of the exchange.

That is the architectural admission. That is the shape of the new market. That is what this whole report is built around, delivered on the record by someone who would know. Part II shows you what it means. Hi, Lori. Thank you.

The publishers being cut don't say anything because they're mostly mid-tier sites without an IR department or a comms person, and because complaining publicly about Amazon is the kind of move that ensures you stay cut. Omertà, but for ad ops.

And Playwire wrote the public number because Playwire is publisher-side and has a commercial interest in publishers actually understanding their own exposure. Which is to say: the only entity in the entire ecosystem incentivized to surface this data was a managed-services SSP. Everyone else preferred you didn't know.

Which brings us to the three things every publisher running APS exposure should already know and almost none of them do.

The standard mental model in this industry treats SSPs and bidders as fungible commodities. Lose one, add two more, rerun the auction, fill the gap. That model is completely wrong when one bidder is 2.35 times larger than the next. You cannot replace Amazon by adding five new SSPs any more than a SaaS company can replace its biggest customer by adding five small ones. The math doesn't work. The CFOs of public publishing companies will eventually have to explain that to their auditors, and it is going to be an uncomfortable phone call.

Rajeev Goel, PubMatic's CEO, on what the new Amazon DSP integration is actually delivering for publishers:

Ten percent. That is the single best public number any substitute SSP has produced about replacing Amazon revenue. It is also, if you're doing the math at home, half the size of the hole. A publisher losing 20.5% of total revenue and gaining a 10% eCPM lift on a portion of their remaining inventory is not whole. They are running a managed decline and dressing it up as "demand diversification" because that's the phrase consultants charge $40,000 to put in a deck.

Goel also said this, which is the part everyone in ad tech should have stopped scrolling for:

Translation from the corporate: buyers are consolidating, publishers are consolidating, and the SSPs not on the short list are getting squeezed out of the middle. When Goel says "buyers are consolidating supply relationships around SSPs," he means PubMatic. The compliment is to himself. And the buyer he's talking about, the one driving the consolidation, is the one nobody in this paragraph is allowed to name out loud.

It's Amazon. It's always Amazon.

And again. That's the best case. PubMatic is the SSP that actually had a number to share. Magnite doesn't have a public substitution number at all. Index has Lori Goode's architectural quotes, which is more than anyone else has produced, but still no revenue number attached. Criteo's current CEO Michael Komasinski (Megan Clarken retired; if you've been reading trade press that still treats her as current, that trade press is more than a year behind) has not engaged the Amazon competition question publicly. The "Amazon alternatives" market is operating on vibes and press releases.

This is the part the industry really missed.

TAM and UAM are being restructured. The dynamic floor logic that used to live inside them has been moved to the Prebid wrapper. The new Prebid adapter is in open beta. The Certified Supply Exchange routes Amazon DSP demand through Magnite, PubMatic, Index, and now Microsoft Monetize, with TripleLift added shortly after. The publishers who think they "lost Amazon" mostly lost the direct path. The demand still flows. It just doesn't flow to them, and they don't control the routing anymore.

Read that one more time. The demand still flows. The 20.5% is still there. It just goes somewhere else first, takes a transit fee, and arrives at a publisher who is now an Amazon-preferred SSP customer instead of an Amazon direct customer.

This is not a purge. This is a rewiring. And the publishers experiencing it as a purge are experiencing it that way because they don't know the new wiring diagram yet. Neither does most of trade press. Which is why nobody is writing about it. Goode described the new wiring on camera. The rest is downstream of that.

Amazon has, without ever announcing it, become the single most important counterparty on the open web. It is now restructuring the terms of that relationship from a position of total leverage, against a publisher base that has no audit framework, no substitution playbook, no procurement language, and no early-warning system for detecting throttling before it becomes access loss. The trade press did not notice. The analysts did not notice. The conference circuit did not notice.

The CMO of one of the three founding Certified Supply Exchange SSPs sat down with us and described the architecture in plain English. A managed-services SSP wrote a research report, buried the number in chart eleven, and the entire industry kept scrolling.

Twenty point five percent. One buyer. No contract. No SLA. No press coverage.

The wiring diagram for the open web changed in 2025. The map most publishers are still using is from 2023. We drew the new one. Part II shows you who's holding the pen, and what Lori Goode told us that the rest of the industry should have been talking about for the last six months.

Quick honest note. If I am going to ask you to upgrade, I would rather just tell you what you are getting than do the cliffhanger thing every other newsletter does.

The free part is the diagnosis. Amazon is 17.6% of publisher revenue at the median, 20.5% at the mean. PubMatic's DTE integration recovers about 10% of eCPM, which is not the same thing as recovering 10% of lost revenue. The pipes are being rerouted through a Certified Supply Exchange that now has at least five preferred SSPs and is still growing. Trade press is not covering it. That is what you needed to know to understand the problem.

The paid part is the audit document. Specifically the artifacts you can copy into a Google Doc and bring to a meeting Monday morning. Which is, after all, the only thing that justifies a B2B subscription.

This is what a CFO actually reads. Cover, argument, calculator with a worked example, scale table, six warning signs of active throttling, a five-row pathway decision table by publisher profile, a five-step activation sequence, the CMO procurement memo verbatim, a methodology summary, and a back cover. No filler. No 16-week activation checklists you will never read. No glossary nobody opens. Twelve pages.

The brief is the version you forward to your head of revenue, your agency review team, or your board prep folder. It is the version that gets read.

This is what your analyst opens when the brief raises a question the brief did not answer. The full version includes everything in the brief plus:

The eleven-point health check with severity scoring, not the compressed six. The full five-pathway matrix comparing Magnite, PubMatic DTE, Index Exchange, Criteo, and the Prebid adapter across eleven columns including contact roles and qualification timelines. A sixteen-week activation checklist with deliverables and walk-away signals at each stage, plus an SSP outreach email template you can send as written. A Prebid adapter migration guide, because the open beta window is the cheapest moment to make this transition and Amazon will not warn you when it closes. A vertical vulnerability scorecard explaining why news, sports, and tech are the canaries and gaming, entertainment, and education are not. An eight-phrase decoder for SSP earnings calls with confidence ratings (HIGH, MEDIUM, INFERENTIAL) on each phrase, so you can read every quarterly transcript correctly and figure out which SSPs are actually picking up Amazon money versus which ones are talking supply path optimization because their CEO read the McKinsey deck.

Plus the methodology disclosures the brief had to compress. What we know. What we do not know. Why Amazon is silent on all of this and what that silence means. The full v1.1 changelog explaining what we corrected from v1.0 and why. The reference document is what makes the brief defensible.

The Excel file with the calculator and the vertical scorecard wired up as live formulas. Blue cells are inputs you edit. Black cells are formulas that recalculate. Enter your monthly revenue, your APS share, your portfolio size, and your assumptions about substitute SSP fill rate and eCPM. The spreadsheet returns your annualized exposure, your modeled recovery, your residual gap, and your concentration risk flag. The scorecard lets you enter your own publisher's dimension scores rather than accepting the editorial defaults.

The spreadsheet is the artifact you save in your board prep folder. The version you used. Dated. Auditable.

Brief on Sunday night. Spreadsheet on Monday morning. Encyclopedic edition open on a second monitor when the brief raises a specific question you need to chase. That is the sequence the documents are designed for, and it is the sequence the documents reward.

The CMO procurement memo deserves its own paragraph. It is the advertiser-side artifact in this package, and it is the only document of its kind currently in circulation. One page. Three findings to confirm about your current Amazon DSP share. Three questions to ask your DSP about supply-path-optimization configuration. Three actions to take this quarter. Designed to be handed to an agency review team Monday morning. Designed to force a structural conversation about Amazon DSP concentration before lock-in is complete.

The same data that quantifies publisher captivity quantifies advertiser leverage on the Amazon DSP side. Today, that leverage runs in the advertiser's favor. On a twelve to twenty-four month horizon, the leverage inverts. The memo is the paper trail showing your marketing leadership recognized the trajectory early.

That is the half of this story nobody else is connecting. It is in all three documents.

If you are a publisher running APS exposure, this is the document you wish you had twelve months ago.

If you are an advertiser running Amazon DSP, this is the document you will wish you had twelve months from now.

If you are an SSP executive, you already know who you are. Hi.

ADOTAT+ subscribers get the brief, the encyclopedic edition, and the live spreadsheet. Updated quarterly as the APS ecosystem evolves. Confidential to subscribers.

Everyone else got the diagnosis, which is roughly 20.5% more than they had this morning. Or 17.6% at the median, if you are reading carefully. The kind of distinction the people behind the paywall are paid to notice.

Stay Bold. Stay Curious. Know More Than You Did Yesterday.