The announcement itself reads like plumbing. On July 30, 2026, Amazon Ads told advertisers that every existing Amazon DSP account would be upgraded automatically into a single advertiser account spanning 34 countries, with no action required, no re-registration, and legacy identifiers left running. The same July day, Amazon reported second-quarter advertising services revenue of $19.8 billion, up 26% year over year, with sports inventory selling out. One document counted the money. The other rebuilt the pipes that collect it.

Taken alone, an account migration is housekeeping. Taken alongside everything Amazon has shipped around its demand-side platform since late 2025, it reads as something else entirely: the latest load-bearing wall in a structure designed to make buying advertising through Amazon cheaper than buying it anywhere else, and to make the humans who traditionally sat between advertiser and impression, agencies above all, progressively optional. Has Amazon ever stated that plan? No. Does the evidence, laid end to end, describe it anyway? That is the question this edition works through, using the fee data, the product defaults, the spending shifts and the earnings wreckage of the past week as material.

One login, 34 countries, and a silence about agencies

Start with who, what and when. Amazon Ads published the launch announcement on July 30, 2026, converting existing Amazon DSP advertiser accounts into unified advertiser accounts. The upgrade lets buyers run programmatic campaigns, Sponsored Ads and Amazon Marketing Cloud workloads across 34 countries from one login, and it groups the change into five capabilities: multi-country buying across the Americas, EMEA and APAC without new registrations, omnichannel campaign management that removes the account switch between programmatic and Sponsored Ads, clean room access provisioned by the upgrade itself, unified billing and permissions on one page, and first-party data connected once for activation everywhere. Amazon's own framing is blunt. Advertising across its product suite and multiple countries historically required separate accounts, distinct registrations, fragmented billing and redundant data uploads, and the announcement states the new structure "eliminates this complexity."

The engineering detail matters as much as the marketing. Accounts receive a new Global Account ID plus regional identifiers, while legacy CFIDs continue to resolve without modification, an additive migration rather than a breaking one. Mappings between old and new identifiers come back from a POST call to the advertiser accounts query endpoint. For campaign creation across borders, an optional country code passes through the campaigns endpoint, and when no code is supplied, campaigns default to the original country. Agencies running scripted campaign creation at volume will therefore see no silent geographic drift. The capability is opt-in at the request level even though the account change itself is not.

Where does it apply? Across five regions: the United States, Canada and Mexico in North America; Brazil alone in South America; seventeen European markets from Germany and France to Luxembourg; seven Middle Eastern markets including Saudi Arabia, the United Arab Emirates and Israel; and six across Asia Pacific, among them Japan, India and China. The 34-market perimeter is deliberately familiar. It matches the footprint Amazon used when it expanded in-market audiences to 32 countries with unified targeting definitions on November 4, 2025. Each release extends the same map rather than drawing a new one.

The why becomes visible only across the full two-year arc. Amazon made Amazon Marketing Cloud directly accessible to sponsored ads advertisers on September 16, 2025, removing the partner coordination step that had gated the clean room. It launched the unified advertiser account in open beta on October 24, 2025, announced by Pratham Sarin, a product and engineering leader, during a UK and EU roadshow. At its unBoxed conference on November 10 and 11, 2025, it collapsed Amazon DSP and Sponsored Ads into a single Campaign Manager interface and opened a consolidated reporting experience in beta. Digiday reported at the time that the platform would enter general availability in December 2025 with a gradual migration through 2026, supported by what an Amazon executive described as extensive roadshows. Reporting reached general availability on June 8, 2026, with two legacy Ads Console tools scheduled for retirement by December 31, 2026. On the data side, Amazon published a technical walkthrough of Ads Data Manager on May 1, 2026, presented by partner solutions architect Mayank Arora, then set a June 30, 2026 consent deadline for advertiser data transmitted in the United Kingdom and European Economic Area, and made first-party paid feature signals free to query in the clean room through December 31, 2026. Interface, reporting, clean room access, data onboarding and now account identity: consolidated in turn, on schedule, without drama.

The practical payoff for buyers is real. A brand advertising in eight European markets previously needed eight account contexts, per-entity billing, per-account permissions and per-market audience uploads. That work now collapses into one account, one billing page, one permission model and one data connection. Cross-product attribution, the question that has dogged retail media buyers for years, becomes answerable inside a single measurement instance, supported by self-service access to more than fifty third-party studies across eighteen countries. Rivals see the same prize: Google shipped its Data Manager API in December 2025, The Trade Desk folded Galileo into a partner portal in March 2026, and Nexxen cut data onboarding to 24 hours in July 2026. Every buying platform wants to be the shortest path between a customer file and an addressable impression.

Read the announcement a second time, though, and notice what it does not say. No completion date for the upgrade wave. No word on whether managed-service accounts follow the same path as self-service ones. Nothing on minimum spend commitments, which Amazon DSP has historically carried and Sponsored Ads has not. And, most tellingly, nothing at all about the layer above the advertiser: multi-account management across client portfolios, the daily terrain of every media agency, goes entirely unaddressed. In a document about who holds the account, the party that traditionally held it on the advertiser's behalf does not appear once.

The price is the strategy, not a promotion

How cheap is cheap? The numbers have been on the record for a while, scattered across trade reporting, and they compound when assembled. AdExchanger reported in November 2025 that Amazon DSP charges no fees for programmatic guaranteed deals on Amazon-owned media and collects a 1% fee for ads on open web publishers. Digiday's reporting from October 2025, drawing on ad buyers surveyed through the year, placed typical Amazon DSP fees between 4% and 8%, dropping as low as 1% and occasionally to zero, depending on spend levels, agency relationships and inventory type, with additional discounts where buyers use the platform for third-party inventory. Set that against the incumbent. The Trade Desk's own annual reporting has placed its take rate between 19% and 21% since 2017, and investment bank Needham and Co. has attributed a 12 to 15% range to the platform in analyst work. The measures differ, headline platform fee against all-in take, yet the gap survives every methodology.

The consequences showed up early. PPC Land documented in August 2025 how an $80 million annual spend migration moved from The Trade Desk to Amazon DSP, driven primarily by the fee disparity. In the LinkedIn debate that followed, Vlad Chubakov, Associate Director of Programmatic at Delve Deeper, put the structural point plainly: "Amazon can slash margins because they own the pipes." The Trade Desk's chief executive, Jeff Green, liked the post. He has since been forced to argue with its arithmetic in public.

On the August 6, 2026 earnings call, Green offered an illustration: a platform charging 8% buys an impression for $1.08 while a rival charging 4% buys one for $1.04, so the fee gap looks smaller once media cost is included. PPC Land's analysis of the call noted that the illustrative figures sit well below the ranges circulating in the trade press. If the real spread runs closer to 1% against 12%, the comparison becomes $1.12 against $1.01, and the performance a decisioning platform must demonstrate to justify the difference rises accordingly. The Trade Desk does not disclose its take rate.

Nor is the pressure limited to headline fees. On August 7, 2026, PPC Land reported that rivals are wrapping programmatic guaranteed deals in agentic AI at 1% fees, a transaction category the IAB Tech Lab wrote into its agentic specifications in April 2026, and that DataBeat measured agentic buyers clearing inventory 13.4% below conventional demand. Green dismissed such approaches on the call as resembling the ad networks of 2006. The market has kept building them anyway. Price deflation is meanwhile arriving on the supply side too: Roku's second-quarter ad prices dropped 12% as impressions jumped 40%, with advertising revenue still climbing 25% to $673 million and third-party demand-side platforms buying nearly three-quarters of its in-stream video.

Why can Amazon sustain a 1% fee where an independent platform lives or dies on its take rate? Green answered the question himself, without meaning to flatter his rival. A walled garden with near-zero cost of goods on owned inventory keeps the whole dollar when spend lands there, which gives it a structural reason to route budgets inward and a structural cushion to price the pipes at cost. Amazon has run this play before. Selling the transaction below the market rate, absorbing the margin pain longer than any competitor can, and collecting the customer relationship at the end is how the company approached books, then logistics, then cloud computing. Advertising fees are simply the next inventory line.

Software where the hands used to be

Crushing the price answers half of the brief. The other half is removing the labor, and July supplied the clearest evidence yet. On July 22, 2026, Amazon Ads published ten new capabilities for Brand+ and Performance+, the two automated campaign products inside Amazon DSP, documented in detail by PPC Land. The announcement is built on the language of advertiser control. Four of the ten features switch on by default and require no advertiser action at all.

The most consequential of them inverts a decision buyers used to make. Automatic Deal Selection for Brand+ streaming television entered open beta enabled by default on new prospecting lines, replacing the buyer-specified deal with a system-chosen set. Nicole Goksel, senior product manager for the capability at Amazon Ads, described the before-state directly: "Previously, Brand+ campaigns ran on a single, pre-configured deal." Now the platform assembles and refreshes the deal set itself, across Amazon properties and the open internet. Supply path control, curation preferences and negotiated terms, the craft an experienced trader charges for, all sit inside the decision the machine now takes. Opting out remains possible. The default has flipped.

The default also reaches further back than July. Amazon DSP first launched automatic deal selection on April 15, 2026 for streaming television campaigns running an Awareness goal with a Reach KPI, as an opt-in checkbox in the Inventory section, with machine learning adding high-performing deals and dropping weak ones throughout the flight, while disclosing neither the model's reaction speed nor its lookback window. On July 1, 2026 the capability spread to online video across the United States, Canada, Mexico and Brazil. Three weeks after that, the checkbox became the starting position on new Brand+ prospecting lines. From optional experiment to default setting in roughly one hundred days.

Follow the investment and the stakes sharpen considerably. Amazon's own API documentation labels the machine-built collection with a curationSourceType of AMAZON_CURATED_AGENT, distinguishing it from advertiser-curated groups, and buyers can retrieve the resulting deal list and deal-level delivery metrics after the fact. The composition decision, though, happens upstream of anything a buyer approves. Which deals enter the set determines where the budget physically lands, across a pool where Amazon charges nothing on programmatic guaranteed deals over its own media and 1% on open web publishers. A model optimizing toward reach carries no visible instruction to favor owned supply, and no published mechanism preventing it either. The feature also pairs with the exclusive reach measurement Amazon introduced for streaming inventory in December 2024, meaning the platform that assembles the deal set also grades its performance. Investment allocation, in short, becomes a model output. The advertiser sets the budget. The seller's software decides its route, and reports on its own driving.

The data pipeline received the same treatment. Audience Signals entered open beta for both products, letting advertisers route first-party audiences through Ads Data Manager as optimization inputs rather than hard targeting caps, at no additional fee. Combined with the free clean room queries running through December 31, 2026, the cost of moving proprietary data into Amazon's models has been engineered toward zero at the exact moment the account upgrade makes connecting it a one-time act. The supply side mirrors the logic: Amazon Publisher Cloud's Outcome Optimizer, launched June 19, 2026, applies shopping and streaming signals to programmatic guaranteed deals inside FreeWheel's ad server, with early tests showing a 33% improvement in on-target reach. Retail data optimizes the buy from one end and the sell from the other.

Feeding those models is an inventory land grab that has run for a year. Amazon DSP added Spotify's global audio and video inventory across nine markets on October 1, 2025, integrated SiriusXM Media through the AdsWizz platform in September 2025, opened iHeartMedia's owned inventory to DSP buyers on June 29, 2026, and gained Triton Digital supply across more than 80 country marketplaces on July 21, 2026. The July release moved audio for Brand+ to general availability across 16 locales, although podcasts remain outside what the automated products can buy, a gap sitting awkwardly beside all that accumulated supply. Digiday's February 2026 review of the sales materials found Amazon pitching more than 800 free ad-supported streaming channels reaching over 55 million monthly viewers as part of a premium streaming case built through 2025.

One July feature deserves separate attention. The Managed Service Campaign Proposal Builder reached general availability, bringing Brand+ and Performance+ into Amazon Media Group's managed service entities through existing planning workflows, accessed via Amazon sales contacts rather than self-service signup. Read that carefully: the automated products now flow through Amazon's own service arm, plan included. The November 2025 unBoxed enhancements built the models. The July 2026 release wires them into a sales channel that requires no agency at all.

Practitioners have watched this movie. PPC Land's May 2026 analysis recorded buyers drawing an explicit Performance Max analogy, recalling how advertisers welcomed full automation from Google before demanding visibility into where budgets went. Digiday's June 15, 2026 scorecard, built on conversations with 13 media buyers ranking the five major platforms, landed on the same tension: agentic integrations make investments easier to manage while transferring decisions away from experienced media buyers. Every default that switches on moves a task from a person Amazon does not employ to a model it does.

What exactly is left for the agency to sell?

The displacement question stops being theoretical the moment holding company money starts moving. AdExchanger reported on November 18, 2025 that two sources at ad tech platforms observing programmatic bidding patterns saw Omnicom agencies shifting spend from The Trade Desk to Amazon DSP, with one source describing a double-digit share of expected third-quarter spend changing platforms and calling the movement an inflection point in the holding company's bidding habits. The Trade Desk flatly denied any shift. Omnicom confirmed nothing, saying only that it works with a broad range of demand-side partners, and Amazon declined to comment. AdExchanger's follow-up discussion raised the awkward second-order question: how does a holding company manage the conflict of routing client budgets into the platform of a company that is also its client?

Amazon's stated destination leaves little room for interpretation. Kelly MacLean, the Amazon Ads vice president overseeing the DSP, told Digiday in January 2025 that the company wants the product to become "the largest AI-driven buying platform for all digital marketers". In the same piece, Kevin Weiss, vice president of retail media at Skai, projected that continued growth across owned inventory, Performance+ and Brand+ would leave Amazon either the market leader or a far larger rival to the incumbents by the end of 2026. The end of 2026 is four months away.

The absorption pattern reinforces the ambition. When Microsoft retreated from its own advertising marketplace, Digiday reported that Amazon folded Microsoft's Monetize supply into its DSP, stitching more of the open web into the platform, with migrating advertisers onboarded through Amazon Ads representatives or an activation partner program depending on size, agency ties and service agreements. Reach, shopper data and pricing power converge in one place, and the concierge greeting buyers at the door wears an Amazon badge.

Two readings of all this coexist, and honesty requires stating both. The generous reading holds that agencies get converted into channel partners rather than removed. Evidence supports it. Kate Greubel, a vice president at the Omnicom-owned commerce agency Flywheel, told AdExchanger that Amazon has trimmed its DSP fee and offers comparatively low rates for streaming and connected television, the sort of comment made by someone building a practice on the platform, not fleeing it. Agencies still supply the multi-client governance, cross-platform arbitration and strategic counsel that the July 30 account upgrade conspicuously fails to address. Agencies also police intermediary costs harder than anyone: Publicis advised clients to stop transacting on The Trade Desk in March 2026 after a FirmDecisions audit questioned fee application, a dispute resolved by joint statement on June 12, 2026, while Dentsu and WPP quietly exited the platform's OpenPath initiative in February 2026 over transparency and fee concerns. An auditor culture, though, rewards whichever platform shows the smallest visible fee. The fee police may be escorting budgets straight to Amazon.

The harsher reading follows from the product decisions. The skills priced into agency retainers, bid strategy, deal curation, supply path selection, data onboarding, cross-market reconciliation, are precisely the tasks Amazon has spent two years automating or absorbing. When the platform selects the deals by default, connects the data once, unifies billing across 34 countries, and proposes the media plan through its own managed service arm, the agency line item must justify itself on strategy alone, at holding companies whose margins were built on execution. Nothing in Amazon's public materials attacks agencies. Nothing in its product roadmap needs them.

A third reading has started surfacing in the trade press, and it may prove the realistic one: agencies survive by ceasing to be agencies. Digiday reported in September 2025 on S4 Capital's plan to derive consulting revenue from systems integration, guiding chief marketers through reshaped org charts and outcome-based deals, a trajectory S4's Wesley ter Haar summarized without varnish: "The reality is AI is eating the agency business." The destination looks less like Madison Avenue and more like an audit and advisory practice in the mold of PwC, paid for judgment, governance and vendor arbitration rather than for hands on keyboards. The operational math points the same direction. PPC Land reported on August 7, 2026 that Omnicom's PHD is drawing only 10 to 20% productivity gains from AI in APAC while agents already cover 15 to 20% of dentsu's regional workload, with the question of who pays the token bill unresolved. Digiday's June 11, 2026 review of agency AI pitches found advertisers pressing on compensation as automation shrinks the labor behind reporting, planning, creative production and audience analysis, while pushing data portability and governance questions into procurement. The catch sits in the economics. Holding company margins were engineered on execution volume; consultancy economics run on partner hours and scarce judgment, a smaller and slower book of business. Amazon's defaults do not merely take work away from agencies. They take precisely the kind of work agency pricing was built on, and hand back the kind that consultancies already sell better.

Why Google's ad network, publishers and The Trade Desk suffer together

Amazon's squeeze explains only part of the pain spreading through the rest of the field. Three forces now compound: advertising demand concentrating onto owned-and-operated inventory, AI answers cutting the traffic that open web pages depend on, and fee compression that punishes anyone charging for the middle. Each casualty absorbs a different mix of the three.

Start with Google, whose suffering is precisely located rather than general. Search still prints money. The segment facing the open web does not. Network advertising revenue, the AdSense, AdMob and Ad Manager line, declined 1% to $7.4 billion in the second quarter of 2025 while Search grew 12% to $54.2 billion and YouTube 13% to $9.8 billion. By the first quarter of 2026 the Network drop had steepened to 4%, reaching $6.97 billion, the sharpest quarterly decline in recent reporting, against research showing AI Overviews correlated with a 58% reduction in click-through rates for top-ranking pages by December 2025. The most recent quarter extended the pattern: traffic acquisition costs rose 10% against 14% advertising growth, a spread consistent with growth concentrating on owned properties rather than revenue-shared inventory. The legal front tightened in parallel, with a bipartisan brief filed August 4, 2026 in the D.C. Circuit urging a tougher remedy and noting that three firms hold 88% of AI model API revenue. Google prospers where it owns the surface and bleeds where it shares one.

YouTube belongs in this story as the gravity well, not the victim. Jeff Green, citing exhibits from the Google antitrust trial, told investors in November 2025 that roughly all of DV360's incremental dollars between 2019 and 2024 went to YouTube, with spend on the platform rising about 800% while Google's open internet buying stayed essentially flat. He put it more sharply to AdExchanger in August 2025, saying Google and Facebook have "largely abandoned the open internet." The flight of Google's own demand into YouTube stranded the publishers its Network monetizes, and the destination now faces its own contest: Amazon arrives at the same video budgets carrying sold-out sports and a Prime Video pool it sizes at 315 million ad-supported viewers.

Publishers take the hit from both directions at once, losing the demand and the audience. On the supply side, Ozone benchmarking data covering roughly 20 billion impressions showed publisher ad request volumes falling 32% to 37% year over year in the United States and 39% to 41% in the United Kingdom between April and June 2026, with rising eCPMs plugging only part of the revenue gap as zero-click search kept visitors away. The earnings tape reads accordingly. Teads halted its 2026 guidance as premium publisher page views dropped 15% to 25% and adjusted EBITDA fell 74% to $7 million. USA TODAY Co. lost 22 million monthly unique visitors in a single quarter, with digital advertising down 9.2% to $79.8 million and chief executive Mike Reed putting a decision on cutting Google off at nine to fifteen months out. Ziff Davis wrote off $54.8 million on health media as ad revenue dropped 6%, with gaming advertiser retention sliding to 81.6% from 93.7%. People Inc. chief executive Neil Vogel went further than Reed, saying that blocking Google entirely is fully on the table. When the referral engine becomes the answer engine, the inventory it once fed simply stops existing, and no bid density survives the disappearance of the page.

The Trade Desk absorbs all three forces simultaneously, which is why its chart looks the way it does. Pressure from above: integrated platforms price the pipes near zero because media margin pays the bill, an economics no independent can copy. Pressure from below: agentic wrappers moving programmatic guaranteed at 1% fees, clearing 13.4% under conventional demand. Pressure from within: consumer packaged goods and automotive, a quarter of platform spend, in retreat, four chief financial officers in fourteen months documented across the leadership churn, and a measurement regime Green himself concedes makes premium open web inventory look expensive under last-click standards. The deeper problem is structural. An independent platform sells access to a healthy open web, and both suppliers of that health, Google's referral traffic and publishers' pages, are withdrawing at the same time. The common denominator across every casualty is dependence on a multi-party web whose parties keep leaving, while the two integrated stacks convert the exodus into share. Of those two, only one pairs the share gain with a price war.

The verdict arrived this week, 24 percent at a time

Whatever the intent, the scoreboard updated on August 6, 2026. The Trade Desk, the largest independent alternative to Amazon's demand side, reported second-quarter revenue of $715 million, up 3%, missing consensus by $37.6 million, and guided third-quarter revenue to at least $650 million, a floor roughly 12% below the $739 million recorded a year earlier. Shares fell 6.8% in the regular session and a further 24.22% after hours to $13.39, against a 52-week range that had opened at $91.45, leaving the market capitalization at $8.91 billion. Guided adjusted EBITDA of about $160 million implies a margin near 24.6%, down from 43% in the comparable quarter. A company that compounded revenue at roughly 34% annually across a decade as a public business now forecasts its first year-over-year quarterly decline.

The mechanics of the decline are more interesting than the headline. Customer retention stayed above 95%, as it has for over a decade, alongside 3% growth. Clients are not leaving; incremental budget is going elsewhere, campaign by campaign, which makes share shifts faster and less visible than churn. Green attributed the pressure partly to consumer packaged goods and automotive advertisers, roughly a quarter of platform spend, moving toward cheap media, programmatic guaranteed and fixed-price transaction types under macro strain. Every one of those escape routes leads through platforms pricing near 1%. Guideline data assembled in May 2026 showed four demand-side platforms controlling approximately 85% of global programmatic spend, up from 75% in 2022, with Amazon nearly doubling its tracked share in roughly fifteen months, from under 10% to just under 20%, while The Trade Desk grew at approximately the market rate. Amazon's advertising segment, already past $70 billion on a trailing twelve-month basis after the first quarter, then posted the $19.8 billion second quarter. Prime Video alone gives its buyers a pool Amazon sizes at 315 million ad-supported viewers.

Fairness demands the counterweights. Amazon publishes no DSP-specific revenue, so its programmatic share rests on third-party tracking rather than disclosure. The near-zero pricing on owned media repeats the pattern Green criticized in Google, where a platform keeping the whole dollar on its own inventory has every incentive to steer spend there, and a buyer surrendering deal selection to a default cannot easily audit where the machine sent the budget. Whether an automatically optimized deal set outperforms a hand-picked one is measurable only after the fact, and Amazon has not specified the reporting granularity for that comparison. The July 30 announcement left minimum spend commitments, managed-service treatment and the upgrade completion date open. Regulators have so far said nothing. Agencies, for their part, have survived every previous automation wave, from search bidding to Performance Max, by migrating up the stack, and holding companies are already selling Amazon expertise as a practice area.

Still, the shape of the plan is hard to unsee. Crush the price until independents cannot follow. Automate the workflow until intermediaries become a preference rather than a necessity. Unify the account until the shortest path from any advertiser in 34 countries to any impression runs through one login, one data connection and one bill, all owned by the same company. Amazon executed that sequence once before, in retail, where the disintermediated parties were bookstores, then wholesalers, then entire high streets. The July 30 upgrade removed the last registration step between an advertiser and Amazon's global inventory. What the document never states is who, if anyone, Amazon expects to remain standing in between. On the evidence of the past two years, that answer is being engineered rather than announced, and Friday's tape suggests the market has already worked it out.

Also noted