A federal judge found that Google illegally monopolized two pieces of the open-web advertising market. On Wednesday, she let Google keep both of them.
Judge Leonie Brinkema rejected the Justice Department’s demand that Google sell AdX, the exchange that runs instant auctions for publisher inventory. Reuters reported that she accepted most of the parties’ proposed behavioral remedies instead. The detailed opinion will remain sealed for 14 days while confidential material is redacted, so the exact final obligations are still unavailable to the public.[1][2]
This is a specific new development in the long fight over Google’s distribution power. Europe recently turned Google’s search and Play Store defaults into a compliance surface. The September 2 ruling moves a separate U.S. monopoly case into a harsher technical environment: ad auctions whose routing, ranking, pricing, and response timing unfold inside proprietary systems in milliseconds.
The monopoly lives between the page and the bid
Open-web advertising uses a chain of software to fill an ad slot while a page loads. A publisher ad server manages the available inventory. Exchanges gather bids. Buy-side systems represent advertisers. The winning creative appears before the reader has time to notice that an auction happened.
Google owns major components across that chain. Its publisher server, DoubleClick for Publishers, now part of Google Ad Manager, controls inventory for many large sites. AdX connects that inventory to bids, including demand from advertisers using Google’s buy-side tools. The April 2025 liability opinion found monopoly power in publisher ad servers and ad exchanges, plus an unlawful tie between DFP and AdX.[3]
The tie worked through access. Publishers wanting real-time AdX bids effectively had to use DFP. AdX mattered because millions of advertisers used Google Ads, and that demand gave publishers a reason to stay inside Google’s stack. The court also found that Google used auction mechanisms including First Look, Last Look, and Unified Pricing Rules to favor its exchange or constrain how publishers priced competing demand.[3][4]
AdX has kept a 20 percent take rate for years. Reuters reported that publishers pay that fee on auctions run as users load pages. The liability opinion described the broader economic consequence in blunt terms: Google’s conduct harmed publishers, competition, and consumers of information on the open web.[1][3]
This is where the case reaches beyond ad-industry plumbing. Advertising remains one of the financing systems for journalism, forums, niche publications, free software documentation, and independent sites. A private tollbooth between publisher inventory and advertiser demand changes which parts of the web can afford to exist.
Divestiture would have changed who controls the machine
The Justice Department proposed selling AdX to an independent buyer. Its plan also included a phased remedy for DFP: move the final auction logic into an open-source implementation administered outside Google, then permit a later DFP sale if competition failed to recover. A court-appointed monitor, interoperability requirements, and an escrow fund for publisher switching costs would support the transition.[5][6]
Google proposed conduct rules. Those proposals included giving rival publisher ad servers access to real-time AdX bids, placing a server-to-server Prebid integration between DFP and AdX, ending First Look and Last Look, retiring Unified Pricing Rules for open-web display ads, and providing publisher data for migrations.[4][7]
The final redacted order will show which provisions survived and how they were written. Wednesday’s public result already establishes the governing choice. Google retains the exchange and publisher server under common ownership. Competition has to emerge through interfaces, promises, prohibitions, and supervision around that ownership.
A protocol can comply on paper and discriminate in production
Behavioral remedies sound concrete when written as verbs: share, connect, answer, export, stop. Each verb hides an engineering specification.
“Provide real-time bids” raises questions about response rate, field parity, timeout behavior, identifiers, privacy filtering, auction context, and latency. “Integrate with Prebid” leaves room to define eligible formats and transactions. “Support migration” can mean a complete, documented export or a pile of files that preserves data while destroying operational continuity.
AdExchanger reported that Google’s proposed Prebid obligation covered open-web display ads while excluding video, connected television, in-app inventory, and some curated or deal-based impressions. It also reported Google’s promise to avoid adding latency, reducing bid-response frequency, or withholding information for publishers using other ad servers.[7] Those details describe the problem better than a generic order to interoperate. A rival connection can technically exist while receiving slower, thinner, or less valuable signals.
The Justice Department made this argument directly in its response to Google’s proposal. It warned that Google could degrade the value of bids passed to Prebid or rival servers by throttling signals, recreating the economic pressure that keeps publishers on DFP while appearing to honor the interface requirement.[8]
This converts the court from an institution that declares a rule into the sponsor of a technical inspection regime. Every product release can change the meaning of compliance. Every new ad format creates a scope dispute. Every performance regression needs attribution. Every private experiment can become a miniature antitrust hearing.
The DOJ anticipated that burden. Its closing position warned that behavioral enforcement would invite Google to test boundaries and force repeated litigation over compliance. Google answered that a sale would create years of migration risk, buyer uncertainty, and damaged technology. Brinkema chose the operational risk of supervision over the operational risk of separation.[4]
The sealed opinion delays the useful argument
The public knows the structural result and little else. TechCrunch reported that the ruling gave no specifics, with the written reasoning sealed for redaction. Reuters reported that most proposed behavioral remedies were accepted. Until the order appears, claims about the final implementation should remain provisional.[1][2]
That delay matters because the quality of a behavioral remedy resides in definitions. A broad duty to treat rival systems fairly can collapse under exceptions. A precise duty with measurable service levels can create evidence. A monitor with access to experiments, source changes, logs, and publisher complaints can investigate mechanism. A monitor limited to reports written by Google will receive compliance theater in a nice binder.
The order also needs a change-control model. Ad tech does not freeze while a decree runs. Formats move toward video, connected television, retail media, and agent-mediated buying. Google can comply with a rule scoped to yesterday’s open-web display auction while value migrates into categories outside it. AdExchanger’s analysis identified this boundary in the proposed Prebid coverage before the decision.[7]
Winning liability can still preserve the operating model
This ruling is the third recent failure by U.S. enforcers to win a Big Tech breakup, according to Reuters. Courts rejected a Meta divestiture after finding the government had failed to prove its current monopoly case, rejected a Chrome sale as a remedy for Google’s search monopoly, and have now rejected an AdX sale after Google lost on ad-tech liability.[1]
The cases differ. Their combined signal is still ugly. A platform can be declared an illegal monopolist while the integrated asset structure that produced the power remains intact. The remedy then asks rivals to connect through interfaces maintained by the same company and asks regulators to prove discrimination inside systems few outsiders can observe.
Behavioral relief can work when obligations are narrow, measurable, and paired with strong access to evidence. The redacted order may contain exactly that discipline. It may also inherit the standard failure mode of platform interoperability mandates: a formal doorway whose dimensions, traffic rules, and maintenance schedule remain controlled by the incumbent.
The useful test will come from publisher behavior and rival economics. Can a publisher leave DFP while retaining competitive access to AdX demand. Can Prebid or another server receive bids with equivalent timing and information. Can publishers set prices without hidden penalties. Can a rival exchange gain volume without Google changing an adjacent path. Can a monitor detect the answer before the market absorbs another year of damage.
Google won the right to keep running the machine. The court now owns the problem of making its internal conduct legible.
Sources
[1] https://www.reuters.com/legal/litigation/google-defeats-us-bid-force-ad-tech-sale-2026-09-02 | Reuters, Google defeats US bid to force ad tech sale [2] https://techcrunch.com/2026/09/02/google-spared-from-ad-business-breakup-but-judge-orders-changes-to-how-it-operates | TechCrunch, Google spared from ad-business breakup [3] https://law.justia.com/cases/federal/district-courts/virginia/vaedce/1:2023cv00108/533508/1410 | U.S. District Court, April 2025 liability opinion [4] https://www.nortonrosefulbright.com/en/knowledge/publications/cb2c3ded/what-you-need-to-know-from-closing-arguments-in-us-v-google | Norton Rose Fulbright, remedies analysis [5] https://www.justice.gov/atr/media/1412586/dl?inline= | DOJ, revised notice of proposed remedies [6] https://techpolicy.press/doj-vs-google-back-to-court-for-remedies-to-break-digital-ads-monopoly | Tech Policy Press, remedies mechanism analysis [7] https://www.adexchanger.com/antitrust/the-doj-and-google-sharpen-their-remedy-proposals-as-the-two-sides-prepare-for-closing-arguments | AdExchanger, post-trial remedy proposals [8] https://www.justice.gov/atr/media/1420131/dl | DOJ, response to Google’s final remedy proposal