This is the memorandum opinion on remedies in United States v. Google LLC, No. 1:23-cv-108 (LMB/JFA), the advertising-technology monopolization case brought by the United States Department of Justice and a group of plaintiff states in the Alexandria Division of the U.S. District Court for the Eastern District of Virginia. District Judge Leonie M. Brinkema signed the 106-page opinion (docket entry 1858) on September 2, 2026, alongside a two-page order (D.E. 1857). Because sealed trial material was referenced, the opinion was filed under seal for fourteen days; on September 16, 2026 the parties filed a joint notice that none of them believed any redactions were necessary (D.E. 1859), and the court unsealed the opinion in full the same day (D.E. 1860).
The opinion holds that "structural remedies are neither realistic nor needed" and that the parties' "substantially overlapping behavioral remedies as modified in this Memorandum Opinion will be sufficient to 'effectively pry open to competition'" the ad tech markets injured by Google's conduct. It rejects all three structural remedies the plaintiffs sought — divestiture of the AdX ad exchange, open-sourcing of the final auction logic in the DFP publisher ad server, and contingent divestiture of "DFP Remainder" — and directs the parties to file one jointly proposed Final Judgment within 30 days.
Posture as the opinion states it
After a bench trial, the court had found in United States v. Google LLC, 778 F. Supp. 3d 797 (E.D. Va. 2025), that Google violated Sections 1 and 2 of the Sherman Act by willfully acquiring and maintaining monopoly power in the worldwide publisher ad server and ad exchange markets for open-web display advertising (Counts I and II) and by unlawfully tying DFP and AdX (Count IV). Count III, concerning an advertiser ad network market, was dismissed. The liability opinion also found anticompetitive the First Look policy (a first right of refusal for AdX on each impression served through DFP), Last Look (letting AdX see rival exchanges' bids before bidding) and Unified Pricing Rules, which removed publishers' ability to set higher price floors on AdX than on third-party exchanges. The acquisitions of DoubleClick and Admeld were found not anticompetitive when viewed in isolation.
The remedies phase followed four months of additional discovery and an 11-day bench trial at which the parties presented 26 witnesses and more than 100 exhibits; the transcripts cited run from September 22 to October 6, with closing arguments on November 21 (2025). The parties then filed post-trial briefs (D.E. 1810 for plaintiffs, D.E. 1811 for Google) and proposed final judgments (plaintiffs' at D.E. 1814-1, Google's at D.E. 1811-2).
Both sides proposed behavioral remedies in three general categories: interoperability and data sharing, anti-discrimination, and limits on Google's buy-side bidding. Both proposed a Monitor and Technical Committee. Google asked for a six-year term; the plaintiffs sought fifteen years and argued that only structural remedies could restore competition, a position the court describes as based on "their lack of trust in Google given Google's history of anticompetitive conduct."
Market findings
Transaction types. The opinion segments open-web display advertising into direct transactions (non-programmatic direct deals, and programmatic direct deals comprising programmatic guaranteed and preferred deals) and indirect transactions (private auctions, also called private marketplace deals, and open auctions). The remedies it adopts reach four of these: programmatic guaranteed, preferred deals, private auctions and open auctions. Google had initially proposed limiting its remedies to open auctions; its vice president Tim Craycroft committed at trial to enlarge the definition, and Google's post-trial proposal covers all four.
Size and trend. Global digital ad spending was $424 billion in 2023 and is projected to exceed $605 billion by the end of 2027; open-web display holds roughly a 30 percent share of ad format revenues and is a market of "tens of billions of dollars per year." Its growth has slowed relative to retail media, streaming, native, audio and AI chats. In January 2019 more than 40 percent of AdWords impressions were indirect open-web display; by June 2025 the figure was 11 percent. Over the same period YouTube's share of AdWords impressions excluding Search rose from just under 30 percent to over 50 percent. The plaintiffs argued that Google's conduct itself contributed to the slowdown.
Artificial intelligence. A dedicated section (pp. 12–15) finds that the integration of AI in ad tech "is in its early stages" and, contrasting the search case, that AI "has not yet had such a disruptive impact on the ad tech industry." Witnesses from The Trade Desk, Kevel and Goodway described no disruption yet to the publisher ad server market; a PubMatic witness called AI in ad tech "an enabler." The opinion records Google's Auto Ads for AdSense, PubMatic's generative-AI buying product "PubMatic for Buyers," The Trade Desk's use of AI in bidding and inventory evaluation, and Omnicom's Omni Assist. It records early work on embedding ads in AI chat interfaces, including Luzia's experiments with Google's AdMob, and concludes that although these "AI-enabled advancements are still in a nascent stage of development, they could threaten the stability and growth of traditional open-web display ads" and "are relevant to the parties' proposed remedies—in particular, the length of time a specific remedy may require."
Supply-path optimization. The opinion treats supply-path optimization, in which buying tools connect directly to publishers and sometimes skip an ad exchange, as a second potential disruptor. It records The Trade Desk's OpenPath product, PubMatic's partnerships with large advertisers and agencies, and testimony that such activity "could be 75 percent of PubMatic's total buyer activity in the next several years."
Legal framework and the fruits of the violation
The court adopts the remedial objectives stated in Microsoft I, 253 F.3d 34 (D.C. Cir. 2001): to unfetter the market from anticompetitive conduct, terminate the illegal monopoly, deny the defendant the fruits of its violation, and ensure no practices likely to result in monopolization remain. It accepts the plaintiffs' argument that terminating the monopoly is an appropriate objective here, because the liability opinion found Google both acquired and maintained its monopolies, while rejecting the claim that divestiture is the only way to achieve it. Structural relief "requires a clearer indication of a significant causal connection between the conduct and creation or maintenance of market power," and the opinion draws on NCAA v. Alston, 594 U.S. 69 (2021), for the caution that "markets are often more effective than the heavy hand of judicial power." It relies throughout on Judge Amit Mehta's remedies decision in United States v. Google LLC, 803 F. Supp. 3d 18 (D.D.C. 2025).
The court identifies three fruits of Google's conduct:
- Monopoly power. Google restricted AdWords' demand largely to AdX and required publishers to use DFP to receive AdX real-time bids; only 11 percent of AdWords open-web display transactions went through non-Google exchanges in 2022, falling to 5 percent in 2024. From 2018 to 2022 AdX handled 63 to 71 percent of worldwide open-web display transactions among exchanges producing data, roughly nine times the share of the next-largest competitor at 6 percent, and in 2022 DFP held 91 percent of the worldwide publisher ad server market by impressions. The court also treats the "black box" character of DFP's final auction logic, raised by industry witnesses at the remedies trial, together with DFP's monopoly power.
- Scale. Google Ad Manager processes at peak about 8.2 million ad requests and about 60 million bid requests per second, over 600 billion ad requests and almost five trillion bid requests per day; PubMatic processes nearly 1 trillion ad impressions daily.
- Profits. AdX charged a 20 percent take rate for over a decade, against rival exchanges that often charged closer to 10 percent.
Structural remedies rejected
General grounds. The plaintiffs bore a "heavy burden." The court finds that where "the gravamen of the violation was an unlawful tie among integrated technology assets, divestiture is unprecedented," accepting Google's argument that "courts in tying cases have not ordered divestiture of either the tying or tied product." The plaintiffs had not shown that behavioral remedies would be ineffective; liability-phase and remedies-phase testimony indicated that access to AdX real-time bids through other ad servers, elimination of the contractual and policy tie, new integrations and data sharing would restore competition.
On the plaintiffs' central argument, the court holds that "the essence of Plaintiffs' rationale for seeking divestiture boils down to a lack of trust that Google will comply with an Order from this Court and an unrealistic desire for certainty," and that "lack of trust alone or desire for 'guaranteed certainty' cannot justify divestiture." It notes that the plaintiffs' economic expert Robin Lee testified that behavioral remedies would suffice if Google acted in complete good faith, that the D.C. Circuit reversed the Microsoft divestiture order that rested partly on untrustworthiness, and that Google would be held in check by "two swords of Damocles": the court-supervised injunction with its contempt power, and the private suits pending against Google, many of which, the opinion notes, were filed by witnesses who testified for the plaintiffs; it cites as examples suits by PubMatic, OpenX, Magnite, Equativ, Index Exchange and Gannett in the Southern District of New York. No prospective acquirer was identified, the proposals carried "tremendous ambiguity," divestiture would take years before any appeal, and every behavioral remedy would be in effect within 15 months. The court declines the plaintiffs' proposed two-step process under Steves & Sons v. JELD-WEN, 988 F.3d 690 (4th Cir. 2021), because it would delay a years-long divestiture and risk being outstripped by "imminent industry disruptions caused by AI and supply-path optimization."
The court also finds the required causal connection not met. Lawful conduct contributed to Google's position, including the AdWords demand built on Search, the approved DoubleClick and Admeld acquisitions, and "multi-billion dollar, lawful investments." It describes the question as "arguably closer" than in Microsoft I or the Google search case but finds insufficient evidence for structural relief.
AdX divestiture. AdX transacts formats beyond open-web display (app, native and instream video), so its sale would not be "tailored to fit the wrong." The plaintiffs' investment banking expert Paul Crisci addressed only initial buyer interest; Index Exchange, PubMatic and Kevel expressed possible interest but none committed, and Kevel lacked the funds to buy AdX alone. Under the plaintiffs' proposal, migration to an acquirer would take up to 540 days after a sale, followed by up to two years of customer migration, and Google would be barred from the ad exchange market for a supervision period of at least ten years. On technical feasibility, AdX and DFP have been integrated with Google's core infrastructure of more than 100 million lines of code for over 17 years. The court cites the seven-year migration of DoubleClick onto Google infrastructure, a five-year replacement of one AdX and DFP dependency by 40 engineers, an eight-year Gmail dependency migration, an abandoned four-year effort to move LinkedIn to Microsoft's cloud, and the seven-year partial migration of Netflix's DVD business to Amazon Web Services, finding the plaintiffs' timelines "unrealistic." Google's own 2023–2024 analysis of a business divestiture, which the plaintiffs relied on, covered only open and private auctions and transferred reference code rather than operational technology. Experts on both sides testified that a divested AdX might not match current functionality.
Open-sourcing DFP's final auction logic. The proposed Open-Source Auction, which the parties agreed should be analyzed as a structural remedy, would reach ad formats beyond open-web display, including the roughly 8,000 publishers who use Google Ad Manager only for direct sales. More fundamentally, the plaintiffs never argued at the liability stage that DFP's final auction logic contributed to the monopolies, and the court never found it anticompetitive: "For this reason alone, the Open-Source Auction will be rejected." The court also finds the proposal technically unproven. It would cover functions such as Enhanced Dynamic Allocation, Billable Event Rate adjustments, Reserve Price Optimization and Optimized Pricing Rules that constitute the "heart and soul of any ad server"; Enhanced Dynamic Allocation uses a seven-day window of bid data computed once a day in about 10 hours on 4,000 computers, which Google's engineering director Glenn Berntson estimated would take 15 to 45 minutes per ad request if run outside Google. Google's internal estimate for a narrower open-source project was four years, and the court finds the remedy would burden publishers, particularly small ones.
Contingent divestiture of DFP Remainder. The proposal would exceed the liability findings because DFP transacts more than open-web display ads. It also lacked commercial detail on asset scoping, a gap the possibility of several buyers made worse, faced the same infrastructure problems as the AdX sale, and, because DFP is currently free for over 90 percent of its publisher customers, carried a "very real" risk of higher costs for publishers. The plaintiffs' proposal would have decided the question almost a decade after entry of judgment.
Behavioral remedies adopted
The court adopts, as modified, an injunction, interoperability requirements, data sharing, non-discrimination provisions and buy-side limits on AdWords. Where the parties proposed different implementation deadlines, the court defers to Google's where shorter (for example, DFP–Prebid interoperability within 12 to 15 months rather than the plaintiffs' 18) and directs the parties to agree a timeframe where Google's is longer (AdX interoperability with rival ad servers: plaintiffs 6 months, Google 12).
- Injunction. Google may not enforce policies or contract terms tying DFP to AdX or recreating the tie, may not reimplement First Look or Last Look, and must deprecate and not reimplement Unified Pricing Rules for all indirect transaction types. Google must allow publishers to set different floor prices across buying tools and exchanges and may not discount DFP fees for uniform pricing. The court cites an estimate by the plaintiffs' liability expert Tim Simcoe that eliminating Unified Pricing Rules alone would lower AdX's take rate from 20 percent to about 16.6 percent.
- Prebid interoperability. Google must build an interface letting Prebid, the open-source header-bidding consortium, solicit AdX real-time bids for all indirect open-web display inventory, and an interface and server-to-server connection letting DFP publishers route all indirect and programmatic direct demand through Prebid, in a "functionally equivalent manner and on materially identical terms" to DFP's current handling of AdX, without added latency and passing the same information. Google also commits to technical support and to letting publishers choose, on a mutually exclusive basis, whether AdX is called from Prebid or from DFP. Non-programmatic direct deals are excluded. The court finds that DFP sharing its pricing and Google audience data with all Prebid participants would remove AdX's information advantage.
- Rival ad server interoperability. AdX must submit real-time bids to rival publisher ad servers "in the same manner it currently submits real-time bids to DFP" for all indirect demand. The court calls this "directly tied to the theory of liability in this case."
- Data sharing. Google must provide publishers with historical and configuration data from DFP and ongoing AdX bid data, including winning and losing bids for each auction, exportable through an interface to rival ad servers, with documentation and support. The plaintiffs' broader version covering all formats, including video and in-app, is rejected as not tailored to the open-web display markets the plaintiffs themselves defined.
- DFP technical documentation. In place of the Open-Source Auction the court adopts Google's proposal to provide publishers with documentation and a per-ad data file showing candidate prices, bids and adjustments used by DFP's "final logic." The plaintiffs' objection that it would not let a rival "kickstart a new business" is rejected as untethered to liability; their objection that it lacks specifics is left to the parties, the Monitor and the Technical Committee to resolve on format and frequency.
- Non-discrimination. Google must use and pass data signals, pass AdX bids and serve DFP ads without regard to whether Google tools are involved, and may not accept compensation for preferential routing. The plaintiffs' drafting is held "too indefinite," lacking specifics and privacy carve-outs, and the parties must craft specific provisions.
- Buy-side (AdWords). Remedies reach AdWords, whose demand the court had called the "golden goose" of Google's conduct at a May 2 hearing, but not DV360, which played no role at liability. The court adopts Google's narrower direct-bidding ban (no direct AdWords–DFP integration, while permitting AdWords to integrate directly with Prebid), a ban on AdWords bidding or routing differently based on Google ownership, a narrow exception for advertiser return on investment (quality, fraud, privacy and similar factors to be specified), and non-discriminatory use of Google first-party data with exceptions for privacy law, internal privacy policy, user opt-outs and children. The first-party data rule "merely puts in writing the status quo."
Monitoring and compliance
The parties agreed to a Monitor chosen from a slate of three candidates, one side proposing the slate and the other choosing; the plaintiffs described which side does which as the main disagreement. The Monitor, once appointed by the court, serves for the term of the Final Judgment with powers to interview Google employees, inspect documents, data, source code, systems and facilities, require reports, interview third parties and hire staff at Google's expense. The court adopts Google's positions that the Monitor's work plan be due in 45 days and that the Monitor report quarterly in writing to both sides. A three-member Technical Committee is formed, each side picking one member and those two picking the third.
Following Judge Mehta, the court gives Google a "broad right to object" to the Monitor's findings, with a warning that abuse could lead to restriction, rejecting the plaintiffs' proposal limiting objections to malfeasance. It rejects the plaintiffs' "sole discretion" language for the United States and requires a status report within 90 days of the effective date. It adopts an Internal Compliance Officer, narrower than proposed: the officer will not run a legal and ethical compliance program or report on Google's document-preservation and privilege practices, which the court holds untethered to liability because the liability opinion declined to sanction spoliation. Google did not oppose, and the court adopts, affidavits every 30 days signed by its chief financial officer and general counsel until a Monitor is appointed, and compliance inspections by the plaintiffs.
Rejected. The plaintiffs' anti-retaliation, non-interference and anti-circumvention provisions are rejected as too vague under Federal Rule of Civil Procedure 65(d), as in the search case, although a policy that employees may report to the plaintiffs, Monitor or compliance officer without reprisal is adopted. The proposed monitored escrow account, funded by 50 percent of AdX and DFP net revenues from April 17, 2025, is rejected because its "lack of specifics is fatal": the plaintiffs did not link the amount to AdX's supracompetitive fees, and the court made no finding on DFP's fees. The court notes, against Google's argument, that federal courts have held disgorgement available under the Sherman Act. Fees and costs require a separate motion.
Term, effective date and geographic scope
The court adopts Google's proposed six-year term, which it describes as consistent with the six-year term in the search case and the five-year term in New York v. Microsoft, and notes it may extend the term if the judgment is not fully satisfied. It holds that the plaintiffs' fifteen-year term "runs the risk of 'impairing rather than enhancing competition,'" quoting the Microsoft court's comparison of remedying a fast-changing industry to "trying to shoe a galloping horse," and adds: "As industry witnesses testified, the rapid emergence of AI and supply-path optimization are likely to dramatically change the industry." The judgment takes effect 60 days after entry, except that work to appoint the Monitor and form the Technical Committee begins immediately.
The Final Judgment applies worldwide. Google had argued, citing the Google Play Store injunction limited to the United States, that international comity barred extraterritorial relief. The court relies on its liability finding that the relevant markets are worldwide, holds that an injunction "must reach worldwide, or else risk adverse effects within the United States," and finds a global application easier to administer given that Google operates its ad tech products on a global basis.
Next steps set by the order
Under D.E. 1857, the parties must meet and confer and file one jointly proposed Final Judgment within 30 days of September 2, 2026, placing unresolved provisions side by side labeled by proponent and without argument; the parties will be told if further briefing or argument is needed. A Word version is also to go to chambers.
Relation to the coverage
PPC Land's account of the full opinion, published September 17, 2026, is consistent with the text on the points it reports, including six years against the fifteen sought, a court-appointed monitor, Prebid and rival ad server interoperability, data sharing, non-discrimination, worldwide scope, the absence of an identified buyer and the "lack of trust" passage (Source: ppc.land). Two refinements follow from the text. Accounts of the release, including The Information's briefing, drew attention to the AI passage (Source: theinformation.com); in the opinion the phrase "threaten the stability and growth" refers to "traditional open-web display ads" and to AI-enabled chat-based advertising described as "still in a nascent stage," in a section that also finds AI has not yet disrupted ad tech. And the AI passages bear on the length of the term and on the risk of a slow divestiture, while the rejection of divestiture rests chiefly on the tying-case precedent, the lack of a buyer, technical infeasibility, tailoring and causation.
Provenance
Retrieved from CourtListener's RECAP archive of the PACER docket (docket 66753787; storage.courtlistener.com). The RECAP entry records D.E. 1858 as a 106-page "MEMORANDUM OPINION. Signed by District Judge Leonie M. Brinkema on 9/2/2026," with text modified and the document unsealed on September 16, 2026 per D.E. 1860. The PDF's text layer is OCR-derived; the raw transcription retains its spacing errors, and quotations on this page restore ordinary word spacing. The accompanying order and unsealing order are transcribed in Raw Sources/United States v. Google LLC - Ad Tech Remedies Order (D.E. 1857, 2026-09-02) and Unsealing Order (D.E. 1860, 2026-09-16).md.
Relationships
- instance-of: AI Antitrust — a monopolization remedy in which the court weighed AI as a market-changing factor in setting the decree's length
- depends-on: United States v. Google (Search) — the opinion follows Judge Mehta's remedies decision on the six-year term, Technical Committee, Google's right to object, the 60-day effective date and the rejection of vague anti-retaliation provisions
- supports: United States v. Google (ad tech) — the primary text of the remedies ruling
- related: Google DeepMind — Google's AI organization; the decree binds Google's ad tech and AdWords businesses
- related: FTC v. Amazon (advertising auctions) — contemporaneous federal enforcement over automated advertising auctions