A federal judge refused to break up Google’s ad tech business. The ruling turned on a practical problem the government never solved: who, exactly, was supposed to buy it.
On September 2, 2026, U.S. District Judge Leonie Brinkema rejected the Justice Department’s demand that Google sell AdX, its ad exchange, and instead adopted most of the behavioral remedies the two sides had proposed. Google had already been found liable in April 2025 for illegally monopolizing the publisher ad server and ad exchange markets. It keeps the business, keeps the roughly 20% fee it charges publishers on AdX, and walks away from the structural threat. What it does not walk away from is the liability finding, which a growing line of publishers is now using to build damages cases in New York.
Two things happened in that Alexandria courtroom, and the wires only reported one.
The reported one: Google won. Alphabet, worth about $4.11 trillion, avoided a forced sale of the plumbing that connects websites selling ad space to advertisers buying it. Shares moved less than a percent. The market had priced this out weeks ago.
The unreported one: the judge did not spare AdX because she thought Google’s conduct was fine. She had already ruled it was not. She spared it because a divestiture needs a buyer, and across three and a half years of litigation, nobody ever produced one.
That is a more interesting business problem than the verdict.
What Happened
Brinkema issued a two-page order. It denies the government’s structural remedy and grants, with modification, most of the conduct changes proposed by both the DOJ and Google. The full opinion stays under seal for 14 days while the parties propose redactions, so the operating details of the remedy are not yet public.
The behavioral menu on the table included limits on self-preferencing, mandatory data sharing with publishers, and non-discriminatory treatment of rival ad exchanges and ad servers. The DOJ had also asked the court to force Google to publish parts of the auction code that decides which bid wins, and to keep divestiture available as a fallback if competition failed to improve. Google’s counterproposal was to give rival exchanges equal access to its ad demand and let publishers plug in competing software.
Google’s regulatory affairs chief, Lee-Anne Mulholland, said the company was pleased the court declined to break apart tools small businesses use to find customers. The DOJ called the relief substantial and said it is “evaluating appropriate next steps.”
Both sides are describing the same order. Both are technically right, which tells you how much of the outcome was written before the ruling.
The Backstory
The DOJ and a coalition of state attorneys general sued in January 2023. Brinkema found Google liable on April 17, 2025, under Sections 1 and 2 of the Sherman Act, ruling that it had monopolized the publisher ad server and ad exchange markets and unlawfully tied its DFP ad server to AdX. Her opinion put a number on the harm: the monopoly let Google charge a 20% fee that cut into what publishers earned. The remedies trial ran in September 2025, with closing arguments in November. The order landed this week.
Filing to remedy: three years and seven months.
Now look at what happened to the asset during that window. Google Network, the Alphabet line that carries AdSense, AdMob and Google Ad Manager partner inventory, has shrunk every year since 2022.
- 2022: $32.78 billion
- 2023: $31.31 billion
- 2024: $30.36 billion
- 2025: $29.79 billion
Over the same span Alphabet’s total revenue went from $282.8 billion to $402.8 billion, up roughly 42%. The disputed business fell about 9% while the parent grew by nearly half again. The decline continued into this year: Google Network revenue fell $285 million in Q1 2026 and another $51 million in Q2, per Alphabet’s own 10-Q filings, driven by AdSense weakness that AdMob only partly offsets.
The 2025 monetization metrics tell you what kind of decline it is. Network impressions fell 7% while cost-per-impression rose 7%. Less open-web inventory, sold at a higher clearing price. That is a business shrinking on volume, not on rate.
The scale comparison is the part that should stop you. In the second quarter of 2026 alone, Google Search & other revenue rose $9.08 billion year over year, and Google Cloud rose $11.14 billion. The entire Google Network line billed $7.30 billion that quarter. A single quarter of growth in either of the other two businesses is larger than the whole quarterly revenue of the business the federal government spent three and a half years trying to break up.
Wedbush, working from court documents, sized Google Ad Manager at 4.1% of Google’s revenue and 1.5% of operating profit back in 2020. The share has not gone up.
The Plan
Google now runs an ad exchange under a court-supervised interoperability regime rather than a divestiture order. It is close to what the company offered.
That matters more than it sounds, because the same script already ran in Brussels. On September 5, 2025, the European Commission fined Google €2.95 billion ($3.45 billion) for self-preferencing in ad tech, ordered it to end the practice, and said its preliminary view was that only divesting part of the business would resolve the conflicts of interest. Google had 60 days to propose something. In November 2025 it proposed interoperability improvements and an option for publishers to set different minimum prices for different bidders. No divestment. It also said it would appeal.
Reuters reported at the time that the Commission was partly waiting on the American courts. If a U.S. judge ordered a breakup, Brussels would not need to fight for one. That fallback is now gone. The EU has a fine under appeal, a behavioral order, and a structural threat it has to prosecute alone.
There is one detail that complicates every clean story about this case. In September 2024, Reuters reported that Google itself offered to sell AdX to settle the EU investigation. European publishers rejected the offer, arguing the carve-out was too small to fix the conflict. Lawyers at the time called it the first asset Google had ever offered up in this kind of case.
So Google has been willing to sell AdX. It just wants to pick the terms, the buyer and the perimeter.
The Business Model Angle
Divestibility runs inverse to dominance. A structural remedy is a forced transaction, and a transaction needs a counterparty. Brinkema pressed on how long a sale would take and noted that no buyer for AdX had been identified. That is not a technicality. The qualified operators of a dominant ad exchange are a handful of firms that would each create the same antitrust problem, plus financial buyers with no capacity to run it. A company that has already absorbed or foreclosed its adjacent competitors has, as a side effect, dismantled the market that a court would need to sell it into. The deeper the monopoly, the less divestible the monopolist. Founders building two-sided intermediaries should read that as a strategic property, not an accident.
Litigation duration is now a line item, not a legal matter. Three courts in a year have declined structural relief, and in each the reasoning was about time. Judge Amit Mehta declined to force a Chrome divestiture partly because he expected AI to reshape search habits anyway. The FTC lost its Meta breakup theory when a court held that TikTok had already changed the market the complaint described. Brinkema flagged that a forced sale would drag on for years through appeal. The pattern is not ideological. It is that the docket now moves slower than the market, so the remedy gets priced against a world that no longer exists when it arrives. For a defendant, delay is a discount.
The cheap half of the problem is what got resolved. The expensive half is still open. Since January 2026, Penske Media, Advance Publications, Vox Media, McClatchy and The Atlantic have all sued Google over its ad tech conduct, joining Business Insider, People Inc. and Slate, with cases consolidated in the Southern District of New York on top of a multidistrict litigation running since 2021 that includes Gannett, Mail Media and Emmerich Newspapers. Ad tech firms PubMatic and OpenX have their own claims. Sherman Act damages are trebled, and Gannett and Mail Media have already won partial summary judgment leaning on Brinkema’s findings. The government’s liability ruling is the reusable asset here, and it is the one Google could not get vacated.
For publishers, the remedy is a price question, not a structure question. The single number Brinkema attached to the harm was a 20% take rate. Interoperability mandates only matter to a publisher’s P&L if that rate compresses or if competing demand actually clears higher. Header bidding was supposed to do exactly this a decade ago and mostly produced complexity rather than yield. Watch the take rate, not the press release.
The Risk
The strongest case against this piece’s reading is Google’s own 2024 offer to sell AdX in Europe. If Google was prepared to divest the exchange on its own initiative, then technical separability was never the real obstacle, and a buyer presumably existed at some price. The counter is that an offered carve-out you design yourself, timed to end an investigation, is a different asset from a court-ordered sale with a monitor, a compliance schedule and a hostile plaintiff defining the perimeter. But it is a real dent in any argument that AdX is simply unsellable.
Second, we are reading a two-page order. The substance is sealed for 14 days. If the unsealed opinion contains aggressive interoperability requirements, publisher data rights and a monitor with teeth, the behavioral package could bite harder than the headline suggests. Judge Mehta’s search remedies were dismissed as toothless on day one and are still being argued about.
Third, the take rate may fall for reasons that have nothing to do with the court. Open-web display impressions are already down 7%. If AI answers keep compressing the inventory that AdX exists to auction, the remedy becomes irrelevant faster than it becomes effective, and Google will have won a fight over a business that was quietly liquidating itself. Google’s own AI products are a meaningful part of that pressure, which is an awkward fact for everyone in the courtroom.
Fourth, this is not the last front. The FTC’s consumer protection unit has been probing both Google and Amazon over whether they properly disclosed advertising terms and pricing, and the FTC sued Amazon over exactly that on August 31. Disclosure cases have a much lower evidentiary bar than monopolization cases.
Here is the falsification test. If AdX’s effective take rate drops materially below 20% within four quarters, or if Google Network revenue stabilizes because rival exchanges genuinely win share, then the behavioral remedy worked and the reading above is wrong. If the take rate holds and the segment keeps sliding, the remedy was scenery.
Quick Questions
Did Google win? On the structural question, decisively. The liability finding stands, which is the part that costs money in private litigation.
Why does the buyer problem matter so much? Because a court ordering a divestiture is ordering a sale into a market it does not control. If no credible operator exists, the judge is being asked to underwrite a transaction with no clearing price. Judges do not like that role.
What is AdX actually worth to Alphabet? Less than the case implies. Google Ad Manager was about 4.1% of revenue and 1.5% of operating profit in 2020 on Wedbush’s read of the court record, and the segment it sits in has declined four years running.
Does this change anything for a small publisher tomorrow? Not yet. The operative number is the roughly 20% AdX fee, and nothing public says it moves.
What is the next real date? The unsealing of Brinkema’s full opinion, within 14 days, and then the SDNY damages docket.
The Business Model Analyst Take
Antitrust obituaries are being written today, and they are aimed at the wrong target. The interesting failure is not that judges have gone soft. It is that the standard remedy for monopoly assumes a functioning market to sell the monopoly into, and monopolization is precisely the practice of removing one. By the time a court is ready to order a divestiture, the defendant has usually spent a decade making sure the buyer list is empty. That is not a loophole. It is the strategy working exactly as designed, one layer deeper than the conduct anyone litigated.
Google did not beat the case. It outlasted the remedy. The bill for the conduct is still coming, in threes, from a courthouse in Manhattan, from the publishers who were on the other side of that 20%.
