Google Had Its Day In Court. Now, It's Amazon's Turn

Full analysis

Two courtrooms, same lesson. A judge said Google monopolized the pipes that connect publishers to advertisers, then declined to break anything up. The FTC and 22 states sued Amazon for allegedly rigging its retail media auctions to make winners pay near their full bid. Both stories land on the same spot for operators: the auction machinery you route billions through is opaque, the referee is soft, and the only people who can inspect the logs are lawyers with subpoenas.

The decision this episode forces isn't "what do I think of these cases." It's "do I keep trusting auctions I can't audit, and where do I put budget and product roadmap now that the government has told me it won't fix them for me." That's hard to undo once you've rebuilt your buying stack around one platform's promises. The deadline is set by two things moving in parallel: Google's behavioral remedies getting written into an order, and Amazon's discovery phase forcing auction logs into the open. Both play out over the next year.

The Market Analyst Two consecutive federal findings that Google broke the law, zero structural breakups. For public ad-tech, that's the story that moves money. Magnite, PubMatic, Index, the independent SSPs all built a bull case partly on "Google gets cut down to size." Brinkema just told them relief comes as bid-data sharing and killing unified pricing rules, not a divested AdX. In plain terms: the courts will make Google share more and behave better, but Google keeps its 20% take rate it admitted it won't lower. The independents get a slightly fairer auction, not a market opening. Amazon's suit matters less to markets today because retail media is private, but it's the first crack in the "Amazon DSP is the transparency standard" story the buy side told itself.

The Skeptic Everyone's screaming "antitrust is dead." Slow down. A remedy that forces Google to hand real-time bid data to rival ad servers and lets publishers set different price floors per bidder is not nothing. Unified pricing rules were the mechanism that stopped publishers from favoring competing exchanges. Kill that, and a publisher can finally set a higher floor for AdX and a lower one for Magnite, steering volume on purpose. Whether it works depends entirely on Google's compliance being real. Google says it already deprecated first look and last look years ago, and the DOJ contests that account. Discount Google's own version. The order's teeth are in the audit, and nobody's seen the order yet.

The Operator The line that should stop you comes from a media buyer who told Ad Exchanger that Amazon's search auction logs are essentially unavailable. You can see your own wins. You cannot see losing bids. So the FTC's claim that winners paid near-full price 80% of the time by 2024, and Amazon's claim that winning bids fell 50%, can both sit on the table and you have no way to check either. That's the actual operating condition. You are spending real money into a box, and the referee needs a subpoena to look inside. Tim Vanderhook's "copying Google's playbook" comparison to reserve price optimization describes a phantom bidder pushing your clearing price up. If you run performance budgets on Amazon, you've been optimizing against a number you can't verify.

The Customer / End User Start with the silence. David Nuremberg of Intermedia Advertising said he was surprised how muted the public response was, and noted The Trade Desk gets hammered for far less. Sarah Caputo had held up Amazon DSP as a transparency model and hit an "et tu, Brute" moment. So why is the outrage private? Because advertisers can't leave. Amazon has the purchase data and the checkout, and no independent measurement to grade it against. That dependency is the whole problem the OpenAI story pokes at from the other side: agencies are genuinely excited about real-time purchase-intent signals from ChatGPT, but they're withholding real budget because there's no ROI measurement and no competitor benchmark. Same complaint, two vendors. Buyers want an auction they can inspect and a number they can check, and neither Amazon nor OpenAI gives them one.

The CFO Look at where the fees actually sit. Google keeps a 20% take on Ad Manager and told the court it has no plans to cut it, and the court let it keep both the take and the structure. That's the cost of the open-web supply chain, unchanged by a monopoly finding. Amazon's soft reserve, if the FTC is right, quietly moved advertisers from paying a penny over the runner-up to paying near their max. Tens of billions in excess spend is the allegation. Whatever the true figure, the direction is one way: the platforms that own the auction set the clearing price, and every behavioral remedy so far leaves them owning it. OpenAI's $1 billion run rate is the number to distrust here, because nobody will say whether the ad credits it's handing buyers are counted as revenue. If credits are in, the earned number is smaller than a billion.

Where the personas split

The real disagreement is whether the Google remedy has teeth. The Skeptic says deprecating unified pricing rules genuinely changes how publishers steer volume, and that's worth more than the headlines admit. The Market Analyst says a take rate that survives two monopoly findings tells you the economics don't move, so the independents got a better auction and not a better market. Both can be right: the plumbing gets fairer while the toll booth stays exactly where it is.

The second split is Amazon versus OpenAI as the same story. The Operator and the Customer see one problem, unauditable auctions and missing measurement, showing up at both vendors. The Market Analyst sees them as opposite trades: Amazon is a mature, entrenched auction now under legal pressure, and OpenAI is a $1 billion toy against Meta's $200 billion bar that hasn't earned trust or scale. The connective thread is that buyers are being asked to trust a number they cannot independently verify in both cases.

What this actually hinges on

Three beliefs. First, whether Google's behavioral remedies are audited by someone other than Google. If the order relies on Google's own reporting, treat it as cosmetic. Second, whether Amazon discovery forces auction logs into daylight. That's the one event that could turn "alleged" into a verifiable clearing-price mechanism and reprice how the whole buy side treats retail media. Third, whether independent measurement ever attaches to any of this, because without it, OpenAI's intent signals and Amazon's DSP are both faith-based buys.

The council leans one way: enforcement will make the auctions marginally more transparent and leave the economics with the platforms. Verify by watching who audits the Google remedy and whether Amazon's logs surface in discovery. De-risk by not rebuilding your buying stack around any single platform's unverified numbers, which is exactly what MCP servers now tempt you to do by putting one agent on top of everyone's plumbing. Unified interface, still-siloed platforms, same opacity underneath.

Prediction: When Judge Leonie Brinkema's final ad-tech remedy order lands, it will let Google keep its Ad Manager take rate at or near the current ~20% with no cap, and enforcement will rely on Google's own compliance reporting rather than an independent auditor with log-level access, by the first post-order compliance review in 2027.

Confidence: Medium. The judge already refused a breakup and leaned toward Google's own proposed fixes.

Why: Brinkema declined the DOJ's structural breakup partly to avoid a 14-to-15-year appeals fight, and the expected remedies (bid-data sharing, killing unified pricing rules) sit closer to Google's own proposal than the DOJ's. A judge who rejects divestiture to get faster relief is not going to bolt on a heavy independent-audit regime that invites the same appeals delay, and Google has every incentive to keep the referee inside its own reporting. The opposite outcome, a hard take-rate cap plus outside auditors with real log access, would require the court to reverse the exact caution that produced the no-breakup ruling in the first place.

Revisit by 2027-06-30: We're right if the entered order leaves Google's Ad Manager take rate uncapped and assigns compliance monitoring primarily to Google's own reporting. We're wrong if the order caps or forces reduction of the take rate, or installs an independent monitor with direct access to Google's auction logs.