Newsletter · · Ashutosh Agarwal

Google Keeps Its Ad Machine as the FTC Turns to Amazon - Digital Ads & Retail Media Weekly - Week of September 6, 2026

Digital Ads & Retail Media Weekly for the week of September 6, 2026: a judge found Google an illegal monopolist in ad tech and then declined to break it up, the FTC accused Amazon of rigging its ad auctions to the tune of 20 billion dollars, Walmart Connect grew 43 percent, and AppLovin got the full investor-podcast treatment.

Digital Ads & Retail Media Weekly

Week of September 6, 2026: Google Keeps Its Ad Machine as the FTC Turns to Amazon

TL;DR

  • The big one: A federal judge ruled that Google is an illegal monopolist in ad technology, and then declined to break it up. No forced sale of its ad exchange, just behavioral fixes that are, for now, secret. The market shrugged, because the business in question is small and shrinking.
  • The quieter, scarier one for Google: In the same breath, journalists and a former top antitrust enforcer kept pointing at the real threat, AI chatbots eating search, and OpenAI reportedly on track for $1 billion a year in ad revenue. The case was fought over a 2010s internet that barely exists anymore.
  • Amazon's turn: The FTC says Amazon quietly rigged its ad auctions to overcharge advertisers by about $20 billion. Amazon says the critics don't understand how ads work.
  • Retail media keeps compounding: Walmart's ad arm, Walmart Connect, grew 43% in the US last quarter, far faster than the stores themselves.
  • AppLovin got the full investor-podcast treatment: an "algorithm with 900 employees" throwing off $7 million of revenue per head and roughly 80% profit margins, with the bulls and skeptics both louder than usual now that growth is cooling from 70% to a guided 30%.
  • Ads inside ChatGPT are real, and they're rough: practitioners testing them say it feels like "Google ads 15 years ago."

What's new

Google was found guilty and kept everything anyway

This was the week's headline, and it was covered everywhere from tech shows to daily market recaps. Here's the plain version.

For years the US Department of Justice (DOJ) has argued that Google rigged the plumbing of online advertising. To see why, picture what happens in the fraction of a second when a web page loads: there's an instant auction where the site sells its empty ad space and advertisers bid to fill it. Google owns the software the publisher uses to sell the space (Google Ad Manager, formerly DFP), it owns the exchange where the auction happens (called AdX), and it owns the tools advertisers use to buy. As the hosts of the Elon Musk Podcast put it, Google is "sitting on every single side of the table" and can effectively peek at everyone's cards.

A court had already found that setup to be an illegal monopoly. This week Judge Leonie Brinkema, in the Eastern District of Virginia, decided the punishment. The DOJ wanted to tear the machine apart: force Google to sell AdX, and publish the secret math that decides who wins each auction. She said no to all of it. Instead she ordered "behavioral remedies", so Google keeps everything but has to change how it runs the auctions to let rival systems plug in more easily.

Two things make this strange. First, as the Elon Musk Podcast hosts noted, the exact new rules are sealed and unpublished for now, so smaller competitors are being told the market is fair again without being shown the new rulebook. Second, forcing Google to open-source its auction logic would have been "the equivalent of forcing a casino to publish" its odds, and the judge refused to do it.

Why the market didn't care. On The Rundown, Zaid Admani of Public.com laid out the numbers that explain the yawn: this ad-exchange business made about $30 billion last year, but that's "only 8% of Alphabet's revenue, and accounts for less than 1% of the company's profits." And it "has been shrinking for 16 straight quarters." Google takes roughly a 20% cut on each sale, but the whole line is fading. On Brew Markets, Dylan Lewis framed the broader pattern: "regulators can prove a company has behaved illegally, a judge can force it to change how parts of its business operate, but courts seem very reluctant to go that final step" and actually break the company up. It's now the second year running that Google has walked away from a threatened breakup: last year it was the search case and a possible Chrome spin-off, as Joanna Stern noted on Marketplace. Reuters called it "the third straight judicial block" of a big-tech antitrust action (Reuters World News).

The FTC comes for Amazon's ad auction

While one big-tech ad case was winding down, another was starting. On Motley Fool Hidden Gems Investing, the team walked through the Federal Trade Commission's new complaint against Amazon: the agency says Amazon manipulated its own advertising auctions and overcharged advertisers by roughly $20 billion.

The specific accusation is neat. For a decade Amazon has said that whoever wins an ad auction on its site pays "roughly a penny more than the next highest bidder", the textbook definition of a fair auction. The FTC says that quietly stopped being true "as early as 2018" for one ad format and soon after for others, because Amazon secretly inserted its own bids to push prices up. Former DOJ antitrust chief Jonathan Kanter, on Prof G Markets, described it as Amazon planting "proxy bids" in its search-ad auction, alongside 22 state attorneys general, with the harm falling on small sellers, right down to people advertising their self-published e-books.

Amazon's response, per the Motley Fool crew, was combative: the FTC "doesn't understand how advertising works." Amazon's counter-story is that it started favoring relevant ads over the highest bidder around 2019, which it says cut the average winning bid by about 50% and saved advertisers $8 billion. Whether both things can be true, cheaper winning bids and $20 billion of overcharging, is what the case will decide.

The investing nuance worth holding onto: as the hosts noted, Amazon's retail business runs at roughly break-even, and it's the advertising business that actually makes the retail segment profitable. So a case that reshapes how Amazon runs ads isn't a sideshow to the retail story. For Amazon, ads are the profit story. (The panel's own view: the $20 billion is "kind of a speeding ticket"; the case with real teeth is a separate 2023 suit about Amazon allegedly stopping sellers from offering lower prices elsewhere, which goes to trial next March.)

The debate

Was this ruling a win for Google, or a warning shot at the whole idea of ad-tech antitrust?

The most thoughtful take of the week came from Catherine Perloff of The Information on TITV. Her argument, in plain terms: the government may have won the battle on paper but is fighting the last war. "A lot of the evidence from this case is from the 2010s," she said, a time when advertising on the open web (think banner ads on a newspaper site) mattered a lot more than it does today.

The judge, Perloff noted, seemed genuinely worried that forcing big changes could further hurt publishers who are already being squeezed, not by ad tech, but by Google's own AI Overviews (the AI-written answers that sit atop search results and keep users from ever clicking through to websites). And the forward-looking punchline: "OpenAI [is] saying we're on track to hit a billion dollars in ARR for our ad business." So the real question, she argued, is "who's developing a monopoly potentially in this landscape", not the one from a decade ago. Google's ad tech, she added, was never as dominant in apps and streaming video as it was on the open web anyway.

Put those together and you get the debate in one line: the government spent years proving Google monopolized a market that AI is now busy shrinking on its own. The bear case on Google was never really the ad exchange. It's that ChatGPT, Gemini and friends erode the search-ad cash machine that actually matters. The ruling is a reminder that the courts move at the speed of the 2010s while the ad market moves at the speed of AI.

Stocks in play

AppLovin (APP), the "algorithm with employees"

No single company got more investor-podcast attention this week than AppLovin, thanks to back-to-back deep dives from The Investor's Podcast (TIP843), its sister show The Intrinsic Value Podcast (TIVP095), and The Canadian Investor.

What AppLovin does: its AXON engine uses AI to place ads inside mobile games (Candy Crush, Solitaire, that genre), figuring out which players are most likely to download the game being advertised, and pocketing the difference between what advertisers pay and what it costs to reach those players. Because AppLovin also owns the software that runs the ad auctions inside those games, it sees all the data, a structural edge.

The numbers the hosts kept marveling at:

  • Roughly 80–85% profit margins (EBITDA margin, profit before interest, tax and accounting depreciation).
  • Around $7 million of revenue for every one of its ~900 employees. For scale, The Canadian Investor put Meta at about $2.8 million per employee, Alphabet ~$2.2 million, and The Trade Desk far lower. Hence the nickname: "an algorithm with employees."
  • Almost no capital spending: one host found just $1.4 million of spend on property and equipment against $869 million of operating cash flow.
  • AXON's mediation platform, MAX, reaches roughly 1.4 billion daily users across 140,000+ apps, and AppLovin's ad spend alone is "more than Pinterest, Snapchat and Reddit's combined revenue."

The bull-vs-bear tension that made these episodes worth listening to:

  • Bull: the stock's earnings multiple has compressed from ~35x forward earnings to ~16x, even though it's still growing 30%+, "dirt cheap on the surface."
  • Bear: growth is decelerating, guided to about 30% in 2027, down from 50–70%+ historically, which the hosts called "inevitable" once you already control more than half of a market. AppLovin has ~55% of ad spend in top-grossing games; competitors sit at 25% and 13%. And the threats are Meta, Google, Liftoff and Unity's LevelPlay, which makes the "moat" debatable.
  • The real overhang: a wave of short-seller reports last year accused the company of everything from fake app downloads to violating Apple's, Android's and Meta's data rules. The Canadian Investor host was blunt that the ToS-violation risk is the one that actually scares him: if Apple or Meta "punted" AppLovin off their platforms, the data edge could vanish. Worth noting on the other side: an SEC investigation into the company was closed "with zero action" a few weeks ago.

Walmart (WMT), Walmart Connect is the story now

Walmart's quarter, dissected on The Watson Weekly, is really a retail-media story wearing a retailer's clothes. "Retail media" means the ad business a retailer runs on its own site and app, Walmart selling ads to the brands whose products it stocks. Walmart Connect is that business.

  • Walmart Connect US grew 43% last quarter (excluding Vizio, the TV maker Walmart owns), and global advertising grew 38%.
  • Membership fees +17%, US marketplace sales +50%+, e-commerce now 23% of US sales.
  • The catch host Rick Watson flagged: reported operating income jumped 28.8%, but about 750 basis points (7.5 percentage points) of that came from one-off tariff refunds. Strip those out and it's closer to 10%. Core store sales actually slowed to their weakest since 2020.

In other words, the flashy headline profit was juiced by a refund, but the durable engine underneath, ads, membership and marketplace, keeps "compounding faster than the overall company." On Watson Weekly's store-ranking episode, the hosts were unusually bullish: Walmart's ad marketplace "has crushed it… they're printing money," and Walmart "is not just a retailer anymore." A small competitive tell in the same episode: Kroger just hired Nate Faust, from the original Jet.com team, to run e-commerce, a sign rivals are trying to copy the playbook.

Amazon (AMZN)

Covered above under the FTC case. The one-line investor takeaway from the podcasts: watch the 2023 pricing-antitrust trial starting next March far more than the $20 billion ad-auction headline, and remember that ads are what make Amazon's retail segment profitable in the first place.

Alphabet / Google (GOOGL)

Also above. The ad-tech ruling barely moved the stock because the exchange business is small and fading. The thing that actually matters for Google, whether AI search erodes its core search ads, was framed this week through the antitrust lens rather than through platform-level ad commentary.

Read-throughs

Ads inside AI chatbots are here, and they're primitive. Two very different practitioners tested ChatGPT's new ads this week. On Inside Commerce, agency operator Josh Duggan (whose clients include AllSaints, Manchester United and Mulberry) compared ChatGPT ads to "Google ads 15 years ago… very, very basic," with high costs and irrelevant targeting: he asked about UK property trends and got ads that had nothing to do with it. On REVB, a real-estate marketer showed how they actually work: you sign up at ads.openai.com, and your ad appears as a small card marked "ad" at the end of a chatbot answer. Crucially, more than 90% of weekly ChatGPT users are on the free tier, and that's who sees the ads. Both takeaways matter for the Google debate: the format that's supposed to threaten search ads exists, but it's nowhere near ready to replace the money machine yet.

The cheapest big ad platform right now is YouTube. Duggan's most quotable line: "You'll never see an ad platform as cost effective again as what YouTube is now… by far the cheapest platform. It could be 10 times less effective than your other channels and it would still essentially break even." A useful signal for anyone modeling Google's video monetization.

Reddit's AI blessing is also its advertising curse. On WSJ's The Journal, reporter Katie Deighton explained why chatbots love Reddit: it's a giant pile of honest, human, crowd-sourced answers, so Gemini and Claude scrape it constantly for recommendations. That's made "brands suddenly care about Reddit after… decades" of ignoring it. But the tension is real: "advertising is the thing that maybe threatens what is Reddit at its core, which is honesty, real people, and no manipulation." She called it "the battle over the soul of Reddit." Separately, on Do This, NOT That, Jay Schwedelson flagged a concrete product change: Reddit is shutting off its on-site lead-generation ad forms at the end of September, forcing advertisers to send people off-platform, which usually hurts conversion, while X is going the other way and bringing those forms back.

Connected TV still runs on too many middlemen. On Marketecture, Cadent CEO Nick Troiano argued that advertisers juggling "10 or 15 vendors" across linear TV, connected TV (streaming) and YouTube need one connected layer of intelligence, planning, buying and measurement that actually talk to each other, even if they keep multiple partners. His swipe at the walled gardens: YouTube, which Cadent only started working with last December, "like a lot of the walled gardens… is still closed." His core complaint is one every CTV bull should track: measurement that takes "one week, two weeks, three weeks to get back" makes real-time optimization impossible.

Meta advertisers say the game is now data, not demographics. Two operators offered a useful ground-level read. On Sub Club, a marketer described scaling Meta spend from zero to $100,000 a month profitably, with better creative, not clever targeting, doing the heavy lifting. On Brands, Beats & Bytes, agency founder Sheldon Poon explained the mechanical shift: old-school demographic targeting now hurts you on Meta, so you go deliberately broad and let Meta's algorithm, drawing on "52,000–54,000 data points" per user, find your buyers by watching who reacts. It's a small window into why Meta's AI ad tools keep gaining share of budgets.

Holiday ad costs are climbing again. On Limited Supply, guest Jen Gray captured the Q4 setup for every brand that buys ads: "attention has become commoditized… the rarest commodity," and customer-acquisition costs are rising again into Black Friday and Cyber Monday. Read that as continued upward pressure on ad prices (CPMs, the cost to show an ad to a thousand people) heading into the fourth quarter, which is good for the platforms selling the ads and painful for the brands buying them.

What changed vs last week

  • The long-running Google ad-tech case finally has an answer. For over a year the open question hanging over Alphabet was whether it would be forced to sell part of its ads business. That shoe has now dropped, and it dropped softly. One overhang removed; the multiple didn't need it, which is itself telling.
  • The antitrust spotlight rotated from Google to Amazon. A week ago Amazon's advertising practices were a background worry; now they're a formal FTC complaint with a $20 billion number attached.
  • AI-search advertising moved from "coming soon" to "you can test it today." The story shifted from speculation about whether ChatGPT will run ads to hands-on reviews of the (rough) ads it's already running.
  • Retail media's momentum held. Walmart Connect's 43% growth keeps the sector's core narrative, retailers' ad businesses compounding faster than their stores, firmly intact.