Newsletter · · Ashutosh Agarwal

The Trade Desk Hits a Seven Year Low as Amazon Opens a New Ad Surface - Digital Ads & Retail Media Weekly - Week of August 16, 2026

Digital advertising and retail media podcast coverage for the week of August 16, 2026. The Trade Desk guided to an outright revenue decline and sits near a seven-year low, Amazon started selling ads inside its Alexa AI assistant, Nielsen agreed to buy DoubleVerify for $2.15 billion, and agentic ad buying moved into live campaigns.

Digital Ads & Retail Media Weekly

Week of August 16, 2026: The Trade Desk Hits a Seven Year Low as Amazon Opens a New Ad Surface

Week of August 9-16, 2026

TL;DR

  • The Trade Desk is the story of the week. The independent ad-buying platform reported growth of just 3% (down from 19% a year ago), guided to an outright revenue decline next quarter, and the stock sits near $13, a seven-year low, down roughly 90% from its late-2024 peak. On The 7investing Podcast, the diagnosis was blunt: a fee fight with one of its biggest agency clients, a new Amazon competitor undercutting its prices, and a take rate that looks too high to defend.
  • Amazon found a new place to sell ads. On its Q2 call, advertising grew 26% to $19.8 billion, and the company rolled out "sponsored prompts," paid suggestions served inside its Alexa AI shopping assistant. Amazon says shoppers who tap them spend 21% more on average.
  • Meta's spending is the debate that won't die. On The Synopsis, two analysts pulled apart why revenue grew 27% last quarter but profit grew only 9%, and argued most of Meta's giant AI bill has little to do with the ad business that pays for it.
  • A big ad-tech deal: Nielsen is buying ad-verification company DoubleVerify for $2.15 billion.
  • "Agentic" ad buying went from slideware to live campaigns. Magnite's head of the effort walked through real deals where software agents negotiate ad placements with each other.
  • Reddit got a reality check. Its stock fell ~20% after earnings on worries about US users, even as analysts on The Investor's Podcast marveled at how profitable and under-monetized it still is.

What's new

Amazon quietly opened a new ad surface, inside Alexa. This was the most concrete new product of the week. On The Watson Weekly (Aug 14), Rick Watson and Jessica Lesesky walked through Amazon's Q2 2026 results: total sales of $200.6 billion (up 20%), and advertising, which Watson called "the fastest growing business," up 26% to $19.8 billion. The new twist is "sponsored prompts": when you talk to Amazon's Alexa AI shopping assistant, advertisers can now pay to have their suggested prompts served up to you. Amazon has even patented the personalization engine that decides which ones you see. In plain terms, the little "you might also want to ask…" nudges are becoming ad slots. Lesesky's key data point: Amazon says shoppers who engage with these prompts "are spending 21% more on average than those who don't." The hosts' verdict was equal parts admiration and alarm. Watson called it "an evil genius and silly idea at the same time," and framed it as Amazon monetizing "the last surface that was supposed to be organic."

Amazon didn't stop there. On Selling on Giants (Aug 11), Will Haire flagged that Amazon is now letting sellers extend their sponsored-product campaigns into creator content through the Amazon Influencer Program, pushing ads earlier in the shopping journey, beyond the search bar. And on Ecommerce Braintrust (Aug 11), Damiano described a new Amazon "Omnichannel Metrics" report that measures a brand's sales off Amazon, at retailers like Walmart and Target, by combining Amazon's ad-buying tool with first-party shopper-panel data. The theme across all three: Amazon is turning its retail-media business into something that reaches well past its own store.

Nielsen is buying DoubleVerify for $2.15 billion. Reported on Sounds Profitable (Aug 14): Nielsen (the audience-measurement company) is acquiring DoubleVerify, which checks whether ads were actually seen by real humans in brand-safe places, in an all-cash deal at $13.60 a share, about a 30% premium. It's expected to close in the first quarter of 2027. Why it matters: the combined company would have more than $4 billion in annual revenue and clients responsible for over $300 billion in ad spend, spanning TV, streaming, digital, social, mobile, audio, and AI ad platforms. As Nielsen CEO Karthik Rao put it, "joining forces with DoubleVerify will extend our capabilities deeper into the digital media industry." It's a sign that "did this ad actually work, and was it real?" is becoming its own consolidating industry, especially as AI-generated and bot traffic muddy the waters.

"Agentic" advertising stopped being a buzzword. On Open Market (Aug 12), Magnite's Jonathan Moffie, whose actual title is "Head of Agentic," described live campaigns where a buyer's software agent negotiates directly with a seller's software agent to place ads, cutting out steps that used to take "weeks of emailing back and forth." The first big use case is connected-TV deals. His proof point: one of Magnite's first agent-run campaigns delivered "99% completion rates" on its direct deals, and agentic is now the company's fastest-growing pipeline. The bigger claim worth watching: this shift may favor the companies that sit closest to the ad supply (like Magnite, the largest seller-side platform) rather than the buying tools, a potentially important wrinkle in a debate we'll come back to.

The debate

Is Meta's AI spending genius or a repeat of the metaverse mistake? This was the richest discussion of the week, on The Synopsis (Aug 13). Here's the setup in plain numbers: for years Meta showed enormous "operating leverage," spend a little more, earn a lot more. Now that's flipped. Last quarter, revenue grew 27% but earnings grew only 9%. Meanwhile capital spending has ballooned from about $35 billion in 2023 to a guided $130-145 billion this year.

The two analysts split the spending into two very different buckets:

  • Spending on the core ad business, better targeting, and AI tools that generate ad images and copy for advertisers, which they think is clearly worth it. One host got genuinely excited about the idea that Meta could let a small business "just chat" with it to run a whole campaign, cutting out the ad agencies and freelance creators that today stand between merchants and their ads: "there are thousands of digital ad agencies and billions of dollars in jobs all about ad creation… if Meta can take out all these layers, it seems like there's an investment to be made there."
  • Spending on everything else, a frontier AI model Meta wants to own so it isn't "beholden" to OpenAI or Anthropic, plus a would-be cloud business, a "business agent" platform, and reselling raw computing power. This, they argued, is where the market gets nervous, because "a lot of this AI spending is not about the core business." One host drew the uncomfortable parallel: this is "eerily similar" to the $100 billion-plus Meta poured into the metaverse (Reality Labs) with little to show for it.

The skeptical gut-check line: "I don't know a single person who's using the Meta AI. It's ChatGPT, it's Claude." The counter: the stock trades around 20x earnings (about 17x if you strip out Reality Labs), so investors are paying almost nothing for any of these bets to work, you're getting "optionality" for free. The honest conclusion from both hosts: Meta hasn't clearly explained the plan, "and I also know they probably themselves aren't entirely sure."

Who wins when software agents do the ad buying? The quieter but important debate, running through the Open Market conversation: as agents take over campaign setup and negotiation, do the buy-side tools (demand-side platforms) get squeezed, while the supply-side players who "own the end-to-end transaction" gain? Magnite's view, admittedly self-interested, is that the sell side is "uniquely suited" to this wave. It's a lens worth keeping on The Trade Desk's troubles below.

Stocks in play

The Trade Desk (TTD), a former darling near a 7-year low. On The 7investing Podcast (Aug 14), Simon Erickson devoted a full segment to what he called an "ugly" quarter, and it's worth spelling out because the deterioration is steep:

  • Revenue growth fell from 19% a year ago to just 3% last quarter. Net profit margin slipped from 13% to 9%; the adjusted EBITDA margin from 39% to 34%.
  • The stock is back to about $13, a seven-year low, a ~$6 billion company, down roughly 90% from over $130 at its late-2024 peak.
  • After beating Wall Street's numbers for 32 quarters in a row, it has now missed two quarters running, and guided to an actual decline in revenue next quarter, something Erickson said he'd essentially never seen from them.

Three things are going wrong at once, per Erickson. First, a fee dispute: French ad giant Publicis audited the fees The Trade Desk was charging, especially for AI features, and pushed back. It was settled quietly in the past month, but it puts a spotlight on the company's roughly 20% "take rate" (its cut of ad spend that flows through the platform). Erickson notes travel and hotel booking platforms take closer to 13-15%, and he thinks TTD's cut "is going to be coming down." Second, Amazon is now a direct competitor, opening its own ad-buying platform to the wider internet at a take rate of "10 or 11%… as opposed to the 20% plus," "the most terrifying thing you can possibly hear if you're the CEO of a smaller company." Third, a loss of credibility: two misses in a row from a CEO (Jeff Green) who "used to always beat and raise."

The response: Green hired three new senior commercial executives in July, many from Google, to relentlessly prove the platform's return on investment to big clients (Erickson likened it to the playbook Splunk once ran). Client retention is still above 95%. The bull-bear framing was refreshingly honest: if growth resumes to the mid-teens, Erickson could see a path to ~$40 by 2027-28 (a triple); if it stays stuck in low-single-digit or negative growth, the stock "actually feels pretty appropriately priced right now." A telling aside: the "death of the cookie / privacy identifier" story that made The Trade Desk so exciting in 2024 now "feels like ancient history," the whole conversation has moved to a single question, what's my return on ad spend?

Meta (META). See The debate above, the whole discussion is a referendum on whether the ad-funded core can carry a $130-145 billion spending year.

Amazon (AMZN). Ads up 26% to $19.8 billion and a brand-new ad surface (sponsored prompts), plus a growing threat to The Trade Desk. Notably, on the same call Amazon confirmed it is stepping back from building its own frontier AI model, leaning instead on partners like Anthropic, a very different posture from Meta.

Reddit (RDDT), punished despite great numbers. On The Investor's Podcast (Aug 16), Daniel Mahncke and Shawn O'Malley dug into Reddit's just-released Q2. The stock fell about 20% after earnings, even though the business hit a 30% operating margin by the end of last year, a level one host had modeled the company wouldn't reach until the end of the decade. Their point: Reddit is still deeply "under-monetized" with about 300 million users. So why the sell-off? A user problem: more than half of Reddit's daily visitors arrive through a Google search rather than the app, and management said it will stop reporting logged-in user numbers, which the hosts flagged as a worrying "hide the ball" move, right as that metric may be turning against them. One analyst's grilling of CEO Steve Huffman on the call was memorable enough that O'Malley told listeners to "jump to the 31 minute and 30 second mark." Verdict: still a hold, not a sell, "Reddit has grown so fast that the valuation looks a whole lot more reasonable than it did."

Read-throughs

  • Alphabet / Google (GOOGL), the AI-search hinge. The most important cross-read of the week came inside the Reddit discussion on The Investor's Podcast. Google now has "a billion Gemini users," and click-through rates from search to the original source websites "plunged by roughly 60%" as AI overviews answer questions directly. Crucially, the hosts noted Google is now saying it can monetize those AI answers "in the same way" as classic search, easing a long-standing fear. The pain, instead, lands on everyone downstream who depended on Google's referral traffic. Also relevant: Google pays Reddit just $60 million a year to license its data for AI, a figure the analysts called "shockingly low" for a corpus that supplies an estimated 5-10% of raw LLM training data and up to 40% of real-time citations in AI answer engines. Reddit is reportedly debating not renewing.
  • Connected TV / streaming. On Chit Chat Stocks (Aug 12), Alex Morris discussed the economics of ad-supported streaming tiers across Disney, Netflix, and Amazon Prime Video, the supply side of the CTV ad boom that platforms like The Trade Desk and Magnite are fighting to sell.
  • Pinterest (PINS). A single useful data point from Stop Scrolling, Start Scaling (Aug 12): Pinterest shared that pins whose titles use brand keywords delivered 28% higher return on ad spend, reinforcing the case that Pinterest behaves more like a search engine than a social feed. The same episode pegged Reddit at about 130 million daily users.
  • Ad-tech M&A and measurement. The Nielsen-DoubleVerify deal (above) is the marquee transaction; Sounds Profitable also noted Acast buying creator network Backyard Ventures for $20 million. Verification and measurement are consolidating.

What changed vs last week

  • The Trade Desk went from "struggling" to "crisis of confidence." The multi-quarter slide became a second straight miss plus negative forward guidance, and, for the first time, the bear case (a structurally lower take rate and a serious Amazon competitor) is being taken seriously rather than dismissed as macro noise.
  • The open-internet thesis lost its old anchor. The "death of the cookie / universal ID" narrative that framed the independent ad-tech bull case in 2024 has faded; the debate has fully shifted to return on ad spend and to a new front, software agents doing the buying and selling.
  • Retail media's frontier moved from search to conversation. Last quarter the story was sponsored search and off-site measurement; this week it's ads inside AI assistants (Amazon's sponsored prompts), a genuinely new surface.
  • Measurement is consolidating. A $2.15 billion Nielsen-DoubleVerify deal puts a marquee price tag on the "prove the ad was real and worked" layer.