# Cannes Flexes AI ROAS While Agents Circle Retail Media - Digital Ads & Retail Media Weekly - Week of June 28, 2026 > A Cannes-heavy week on the tape for the week of June 22-28, 2026: Meta touts a $4.13 Advantage+ ROAS while operators warn AI shopping agents could route around the retail-media discovery funnel that 90% of the money depends on. ## Digital Ads & Retail Media Weekly ### Week of June 28, 2026: Cannes Flexes AI ROAS While Agents Circle Retail Media --- The whole industry decamped to the French Riviera this week, and the tape reflects it: nearly every meaningful podcast hit came out of Cannes Lions 2026. The throughline is a single uncomfortable question. The walled gardens are putting up the best AI-driven return-on-ad-spend numbers they've ever shown, and at the very same conference, the smartest operators are warning that AI shopping agents could route around the exact ad inventory those numbers are built on. Bullish proof points and an existential bear case, shouted from the same stage. Let's get into it. ## TL;DR - **Meta's flex is real.** New Berkeley research cited by a Meta VP pegs Advantage+ ROAS at **$4.13 per dollar**, up 25%+ since 2022, with 8M advertisers now on its GenAI tools. The walled garden is widening its lead. - **Agentic commerce is the bear case nobody can size yet.** Amazon Ads (~$70B, 60-70% of all retail media) sits on a "discovery funnel where 90% of the money is made," and AI agents that transact directly could bypass it. - **Retail media is consolidating and shopping for reach.** Walmart bought streaming-ad platform Vibe for a reported $1.4B; Walmart Connect + Sam's Club are merging org charts. Land-grab mode. ## What's new **Meta turned Cannes into a product dump, and a ROAS victory lap.** On [The Digiday Podcast](http://url7324.matterfact.com/ls/click?upn=u001.idHmPrr2Geh7KYLAsTy7NkrIVb-2FgA4pmf2rMXQwGcOhqbKaQNYt-2BOM3uKRXS6GxdngpXncwwTGFLim8u5VDIXHdSHDT2NzuPpfHa0HBLNuFKekkd4cpb94qZcrInwoSEM24vZrzI5D3Q1N5vyonAEg-3D-3DfzX9_7mLGwmUci-2BLaXswv9WX1yTgqn3Wad-2FotHhzHgSNAZbX1q2pRF58etHRLoEOQIAGh4iNXiPQ79gWK4KUIFDZk2VgYRRAgE-2F3r-2F-2F2DBZHKSz7fPRn4UVO-2BUXdVpBwEwAnisEziUJw2Zpc0qcyHeAvRA4SovmODW9W5LUZS4B1CpR071UzpXNf6eqAazt8B0r0ZrNmXAEmlxOlT5XfKx6rPAw-3D-3D) (June 25), Meta VP of Global Business Group **Nicola Mendelsohn**, an operator, so weight it accordingly, cited Berkeley research showing Advantage+ delivers **$4.13 in return for every U.S. ad dollar, up over 25% since 2022**, with 8M advertisers using its generative-AI tools and 15 new products launched ("biggest Cannes ever"). The headline release: a creative workspace inside Ads Manager that stores campaign history and surfaces live performance so advertisers can "double down in the moment on what's working." This moves numbers: it's the clearest data point yet that Meta's AI flywheel is compounding ad efficiency, which underwrites pricing power even if impression growth slows. **The agentic-commerce bear thesis got its sharpest articulation.** On [TD Cowen Insights](http://url7324.matterfact.com/ls/click?upn=u001.idHmPrr2Geh7KYLAsTy7NkrIVb-2FgA4pmf2rMXQwGcOgWyqK3JTSd352g9jijm1VAmk-2BW01c4QQDoMpculsH6ugqr7KyUuAAyu8sbdRr5RRFWLW94iyvByL5thvxOFqyrgWc2JXyZaVNDcITGIpbM7w-3D-3DN8j1_7mLGwmUci-2BLaXswv9WX1yTgqn3Wad-2FotHhzHgSNAZbX1q2pRF58etHRLoEOQIAGh4iNXiPQ79gWK4KUIFDZk2eD0JT54dX1EmO0CQlk7EaO5wooNSLa0FxRe3CUo1b6HQHxD0-2B-2FFpYgPIt4EYMkifOOX6CvO16Komqk4y3vClyOhi6c0W648sG4WK2ZGO-2BvbrNOpr6dudA8CAYv07vtyeg-3D-3D) (June 22), longtime Amazon expert **Scot Wingo** (industry pundit) laid it out: Amazon Ads is ~**$70B and 60-70% of the entire retail media industry**, almost all earned in the sponsored-search "discovery funnel where 90% of this money is made." If answer-engine agents (ChatGPT, Google AI Mode, Perplexity, Copilot) complete purchases without ever loading a search results page, that inventory gets skipped. Even a "best case" handoff to a product detail page bypasses the highest-value slots. This is the single most thesis-relevant idea on the tape: it's an attack on the structural margin of the entire category. **ChatGPT formalized its ad ambitions.** On [The Watson Weekly](http://url7324.matterfact.com/ls/click?upn=u001.idHmPrr2Geh7KYLAsTy7NkrIVb-2FgA4pmf2rMXQwGcOgjOM1Jea47cf3o40qtaVlsLyz-2BJW6RkPgNbd7Q3HInC-2BKHGuaSgU8mMHG2QhOTK-2BH7r6w-2BVvWzlH6NTuEsN-2BAYAEXcLqT8RAlFMudqYjyGcg-3D-3D13mx_7mLGwmUci-2BLaXswv9WX1yTgqn3Wad-2FotHhzHgSNAZbX1q2pRF58etHRLoEOQIAGh4iNXiPQ79gWK4KUIFDZk2UgK3VNOy40PSonQvfukloFIx5ua1mE8xlPzGxZVY9c0Ekcm7pEMylWvyDNwQXNx-2BMA68jA8tEJKQRuEMd7bpx1YElgD-2BNmg2vMZJ6UXpmCVdctktbWILNYOmiMYUZT8Rg-3D-3D) (June 26), **Rick Watson** (ecommerce pundit) reported OpenAI's Cannes pitch: ads head Dave Dugan (poached from Meta) says ~**20% of ChatGPT queries carry direct commercial intent**, OpenAI is targeting a **$100B ad business by 2030**, and weekly actives are now 900M+. Watson's read: this is aimed squarely at Google, because both are intent platforms, and conversational intent may be "an order of magnitude" richer than keyword search. Caveats he flags are real: no trust-and-safety apparatus yet, and attribution from chat to checkout is unsolved. **Walmart bought its way into streaming ad-buying.** On [This Old Marketing](http://url7324.matterfact.com/ls/click?upn=u001.idHmPrr2Geh7KYLAsTy7NkrIVb-2FgA4pmf2rMXQwGcOhjoJr3xe7bL4At2Wnl8UJy6GqX6L6AWJRViuH-2Bb4s1GCSNJifuIXJpIkqM55TpWX2uVEZrBREeX-2FcrK1IzuK88S5BX84IWXDmIDfjeFygHPA-3D-3DcLq-_7mLGwmUci-2BLaXswv9WX1yTgqn3Wad-2FotHhzHgSNAZbX1q2pRF58etHRLoEOQIAGh4iNXiPQ79gWK4KUIFDZk2QO0QIUtMPzL1Bs-2Fw9RgDbg-2BjyLszeg9S8qcaiktktYtVI5HDiv86oJQ3VwNPCsI-2FJ-2Bdll4hVhKPheccgLRFh70JdAqF3fy9j7HU5-2F0ek7WTXWFwtMig89rr8PjjGIKHvA-3D-3D) (June 26), **Joe Pulizzi and Robert Rose** (commentators) covered Walmart's reported **$1.4B acquisition of Vibe.co**, a steep markup on Vibe's $410M valuation last September. Vibe makes buying streaming TV ads "as simple as Google or Meta." Pair it with Walmart's Vizio hardware and you have a vertically integrated CTV retail-media stack. Context they cite (eMarketer): Walmart's ad business is ~$8.23B worldwide vs. Meta $243B, Google $239B, Amazon $82B, "a drop in the bucket," but the strategic logic is sound. ## The debate **Bull:** AI tooling and retail media are *expanding the pie* and *entrenching the walled gardens*. Meta's $4.13 ROAS and AppLovin's 59% growth are not hype, they're advertisers voting with budgets because the AI targeting works. Retailers aren't passive victims of agents; they're embedding ads *inside* their own agents (Amazon's Rufus/Alexa for Shopping and Walmart's Sparky selling "sponsored prompts," per Wingo). The first-party-data moat gets deeper, not shallower. **Bear:** Retail media is saturating, CTV is fragmenting, and AI search is a wildcard that could gut the open web and pressure even the giants. If agents disintermediate the discovery funnel, Amazon's ~$70B and everyone's retail-media economics face a structural, not cyclical, threat. CTV inventory is moving off the open market (Fox/Roku), squeezing the programmatic middle. And a credible $100B ChatGPT ad business by 2030 is share that has to come from somewhere, most likely Google's search dollars. The honest answer: nobody on this week's tape claimed to know how big the agentic hit is. Wingo himself said **holiday 2026** is when we find out which formats survive. That's the catalyst window. ## Stocks in play **Meta (META).** *Bull:* $4.13 ROAS, 8M advertisers on GenAI tools, AI flywheel compounding efficiency, pricing power intact. *Bear:* agency-relationship friction ("black box" concerns) and an unproven monetization path on AI Glasses ("wait and see," per Mendelsohn). *Watch:* whether the new Ads Manager creative workspace lifts advertiser spend per account into Q3. **Amazon (AMZN).** *Bull:* ~$70B ad business, 60-70% category share, building ads into Rufus/Alexa and launching "ShopDirect" plus a "Buy for Me" agent; blocking ChatGPT/Google bots to defend its funnel. *Bear:* most exposed name to agentic disintermediation of sponsored search. *Watch:* holiday 2026 read on agentic ad-format adoption. **AppLovin (APP).** *Bull:* per [Chip Stock Investor](http://url7324.matterfact.com/ls/click?upn=u001.idHmPrr2Geh7KYLAsTy7NkrIVb-2FgA4pmf2rMXQwGcOh-2Bu8seyIOsGDO7arcUAoeSnQuTlBW9NErHqxWQB4LUUShy3d6r55v13xeKb6qVZHEAI-2FpLMuDtQzHHhld-2BjXuiM4WocQ0ZxlM7-2F1Z4T0H7bA-3D-3DLOdA_7mLGwmUci-2BLaXswv9WX1yTgqn3Wad-2FotHhzHgSNAZbX1q2pRF58etHRLoEOQIAGh4iNXiPQ79gWK4KUIFDZk2Y1jR8I-2Frdumq0Wk0TELh6-2BauhRiHvKbtzcvT2ZsKnitmKdxTdTkpXRLRQ6vdJyJ8SdNYOyE8SV1KptvRscHqWU13mhVRKOLWUaLUfrlpD5y2v4OlwyJY11yvsH-2FpNYsNQ-3D-3D) (June 27), Q1 2026 revenue +59% YoY, 70% FCF margin, Q2 guide +54% revenue / +60% adj. EBITDA; AXON 2.0 trained on NVIDIA L4 via Google Cloud; ~$8B 2026 revenue with a $70B+ long-term target. *Bear:* that target *requires* cracking e-commerce and CTV, and CTV attribution is unsolved (no QR codes on a TV). Hosts peg fair value near ~$464 on ~18% 5-yr growth. *Watch:* traction in the consumer/e-commerce vertical and any hard CTV attribution proof. **Walmart (WMT).** *Bull:* Vibe + Vizio = integrated CTV retail-media stack; aggressive consolidation. *Bear:* still a "drop in the bucket" at ~$8.23B vs. the giants; integration risk. *Watch:* whether Vibe converts streaming inventory into measurable Walmart-attributed sales. ## Read-throughs - **Emerging ad platforms (RDDT, PINS, SNAP):** no podcast coverage this week, no commentary on their ad revenue, CPMs, or formats. - **Retail-media operators (WMT, CART):** Walmart was the week's busiest name (above). **Instacart** got only a passing mention in the TD Cowen episode as one of many Amazon-template networks ("Uber, DoorDash, Instacart... they all have it"), no standalone detail. - **CTV/streaming:** On [Marketecture](http://url7324.matterfact.com/ls/click?upn=u001.idHmPrr2Geh7KYLAsTy7NkrIVb-2FgA4pmf2rMXQwGcOhRJnlz4oUI3KAw2hduIFSOqMgLHDzD7oX-2F0NZXQGTL93ICryLAuoogfAGdJk4xClCJhiaqjMClnBW25N72TvBICvqeXyS2AgJKxEdZQJ-2BGGA-3D-3DCFN0_7mLGwmUci-2BLaXswv9WX1yTgqn3Wad-2FotHhzHgSNAZbX1q2pRF58etHRLoEOQIAGh4iNXiPQ79gWK4KUIFDZk2bK3Xb78TWmIOxQswuPoOemDh82aK9XUxnPUZBGoJK6z9p-2BeM-2BnpNyNrgojJldrXC4GSXfFvSJPGjYFJ7UWAq1BGx9pL0UYBemJCGXLmwu3AwdFtin6RXWYsx4Bfy1vmVQ-3D-3D) (June 24), Tatari CRO **Andy Schonfeld** (operator) argued CTV is supply-constrained and data-sparse, a bad fit for DSP architecture, and launched "Upstream," a direct-buy product wired into Paramount, Warner, Fox, NBC and Disney. He read the Fox/Roku deal as "a major change for the ecosystem," signaling premium CTV inventory moving *off* the open market, and the Publicis/Trade Desk spat as agencies starting to audit programmatic-TV fees. Read-through: pressure on open-market SSP/DSP economics. - **Ad-measurement vendors (DV, IAS, RAMP, CRTO, MGNI, PUBM):** no coverage. No episodes on cookieless tracking, identity, UID2, or verification this week. - **The Trade Desk (TTD), Kokai, UID2:** no direct coverage, TTD surfaced only secondhand via the Publicis spat. **Google/Alphabet's** ad business also went uncovered on its own merits this week. ## What changed vs. last week Narrow but dense. This was a **Cannes week**: high-signal coverage concentrated on Meta, Amazon/agentic commerce, Walmart, AppLovin, CTV and OpenAI's ad push. The flip side is real gaps, no coverage of Reddit, Pinterest, Snap, the ad-measurement complex, The Trade Desk directly, Google's ad fundamentals, or CPM/macro ad-spend trends. If those names matter to your book, this week gave you nothing new on them.