Newsletter · · Ashutosh Agarwal

Alphabet Ad Growth Holds as Retail Media Turns to Incrementality at Cannes - Digital Ads & Retail Media Weekly - Week of July 20–26, 2026

Alphabet's Q2 ad engine stayed healthy (Search up 17%, YouTube up 13%) even as a raised $195–205 billion capex guide sank the stock, while Cannes Lions marked retail media's shift from a growth story to an incrementality reckoning, for the week of July 20–26, 2026.

Digital Ads & Retail Media Weekly

Week of July 20–26, 2026: Alphabet Ad Growth Holds as Retail Media Turns to Incrementality at Cannes

TL;DR

  • Alphabet's Q2 print was the whole week. Total revenue grew 24% year over year, a third straight quarter above $100 billion. But the number everyone fixated on was Search: up 17%, yet a hair light versus expectations ($63.3 billion vs. a $63.4 billion estimate). Cloud grew a staggering 82%. The stock still fell, roughly 2.5–2.8% after hours, then about 7% the next day, because the company raised its 2026 capital-spending guide to $195–205 billion and its Gemini Pro 3.5 model slipped. The ad business is fine; the market got scared of the bill for AI.
  • Retail media "grew up" this week. From Cannes Lions and a Criteo interview, the same message kept coming: the roughly $75 billion retail-media market (Amazon is about two-thirds of it) is shifting from a land-grab to a reckoning on whether the ads actually cause incremental sales. As Criteo's retail-media president put it, incrementality has moved "from the measurement team to the boardroom."
  • Instacart went on an ad-product offensive at Cannes, an AI shopping assistant, shoppable vertical-video feeds, and an in-house creative agency it's now selling to brands, while deliberately not putting ads inside its AI chat experience yet.
  • The measurement debate is the through-line of the whole week. Practitioners hammered the gap between "platform-reported" returns and true incremental returns, whether the platform is Google, AppLovin, Spotify audio, or Amazon.

What's new

Alphabet's earnings dominated, and the ad story was better than the tape suggested

Alphabet reported its second-quarter 2026 results on July 22, and the podcasts spent more airtime on it than on everything else combined.

Start with the plumbing. On Wall Street Unplugged (July 23), the host walked through the segment lines: overall services up 15%, Google Search up 17% year over year, and YouTube ads up 13%. On CNBC's "Fast Money" (July 22), the desk pinned down exactly why the "Search miss" narrative took hold, and how small the miss actually was: "search revenue, $63.3 billion versus $63.4 billion estimates. So it's just a little miss there." Guy Adami's framing was the fairest read of the week:

"The search was disappointing, but I don't think it was catastrophic. And remember, it was a year and a half or so ago where people were talking about the existential risk that AI presented to Google in terms of the search business. So it's not falling off a cliff."

The bigger story was Cloud and the capital it takes to feed it. Fast Money reported Google Cloud grew about 82% year over year at a 35.5% operating margin, genuinely rare for a business at that scale. On Bloomberg Intelligence (July 22), Mandeep Singh, the firm's head of global technology, called it "a $100 billion run rate business now, which is phenomenal." He also flagged the one soft spot the bulls glossed over, the cloud backlog, or contracted-but-not-yet-recognized revenue, came in below the "whisper number," which is odd for a business growing 82% with Anthropic as a marquee customer.

So why did the stock fall on a 24% revenue quarter? Two words that came up over and over: the spend. Fast Money reported the new 2026 capital-expenditure guide of $195–205 billion, roughly $15 billion higher at the midpoint than prior expectations, with Gemini Pro 3.5 delayed on top of it. Karen Finerman's take: "I think that's why it's down." And the read-through she drew is the one ad investors should sit with: "this isn't great for Meta, though. The spend... if they are in the arms race... everybody else is pushed to raise as well."

Wall Street Unplugged added the cash-flow detail that explains the anxiety: operating cash flow of $39.1 billion in the quarter ($185.7 billion over the trailing twelve months), capital expenditure of $44.9 billion in the quarter (roughly 60% technical infrastructure, 40% data centers), enough to push free cash flow to negative $5.9 billion for the quarter. The counterweight: about $242.5 billion in cash and marketable securities, and a balance sheet that took on debt fast (long-term debt near $98.2 billion, up from the mid-teens a year ago) plus an $80 billion equity raise, with Berkshire Hathaway buying $10 billion of it. On Bloomberg Intelligence, the panel put it bluntly: "believe it or not, this company will have negative free cash flow next year" if capex keeps climbing.

On the AI-usage bragging rights, 90% of the Fortune 100 on Gemini Enterprise, 22 billion API tokens processed per minute, and a Gemini app with 950 million monthly active users, tripled over the past year, Singh offered the most important piece of skepticism of the week. Those user numbers, he argued, mostly reflect Google's distribution, not genuine preference:

"The 950 million is a reflection of the high attach rate that Google has because of the distribution through search and the operating system and browser... all these companies are reporting very high MAU numbers, but it's really the usage that counts."

Why it matters for ads: the core investment question underneath the whole print is whether AI-driven search behavior slowly erodes Google's highest-margin cash cow. This quarter is not evidence that it is, Search grew 17%, but the "slight miss" is exactly the data point bears will keep poking at.

Retail media stopped asking "does it work?" and started asking "how well?"

The second big theme came from Cannes Lions and CPG-focused shows, and it's a genuine phase shift.

On The CPG View (July 21), Sherry Smith, President of Retail Media at Criteo, laid out the market in round numbers: a roughly $75 billion retail-media business, about two-thirds of which is Amazon, growing at a compound rate of "almost 18 to 20%", against brick-and-mortar retail growing 2–3% and e-commerce around 8% in North America. In plain English, "retail media" means the ad businesses retailers build on top of their own shopping traffic and purchase data (think sponsored products on Amazon or Walmart's site). It has been the fastest-growing corner of digital advertising for years.

Smith's central point is that the easy-money era is over:

"Brands are no longer asking whether retail media works. They're asking how well it works. And they're asking compared to what?... A strong ROAS number could move the conversation forward. It doesn't work anymore. Today, brands want a more complete picture. They want to know what was truly incremental."

("ROAS" is return on ad spend, revenue divided by ad dollars. "Incremental" means sales that would not have happened without the ad, the distinction that runs through this entire issue.) Her line of the week: "Incrementality has moved from the measurement team now to the boardroom." She also predicted consolidation, but on the infrastructure and measurement layer, not the retailer front-end: "the future is more about a connected ecosystem where differentiated retailers still have a place, but only if they can execute well." Her advice to smaller regional retail-media networks was to stop trying to look like Amazon and instead lean into loyalty and relevance: "brands are not just buying size. They're buying relevance."

Instacart's Cannes product blitz, and a deliberate pause on AI-chat ads

The single richest retail-media source of the week was The CPG Guys (July 22), a live Cannes interview with Instacart's Ali Miller (General Manager of Advertising) and Tim Castelli (VP of Global Advertising).

The scale numbers: Instacart's marketplace now spans 2,200 retailers, built on 14 years of operating history, 1.6 billion transactions, and over 100,000 storefronts. Its advertising pitch has moved well beyond its own app, it now uses its first-party purchase data to target and measure ads that run on TikTok, YouTube, Meta, The Trade Desk, Roku, and NBCUniversal, and it offers that ad technology to other retailers through its "CaratOut" partnerships (over 310 partners). Miller's framing: "taking this e-commerce data out of its own little corner."

The Cannes announcements:

  • An AI shopping assistant that is "inventory-aware, based on your local retailers," and learns from every transaction. Castelli cited a Harris poll that "3 out of 4 Americans are more stressed about... deciding what's for dinner than actually making dinner," the friction Instacart says the assistant solves.
  • Shoppable vertical-video feeds, an "Immersive Feed" of the snackable, TikTok-style recipe content people scroll, with an "add to cart" button attached.
  • "Ads to Views", Instacart is opening up its in-house creative agency, Local Produce (recently named AdAge's number-one in-house agency of the year), as a service to brands. It has already run about 400 campaigns for roughly 200 brands, seeing "80%, 90% new-to-brand impressions" when brands integrate into its seasonal moments.
  • Caper smart carts, the in-store, screen-equipped shopping carts have tripled deployments and are now in "100 cities across 15 states." A striking cross-channel stat: even a shopper who's exposed to an Instacart ad but doesn't buy shows "15% to 20% increase of buying those products in-store."

The most investor-relevant nuance was what Instacart is not doing. Asked about running ads inside its new AI chat/assistant experience, Miller was explicit that they're holding off on purpose:

"We're also really early in terms of how this is taking off... towards the second half of this year, we'll be building experimentation, we'll be looking at what resonates, and we'll be using that to develop our ad strategy in these new surfaces."

Castelli's blunt version: "if anybody is sitting here... [saying] they've got this figured out on the ad experience [in AI], I think they're lying." That "test-and-learn, let the consumer lead" posture is the honest read on where everyone is with ads inside AI assistants right now.

Costco's transparency-first retail-media build

A companion Cannes interview on The CPG Guys (July 25) featured Costco's Mark Williamson (AVP, Retail Media), and it's a useful contrast to the incumbents. Costco is a late entrant that deliberately built its stack on a "clean sheet of paper," running on a Google Cloud "unified data platform" that ties everything back to the member. Williamson's edge is physical signal most retailers don't have: "it starts with the card scan on entry and it ends with the card scan on exit... to my knowledge, we may be the only retailer that is scanning a card on the way in." His notable new move is pushing Costco's first-party data into Google Product Listing Ads (the shopping ads in Google search), betting that high-intent shoppers searching "washing machine" or "refrigerator" can be routed to Costco's conversion channel in a way a brand buying generic keywords alone can't match.

The debate: is the digital-ad pie still growing, and who captures the growth?

The cleanest framing of the central debate came from Motley Fool Hidden Gems Investing (July 20), where analysts Rachel and Matt tackled a listener question about whether Alphabet and Meta grew by taking share or by growing the whole market.

The scale-setting facts: Meta relies on advertising for about 98% of its revenue; Alphabet for about 70%; together they control roughly half of the entire global advertising market, at a combined market cap around $6 trillion. Rachel's argument is that AI has been a tailwind, not the feared extinction event:

"There had been [fear] these companies would see their business models completely crushed, but in fact, the AI-driven targeting tools have actually triggered an advertising boom... [the incumbents] have all of the data to fuel that growth."

Matt pushed back on the tidy "duopoly wins forever" story with two caveats worth holding onto:

  1. Don't count out Amazon (and Walmart and Target). Amazon's ad business "sits a lot closer to the actual purchases you make than either Google or Facebook ads do," the structural reason retail media keeps outgrowing everything else.
  2. Agentic AI shopping is the real wild card. "The question of who gets paid when an AI agent clicks buy, there's not really a clear answer there just yet." If autonomous shopping agents disintermediate the sponsored link, a lot of business models built on that link have a problem.

So the debate isn't "does AI kill digital advertising", this week's evidence says no. It's "does AI reshuffle who captures the next dollar," with agentic commerce as the open, unresolved threat.

Stocks in play

Alphabet (GOOGL) was the most-discussed name of the week by a wide margin. Ad engine healthy (Search +17%, YouTube +13%), Cloud spectacular (+82%), but the stock sold off on a raised $195–205 billion capex guide, a delayed Gemini model, negative quarterly free cash flow, and an $80 billion equity raise. Bull case (Wall Street Unplugged, Fast Money): distribution moat across five 3-billion-user properties, a differentiated in-house chip story, and Berkshire's $10 billion vote of confidence. Bear case (Bloomberg Intelligence): a soft cloud backlog and the slow-burn risk that chat behavior erodes Search over two-to-three years.

Meta (META) drew almost no direct earnings/ad-revenue commentary this week (its print hadn't landed in the window). The one hard read-through is negative-ish on near-term sentiment: if Alphabet's capex forces the whole field to spend more, Meta is "in the arms race" too (Fast Money). On the product side, Meta's Advantage+ (its AI-automated ad campaign product) keeps showing up as the default that operators actually run, see Read-throughs.

Amazon (AMZN) was framed repeatedly as the retail-media gorilla (about two-thirds of a ~$75 billion market per Criteo on The CPG View) and as the ad platform whose ads sit closest to purchase (Motley Fool). No dedicated Amazon ad-segment coverage this week.

The Trade Desk (TTD) was thin this week. No dedicated episode. The only substantive mention: Instacart listing The Trade Desk among the off-platform surfaces where it activates first-party data (The CPG Guys). Nothing new this week on Kokai, UID2, or the open-internet thesis.

AppLovin (APP) got the best independent read of the week, from Olivia Kory on Marketecture (July 24). Her verdict: it works, but understand why. "The effects of AppLovin ads are short-term. They work immediately," which suits low-consideration, low-average-order-value impulse categories (apparel, footwear) and struggles with subscriptions. Crucially, AppLovin "opened it up to third parties right away" for incrementality testing, "as easy as it is in any like Meta or Google." That measurability is a genuine differentiator versus other channels brands are skeptical of.

Read-throughs

The measurement/incrementality war (the real story under everything). On Ecommerce Playbook (July 21), Tony Chopp (VP of Paid Media at Common Thread Collective) gave the cleanest tutorial of the week on why platform-reported returns lie. His example: "if Google brand search has an incrementality factor of 0.3, then a platform-reported 10:1 ROAS actually represents a 3:1 incremental return." His agency's working benchmarks: Google non-brand incrementality median of 0.6, Google brand of 0.27, meaning much of "brand search" revenue would have happened anyway (last-click attribution "is the most generous... in digital advertising"). Olivia Kory's coffee-shop analogy on Marketecture nailed the concept for a lay reader: the cappuccino sign at the register, "Did that sign make you buy the cappuccino, or were you going to order it anyway? That is the difference between attribution and incrementality." The investable read: the platforms and channels that let brands measure true incrementality (AppLovin, Walmart's new incrementality product, MRC-accredited Instacart) are winning trust, while the ones that don't (parts of retail media, affiliate, Amazon Search) are, in Kory's words, "limiting their advertising TAM."

Signal loss / privacy. On The Agile Brand (July 24), Infolinks CEO Bob Regular argued the deterministic signal advertisers rely on, cookies, matched emails, "is continually deprecating" thanks to privacy laws, walled-garden and browser changes, and shifting behavior ("using AI more... surfing the web a little less... more time on vertical video"). The forced migration is "greatly towards a predictive signal environment," with AI-driven contextual targeting (matching ads to page meaning and sentiment) as the replacement. He also called the industry's obsession with "supply path optimization" (buying media via the shortest route to the publisher) a partial distraction, the shortest path isn't always the best-performing one, "a nice analogy to Google Maps or Waze."

Audio / CTV and Spotify's ad-tech build. On Next in Media (July 21), a Spotify advertising executive detailed the "Spotify Ad Exchange" (about 18 months old), noting over 7,000 advertisers have used its generative-AI creative tool and that Spotify has launched plugins for Claude and Codex that hook directly into its ads API ("if a buyer wants to buy from the command line, they're now able to do that"). The universal theme returned, "everybody is seeking... ROAS numbers", with Spotify leaning on first-party measurement plus AppsFlyer, Nielsen, and Kantar, and on Bayesian attribution models to properly value audio's delayed effect. (This is the closest the week came to substantive CTV/streaming-ad commentary; there was no dedicated coverage of Roku, Disney, Netflix, or Warner Bros. Discovery ad strategy.)

Ads inside AI search (ChatGPT/Gemini/Perplexity). Still more speculation than substance. The most concrete discussion was on Marketecture: reporting that OpenAI faces an "inventory bottleneck" (its queries are so long and context-rich that advertisers can't spend enough), and speculation it may be building an audience-extension ad network, "FAN [Facebook Audience Network], but for ChatGPT." Practitioners testing OpenAI ads reported they "just can't scale to the levels we would need to even test it," and described a confusing strategy that lurched from a Netflix-style, high-CPM curated brand play to a self-serve, ad-tech-forward one. ("CPM" = cost per thousand impressions, the standard price unit for brand ads.)

How operators actually use the big platforms (small-business read-through). Two episodes showed the incumbents' AI ad tools working at the ground level. On The Unofficial Shopify Podcast (July 21), a Shopify product director detailed "Campaign Autopilot," an AI ad agent that picks Meta's Advantage+ as its single campaign type, optimizing purely for checkout conversions, and walked through Meta's learning period ("something like 50 conversion events in a week"). Notably, Shopify named its next ad-channel partners after Meta: Microsoft Ads, then ChatGPT and Snap. And on The Foundr Podcast (July 22), a DTC founder described lifting monthly sales from an inconsistent $68K–$80K to a steady $140K–$150K largely through disciplined Meta-ads creative testing, a reminder that, whatever the AI-search noise, Meta's performance-ad machine is still where small brands go to scale.

What changed vs. last week

  • The center of gravity shifted from Meta to Alphabet. Alphabet's earnings pulled essentially all the market/analyst attention this week; Meta's own ad-revenue story was conspicuously absent (its print hadn't landed).
  • The "AI kills Search" panic keeps fading. A 17% Search growth quarter is not what an existential decline looks like, the debate has moved from "will the ad business survive AI" to "can the ad business justify the AI capex bill," which is a very different (and more bullish-for-ads) question.
  • Retail media crossed a psychological line. The dominant tone shifted from growth-story cheerleading to accountability and incrementality, a maturation that favors the scaled, well-measured players (Amazon, Walmart, Instacart) and squeezes sub-scale "me-too" networks.
  • "Ads inside AI" is still pre-revenue everywhere. Instacart is explicitly waiting until 2H 2026 to experiment; OpenAI can't scale its inventory; Shopify lists ChatGPT as a future partner. Nobody has cracked it yet, which is itself the signal.