S09.10 · Retail & Commerce
The fastest-growing ad channel, monetizing shopper data at software margins, now facing an existential agentic-shopping threat.
US retail media ad spend is forecast to reach nearly $70B in 2026, up from roughly $31B in 2021 (eMarketer) — the fastest-growing advertising channel in the market. Amazon alone is projected to exceed $75B in global ad revenue by 2028. The channel layers roughly 70%+ software-style margins on top of low-margin retail operations, effectively reallocating retail's profit pool from merchandise to media. It also carries the sector's single largest structural risk, because the same agentic AI reshaping shopping behavior elsewhere strikes directly at the mechanism that funds it.
Off-site retail media — using retailer data to target ads on other platforms — is growing faster than on-site retail media (eMarketer). The US leads and is the most measured market; adoption is moving fast in the EU and UK through retailers such as Tesco and Carrefour, and Alibaba's and JD's advertising businesses in China rank among the largest globally. Concentration is extreme. Amazon dominates; a "scaled second tier" — Walmart Connect, Instacart, Roundel, Kroger Precision Marketing — has separated itself from a long tail of sub-scale networks that lack sufficient first-party data or traffic to compete. What is being monetized is first-party shopper data and owned digital real estate, sold on behalf of CPG and brand advertisers.
Scale decides who gets to participate at all. A retail media network justifies its build only when the retailer holds enough first-party purchase data and enough owned traffic to make targeting meaningfully better than open-web advertising; below that threshold, the economics never clear the cost of the ad-serving and measurement stack. Hence the sharp bifurcation: Amazon, a handful of credible second-tier networks with real scale in their category, and a long tail of retailers with a network in name only.
For a retailer, retail media is about as close to pure expansion as this sector offers — new high-margin revenue with no cannibalization risk against the core merchandise business — while consolidation among sub-scale networks is a defense against irrelevance. Measurement and targeting automation is lowering campaign costs for advertisers; that is efficiency, nothing more.
The structural risk is larger than any of it. Sponsored search and sponsored placement earn only while a human being is scrolling a results page and can be influenced by what appears at the top of it. Agentic shopping — AI agents purchasing on spec and price rather than browsing sponsored placements — removes the human attention the entire advertising mechanism depends on, which makes it the single biggest threat to this profit pool. Growth in the channel is already underway and strong; the agentic threat to the underlying ad model is expected to bite within 2-5 years and, unlike most disruption timelines in this sector, is described as potentially existential to the category as currently constructed.