Amazon Bets on AI Buying Agent as Ad Revenue Nears $20bn Quarterly
The Development
Amazon's advertising revenue is approaching $20bn per quarter, and the company is positioning an AI buying agent as the next growth driver. The same day, Walmart announced it is bringing its shopper data to Vibe's self-serve connected TV platform, extending its first-party audiences into a channel where Vibe already serves performance-minded advertisers. The two moves landed alongside a separate publisher-side data point: Reach titles recorded the steepest audience declines in the latest UK publisher figures. Together they mark the continued shift of ad spend toward retail-data-powered, increasingly automated buying environments, and away from open-web publishers.
Our Take
Amazon is not building an AI buying agent as a convenience feature. It is building the interface through which budget gets allocated, and an interface that sits on top of purchase data compounds its own advantage. Every campaign the agent runs sharpens its optimisation against Amazon's own sales signal. Walmart's Vibe move is the counter-strategy: if you cannot own the agent, own the data supply behind someone else's. Both approaches bypass the open-web publishers losing audience, which is why Reach's declines belong in the same story. Budget is following verified outcomes and automation, and the middle of the stack is getting squeezed.
What Changed
Retail media networks can now execute campaigns through autonomous buying agents and syndicate first-party purchase data into third-party self-serve CTV. Planning, audience selection, and activation collapse into a single agent-mediated workflow tied to closed-loop sales data, with no agency in the middle.
Marketing Impact
Media and retail media teams face a decision about delegating allocation to platform-owned agents. Performance marketers gain cheaper, faster CTV access with closed-loop sales attribution via Walmart and Vibe. Agency-held planning and buying margin is the line item most exposed as agents absorb workflow.
Competitive Implication
Retailers with first-party purchase data and a buying interface gain structural advantage because their agents optimise against real sales. Open-web publishers dependent on declining referral audiences and agencies whose value is manual media execution become vulnerable, since both are disintermediated by data-plus-automation stacks.
Strategic Outlook
Expect Google, Meta Platforms, and The Trade Desk to accelerate their own agent offerings through 2027. Walmart will likely extend data syndication beyond Vibe to other self-serve channels. Advertisers will demand agent transparency and cross-platform measurement as walled-garden agents begin competing for the same budgets.
The Exploit
Action Item
Heads of media should run a controlled head-to-head test in Q1 2027: allocate 10% of retail media budget to Amazon's AI buying agent against human-managed campaigns, with identical KPIs, and decide on wider delegation from the results.