The debate over Amazon is not whether e-commerce grows. It will. The debate is whether Amazon can convert AI-mediated shopping and retail advertising into durable, high-margin revenue without weakening two assets it cannot easily rebuild: advertiser trust and seller economics. A modest gain in advertising monetization may matter more to earnings quality than a point of retail market share. That is a claim about margin mix, and margin mix is a measurement problem. The history of advertising is a history of unmeasured waste. Read what follows the way an auditor reads a ledger: what is counted, what is assumed, and what is quietly not counted at all.
The Demand Baseline: Steady, but Not Decisive
Demand is not the constraint. The global e-commerce market is projected to grow about 6% annually from 2026 through 2031 7. U.S. retail e-commerce sales reached $340.2 billion in Q2 2026, up 12.2% year over year, after 5.0% growth in Q2 2025 10. Any catalog merchant would take that trade. But growth of this kind is a floor, not a verdict. It tells you customers keep buying online. It says nothing about who captures the profit on the transaction, or at what margin.
Where the Margin Actually Lives
The margin lives beside the merchandise, not in it. Retail media and advertising carry EBIT margins of 60–80%, against 2–5% for first-party retail 7. The arithmetic is old. Selling goods requires buying inventory, moving it, and standing behind it. Selling the space beside the goods requires none of that. This is why a modest share or monetization gain in advertising could move earnings quality more than baseline online-sales growth ever will. The strategic variable is not volume. It is mix — and whether the high-margin mix can expand without taxing the sellers who stock the shelf.
AI Discovery: A Real Channel, an Unproven Conversion
AI is becoming a discovery channel, and the traffic is real. Adobe reported 393% year-over-year growth in LLM and AI traffic in the first quarter of 2026 7. Consumers are willing to be assisted. In a Visa survey, 40% of Americans said AI recommendations had influenced a purchase, but acceptance fell from 58% for AI price comparison to 27% for autonomous execution 7. Read that gradient carefully. Shoppers will take advice from the machine before they let it spend for them. Assisted discovery scales first; delegated purchasing waits.
Advertising has already followed the traffic. Retail advertisers accounted for 39% of ChatGPT ads 1,5,6, and retail queries represented 24% of U.S. ChatGPT search queries 6. AI search is shaping up as a contested customer-acquisition channel, and Amazon has every reason to contest it. But mark what the evidence establishes and what it does not. Rapid AI referral growth supports strategic relevance. It does not demonstrate Amazon-specific conversion or revenue gains. A referral is a click, not a sale. Until someone runs an incrementality test on that traffic, it is booked as return on faith. This creates undetected risk. The question is not whether AI-assisted commerce works, but how you know it works.
The Advertiser Trust Ledger
Advertising is the clearest high-margin opportunity in the material and the most consequential risk. The material alleges that Amazon advertisers were charged approximately $20 billion in surcharges affecting 1.2 million advertisers 4. It further reports that the share of advertisers charged their full budget cap rose from about 4% in 2020 to about 79% in 2024 9. These are allegations, not verified operating results, and should be weighed as such. Their strategic weight is nonetheless plain. The advertising product sells one thing: cost-per-acquisition integrity. Merchants buy it because returns are measurable and repeatable. If auction mechanics quietly absorb budgets — if the party selling the measurement also controls the measurement — the waste fraction does not disappear. It migrates from the advertiser's ledger to the platform's revenue line. That is the quiet first step of attribution collapse, in a business whose entire premise is efficient, measurable returns.
Seller Economics Set the Floor
Merchant economics are the complementary constraint, and the arithmetic turns punitive at low ticket sizes. One marketplace example — not Amazon's — makes the point: an 8% commission plus 2.9% and $0.30 per transaction at Whatnot equates to 25% of a $2 sale 10. Fixed fees bite hardest at the bottom of the price ladder. Every mail-order operator learned that from postage minimums. When a platform presses for more monetization from already thin-margin sellers, merchant retention and assortment quality are the first casualties.
Fulfillment costs point the same direction. The material states that a company plans to raise peak-season storage fees by 250% or more 8. The claim carries no named-company attribution, so treat it as directional evidence of fulfillment-cost risk rather than an Amazon-specific fact. The wider lesson holds either way. Advertising and fulfillment monetization can lift platform revenue this quarter while quietly weakening the seller economics that sustain selection, price competitiveness, and — in time — the advertising spending itself.
What the Insider Filings Do Not Say
The insider record supports no directional conclusion about the operating business. Andrew R. Jassy converted 50,000 RSUs at a $0 price on August 21, 2026, and his direct holdings rose by a net 30,000 shares after the reported transactions 2. Separately, Douglas J. Herrington acquired 7,500 common shares through an RSU conversion and held 474,638 shares directly afterward 3. These are compensation mechanics and ownership records. They disclose no view on fundamentals or trading prospects. Reading conviction into RSU schedules is an attribution error of the filing-page variety.
How You Would Actually Know
The evidence sorts into four findings. Margin mix is the central upside: retail-media network economics can make ad growth and AI-enabled discovery worth more to earnings than incremental first-party volume. Trust is the central downside: the surcharge allegations and the sharp reported rise in full budget-cap charges pose a governance and retention question, if substantiated. AI-commerce evidence is promising but indirect: rapid referral growth and consumer openness support strategic relevance, not demonstrated revenue. Merchant economics require balance: monetization that outruns seller viability buys short-term revenue with long-term selection.
A more rigorous approach would follow the merchant's own discipline. Separate assisted discovery from delegated purchasing in the measurement. Test incrementality on AI-referred traffic before crediting it. Treat auction mechanics as a cost-per-acquisition question, because that is how the advertisers who fund the margin already treat them.
The market will grow about 6% a year whether or not anyone audits the machinery behind the margin. The oldest lesson in this trade holds that half of advertising is wasted and no one knows which half. For Amazon's advertising-led margin story, the sharper question is this: with the platform selling the ads and grading them, who is even trying to find out?