Amazon is no longer simply an e-commerce marketplace. It is combining commerce, advertising, media, logistics, and cloud-based AI into one operating system for customer acquisition and conversion. The opportunity is clear: Amazon can use first-party shopping data to connect advertising exposure with product discovery, comparison, retargeting, and purchase 15,18. The measurement risk is equally clear. Inventory demand, brand lift, conversion, recognized advertising revenue, and incremental profit are not the same thing.
The question is not whether Amazon’s advertising ecosystem works, but how the company and its advertisers know that it works. The history of advertising is a history of unmeasured waste. Amazon’s advantage is the possibility of reducing that waste through purchase-intent data and closed-loop measurement. That advantage will persist only if product data is accurate, marketplace enforcement is credible, and reported sales lift represents genuine incrementality rather than attribution capture.
The evidence is concentrated in late July and early August 2026 and is therefore relatively current. Much of the detail, however, comes from single-source reporting. The strongest corroboration concerns advertising demand, international television targeting, Prime Video reach, delivery expansion, and the relationship between retail-media data quality and transaction performance.
Advertising Is Amazon’s Most Interconnected Growth Engine
From sponsored search to full-funnel commerce media
Amazon’s advertising proposition rests on purchase-intent and shopping-behavior data. Advertisers can connect exposure with product-page visits, comparisons, retargeting, and purchases 15,18. That is more commercially actionable than impressions, clicks, reach, or engagement because retail media can incorporate product discovery, shopping behavior, purchase signals, and potentially closed-loop measurement 18. Advertisers are consequently shifting budgets toward commerce platforms that can connect media spending directly to sales outcomes 18.
Amazon’s stack now spans Sponsored Products, Sponsored Brands, Sponsored Display, Amazon DSP, sponsored prompts, Prime Video, sports programming, and creator placements. Sponsored Products remain the largest and most conversion-oriented offering 18,19. Sponsored Brands support discovery and consideration through headline search, Stores traffic, and video 18. Sponsored Display adds retargeting and audience-based visibility on Amazon and in selected off-Amazon placements. DSP applies Amazon audience and shopping data across Amazon-owned and third-party inventory 18. Sponsored Products are also moving into creator placements, extending the system beyond conventional marketplace search 34.
Demand indicators are strong, but they should not be converted mechanically into revenue estimates. Amazon’s sports inventory was reportedly sold out across Thursday Night Football, the NBA, WNBA, and NASCAR 20. Other reporting likewise described sports-related inventory as sold out 39,46. Amazon added 30 new NBA advertisers during the first year of its partnership 20. Sports rights are therefore functioning as both audience assets and customer-acquisition vehicles for Amazon Ads.
Amazon Ads has been characterized as the world’s third-largest advertising network 16, and advertising is described as a large, fast-growing revenue stream 21. These claims are directionally consistent. Most individual metrics, however, have only one or two sources. They should be treated as indicators, not audited evidence.
Television measurement is improving, but scale remains constrained
Amazon’s expansion of zip-code-level television targeting into Canada, Mexico, and Brazil is a more concrete product-development signal. The rollout is supported by five sources 2,3,48 and includes geographic and creative localization, including state- and postal-code-specific variants 47,48. The initial restriction to first-party video inventory and 1PAS content 3,47,48 limits near-term scale. It also gives Amazon a controlled environment in which to bring upper-funnel television advertising into a commerce-linked measurement framework.
The XCM pilot is intended to improve cross-platform and Total Video campaign measurement 50. Prime Video’s inclusion in NMO’s independent Total Video reporting framework should improve comparability with other premium video platforms 50. This is meaningful because measurement comparability is a prerequisite for budget reallocation. Advertisers cannot judge incremental return if each platform defines reach and conversion differently.
The central tension remains. Amazon’s value proposition is strongest when campaigns can be tied to purchases, but awareness and acquisition campaigns should not be evaluated solely through conversion metrics 18. In the Galderma test, combining Prime Video streaming video, Amazon DSP, and Sponsored Ads increased consumer consideration by 4.5% 6,7,35,42. Consideration is an upper- or mid-funnel measure, not a direct sales outcome 42. Amazon is moving toward full-funnel advertising, but advertisers and investors must keep inventory demand, brand lift, conversion, and recognized revenue separate.
Prime Video Builds Reach Into the Advertising Flywheel
Prime Video has a different economic role from Netflix. Amazon uses video to increase the value, acquisition, and retention of Prime, while Netflix relies more directly on paid subscriptions 21. Prime Video’s economics therefore include bundle value, advertising, the paid ad-free tier, Prime Video Channels, and transactional rentals and purchases 21.
The audience is distributed across a global Prime membership base numbering in the hundreds of millions 21. Global Prime membership was estimated at 285 million in 2025 11. Advertising was enabled by default on Prime Video, allowing Amazon to create a large ad-supported audience without acquiring each viewer independently 21. Prime Video advertising is described as the service’s fastest-growing direct revenue stream 21. The service also supports Prime retention, shopping activity, add-on subscriptions, and transactional video revenue 21.
Prime Video Channels increases engagement frequency and turns the application into a broader entertainment destination 21. Live sports provide scarce premium inventory with demonstrated advertiser demand. Reported audience figures reinforce the scale of the asset: “Off Campus” reached 36 million global viewers in its first 12 days; NBA Eastern Conference Semifinals Game 7 reached 6.5 million U.S. viewers; European NBA viewership more than doubled year over year; and NASCAR averaged 2.3 million viewers in its second Prime Video season 10.
The economics are not established by reach alone. Prime Video depends on advertisers’ willingness to pay for available inventory 21, and public sentiment regarding recommendation quality has reportedly been skeptical and negative 28. Amazon also withdrew from cricket broadcasting rights on June 5, 2026 49. That decision suggests the company is not pursuing every sports asset regardless of price.
The relevant investment question is not whether Prime Video can build reach. It is whether content, advertising load, measurement, and rights costs can increase total Prime ecosystem value. The available evidence supports audience traction and advertiser demand. It does not establish the return on invested content capital.
Product Data Determines Retail-Media Quality
Criteo’s findings identify a foundational constraint. Two sources report a positive correlation between deeper product feeds and larger consumer baskets 40,41. Related observations link product-feed depth to basket size and potentially higher transaction values 38,43,44. The evidence is correlational. No numerical uplift or methodology is provided 37.
The implication is nevertheless material. Better titles, attributes, imagery, availability, pricing, and structured product information can improve discovery, recommendation quality, ad relevance, and basket size. Claims that basket-size gains depend on accurate, deep, and well-maintained feeds 43 make data governance an economic issue rather than a catalog-management detail.
That issue becomes more important as conversational commerce develops. Chat interfaces may displace or supplement conventional search and marketplace discovery 37,43,54. Amazon has a significant data advantage, but its scale also increases the cost of inaccurate or inconsistent catalog data. Amazon recorded 984 million product-title updates since June, and AI-rewritten titles are being rolled out across experiences 36. Automation may improve catalog quality at scale. It may also produce inaccurate or context-stripped content if controls are inadequate.
The evidence does not show that AI shopping has already displaced established search. A Productrise study found that standard Google Search returned products for 88% of queries versus 23% for Google AI Mode 53. This is not an Amazon-specific comparison and comes from a single source. It does, however, caution against assuming that AI shopping interfaces will immediately replace existing search funnels. Amazon’s opportunity is to use AI to improve discovery while preserving the high-intent transaction environment that supports its advertising advantage.
Marketplace Integrity Creates Attribution and Regulatory Risk
Amazon’s marketplace scale creates a persistent trade-off among selection, seller participation, and trust. Abusive sellers reportedly return after temporary removal under different names, through new listings, or with modified accounts 31. Monitoring systems must track unauthorized sellers, copycat listings, pricing disruption, listing edits, and abuse patterns 31. Static enforcement is insufficient because seller tactics change continuously 31.
Enforcement performance is not uniform across marketplaces 33. Trademark takedowns may be slower or less predictable on Amazon.com.mx and Amazon.com.au than on Amazon.com 33. Rights holders may also be unable to act in European storefronts without locally applicable patent or trademark rights 33. International scale therefore increases the enforcement burden, and successful U.S. controls should not be assumed to translate uniformly overseas.
Product-origin claims illustrate the trust problem. Misleading country-of-origin claims were found on both Amazon and Walmart marketplaces 22, including third-party listings 14. Amazon sellers remain responsible for accurately labeling ingredients, warnings, usage instructions, and country of origin 32. The FTC’s “Made in USA” standard requires products to be all or virtually all made in the United States 24.
Amazon’s Alexa for Shopping and Walmart’s Sparky reportedly recognized contradictory or potentially deceptive information but did not flag it to users or intervene 12. Walmart’s Sparky characterized enforcement responsibility as falling on manufacturers rather than retailers 12,23, while Amazon and Walmart provided limited or differing responses 24. These claims do not establish systemic Amazon misconduct. They do identify a reputational and regulatory risk: AI can detect questionable information without converting detection into consumer protection.
Brand-protection tools remain useful but fragmented by jurisdiction. U.S. rights holders can record trademarks and copyrights with Customs and Border Protection through e-Recordation to prevent infringing goods from reaching Amazon fulfillment centers 33. Trademark owners may have Lanham Act claims where unauthorized selling creates a likelihood of consumer confusion 31, and Amazon sellers must avoid infringing third-party rights 32. Brand Registry enrollment generally requires a trademark in the relevant territory. A U.S. trademark does not automatically extend to European or Asian storefronts 33.
This creates undetected risk for advertising as well as compliance. Bad listings can attract impressions and clicks while weakening conversion, trust, and repeat purchase. A retail-media network cannot maintain cost-per-acquisition integrity if the underlying product information is unreliable.
Logistics and Grocery Expansion Increase Frequency—At a Cost
Amazon continues to use fulfillment speed and category expansion to increase engagement. Amazon Now, a delivery service offering delivery in 30 minutes or less, is expanding to dozens of additional U.S. cities, a development corroborated by four sources 1,53. The broader rapid-delivery trend is moving beyond food and small parcels. Uber has been used to deliver Bunnings lawnmowers, and other platforms are testing bulky home-improvement categories 37,38,41.
Amazon’s delivery model depends heavily on Delivery Service Partners, which move packages from warehouses to customers’ homes 26. These partners are strategically important to delivery speed 25,27, but they expose the model to partner economics, labor availability, service quality, and regulatory scrutiny.
Customer-frequency indicators are encouraging. The number of customers ordering perishable groceries from Amazon reportedly increased 50% since the beginning of the year 20. Amazon Pharmacy same-day prescription deliveries nearly quintupled during the first six months of 2026, while customer savings from automatically applied manufacturer discounts increased more than 400% year over year 10. These are single-source claims, but they fit the broader strategy: faster fulfillment and recurring or time-sensitive categories can raise order frequency and strengthen Prime retention.
The risk is that rapid delivery and healthcare logistics may carry lower margins or greater operating complexity than headline growth rates imply. The question is not whether Amazon can subsidize more frequent orders, but whether those orders produce profitable frequency.
Competition Is Moving Into Amazon’s Measurement Territory
Amazon remains a major traffic and intent platform. Approximately 74% of U.S. consumers reportedly begin product searches on Amazon 15. Competitive pressure is nonetheless widening. Amazon faces Shopify, Etsy, and eBay in marketplace services 11, while Temu and Shein are gaining ground as low-cost digital marketplaces 13.
In April 2026, Walmart, Temu, and eBay collectively attracted 342.5 million U.S. unique visitors versus Amazon’s 243.8 million, although Amazon still represented 71.2% of the combined audience on the stated comparison 45. These figures measure traffic, not purchase quality, profitability, or customer lifetime value. They do show that Amazon’s reach advantage is not unchallenged.
Argentina illustrates the international dynamic. Amazon, Temu, Shein, and AliExpress are maintaining or expanding consumer traction 52. International price differences, broader selection, eased import restrictions, and lower prices than domestic alternatives are driving demand 52. Shein and AliExpress are specifically attracting consumers through low pricing 52. Amazon must therefore combine price competitiveness with trust, fulfillment reliability, local payment and tax convenience, and advertising relevance. Low-price international marketplaces may also impose hidden delivery, duty, and administrative costs, including parcel fees charged before delivery 53.
Shopify’s expansion of Shop Campaigns is another competitive signal. The campaigns charge merchants only when a sale converts, initially run within the Shop app, and are intended to add Meta and Google placements later 54. This moves media buying and conversion measurement into Shopify’s merchant ecosystem 54. Amazon retains the advantage of deeper purchase-intent data and marketplace demand. Shopify’s merchant-centric model could nevertheless attract brands that want performance marketing without relying entirely on Amazon’s marketplace.
Investment Implications
The cluster identifies four linked investment themes: retail-media monetization, Prime Video and sports-led audience expansion, operational density through fast fulfillment, and AI-enabled commerce infrastructure. Their common mechanism is a platform flywheel. More Prime members and faster delivery can increase shopping frequency. Higher traffic attracts sellers and creates more behavioral data 29. Richer product data can improve discovery and basket size 38,40,41. Stronger purchase signals can make advertising more measurable. Advertising revenue can then support investment in content, technology, and logistics.
Advertising appears to be the most immediately monetizable opportunity. Sports sellouts, Prime Video’s default ad load, DSP targeting, sponsored placements, creator expansion, and cross-platform measurement all increase addressable inventory and strengthen the advertiser proposition. But revenue quality matters. Search inventory is limited by the number of active shoppers at any moment 18. Keyword auctions and sponsored placements are becoming more expensive 18. Amazon’s “Advertising Services Sales” may not match estimates of U.S. digital advertising spend, global net advertising revenue, or retail-media revenue 18.
Investors should track advertising growth alongside incremental margins, advertiser retention, sales lift, and the mix between high-intent Sponsored Products and more capacity-constrained upper-funnel inventory. This is the required test for attribution integrity. Gross demand is not incremental return.
Prime Video is strategically valuable even if direct video profitability remains difficult to isolate. Its primary role is to raise the value of Prime, create a scalable ad-supported audience, and add cross-selling opportunities 21. The key debate is whether sports rights and original content generate sufficient incremental Prime retention, shopping frequency, and advertising demand to offset rights and production costs. Current evidence supports audience traction and advertiser demand, but not the return on invested content capital.
AI is an enabler, not a standalone thesis. Amazon is applying it to title rewriting, shopping assistants, campaign management, sponsored prompts, and cloud infrastructure. AWS supports third-party models such as Moonshot AI’s Kimi K3 9, hosts models including Llama, Mistral, Nova, and marketplace offerings 51, and provides large-scale deployment through SageMaker HyperPod and EKS 9. CloudWatch managed Prometheus collectors reduce the operational burden of monitoring workloads 9. Bedrock supports prompt caching and web grounding 8,30.
Inference demand is more price-sensitive than training demand, even if it is more recurring and diversified 17. The commercial outcome will depend on adoption, workload retention, pricing, and infrastructure margins—not feature count.
The principal downside is that Amazon’s ecosystem amplifies both good and bad data. Product-feed depth can increase basket size, but inaccurate listings, deceptive origin claims, AI-rewritten titles, and recurring abusive sellers can undermine trust and conversion. Onton’s product-trust benchmarks illustrate the uncertainty. Onton claims superior accuracy versus Google Shopping and Amazon 4,5, but the figures are based on internal company benchmarks and have not been independently verified 4,5. This is an isolated, unverified competitive claim, not evidence of Amazon underperformance. It does reinforce the need for independent validation when assessing AI shopping quality.
Bottom Line
Amazon’s platform breadth supports a constructive strategic view, particularly in advertising, Prime Video, and high-frequency fulfillment. The advertising opportunity is substantial because Amazon can connect media with purchase intent across Sponsored Products, DSP, Prime Video, sports, and creator inventory 15,18,20. Prime Video is most valuable as a Prime-retention, shopping-frequency, and advertising asset—not simply as a standalone streaming business 21.
The investment case remains favorable but execution-sensitive. Product-data depth, catalog accuracy, and marketplace enforcement are foundational to advertising and conversion economics; rapid AI-driven catalog changes increase both efficiency potential and trust risk 31,36,40,41,43. Investors should monitor advertising incremental margins, content and sports-rights returns, delivery economics, international price competition, and consistency of marketplace enforcement across jurisdictions 33.
The decisive question is simple: how much of Amazon’s reported advertising growth represents new, profitable demand, and how much is merely better credit for demand that would have existed without the ad?