CompanyGraph
Amazon.com, Inc.
AMZN · United States
Warehouse and data-center infrastructure is shared across third-party marketplace fulfillment and enterprise cloud hosting so each stream absorbs fixed costs the other could not carry alone.
Amazon's shared infrastructure model — fulfillment centers processing both proprietary and third-party seller stock on the same physical floor, and AWS data centers running internal e-commerce workloads alongside external enterprise applications on the same hardware — spreads fixed capital costs across two demand sources that neither stream could amortize alone, but this co-location creates a single structural vulnerability: because geographic concentration is what makes the cost-sharing possible, a failure at a node like us-east-1 cascades into simultaneous retail and cloud disruptions from one point. The two streams do not scale symmetrically, however, because AWS compute replicates at near-zero marginal cost as demand grows, whereas last-mile delivery requires linear additions of vehicles, drivers, and sortation facilities that cannot be virtualized, meaning peak-season metropolitan delivery capacity becomes the hard ceiling on fulfillment throughput regardless of how efficiently the warehouse floor operates. That asymmetry is reinforced by data localization regulations, which force AWS to replicate physical infrastructure across jurisdictions rather than consolidate it, adding capital cost precisely where the dual-use sharing logic would otherwise reduce it. Replacement friction — FBA sellers with inventory already inside the facilities, AWS customers with applications built around proprietary services, and Prime members on auto-renewing 12-month billing cycles — keeps volume on the shared infrastructure, sustaining the unit economics that the entire cost-amortization structure depends on.
How does this company make money?
Third-party marketplace sales generate a portion of each transaction through seller fees. AWS bills enterprise customers on a usage-based model tied to compute and storage consumption. Prime membership operates as a subscription paid on recurring cycles. Sponsored product placements in search results generate advertising payments from sellers. First-party retail product sales — goods the company buys and resells directly — form an additional distinct income stream.
What makes this company hard to replace?
FBA sellers already have physical inventory stored inside the fulfillment centers, with warehouse-management systems integrated directly into those facilities, creating a stranded-asset problem if they were to leave. AWS customers whose applications are built around proprietary services such as DynamoDB and Lambda would need to rewrite substantial amounts of code to migrate to another provider. Prime membership auto-renewal billing cycles create 12-month retention windows that require an active cancellation step to exit.
What limits this company?
Last-mile delivery capacity in metropolitan areas during peak seasons is the hard throughput ceiling. Prime-eligible order volume exceeds the combined capacity of internal delivery fleets, delivery service partner networks, and USPS rural coverage. Unlike AWS compute — which replicates at near-zero marginal cost — each additional delivery unit requires a physical vehicle, a driver, and a local sortation facility that cannot be virtualized or centralized.
What does this company depend on?
USPS handles rural and residential delivery where the company's own logistics cannot reach economically. Visa and Mastercard payment processing networks underpin transaction handling. Third-party sellers supply approximately 60% of units sold on the marketplace. AWS availability zones across 33 regions require utility grid power to operate. UPS and FedEx networks serve as backup fulfillment capacity when internal capacity is exceeded.
Who depends on this company?
Third-party marketplace sellers whose inventory management systems integrate with Fulfillment by Amazon APIs would face stranded logistics investments if that access were terminated. AWS enterprise customers running production workloads on EC2 instances (virtual computing servers) would face application downtime during any migration away from the platform. Whole Foods stores depend on supply chain integration with the broader network for inventory replenishment.
How does this company scale?
AWS software services replicate at near-zero marginal cost across global data center infrastructure as demand grows. Last-mile delivery, by contrast, requires linear scaling of delivery vehicles, drivers, and local sortation facilities — physical assets that cannot be virtualized or centralized.
What external forces can significantly affect this company?
GDPR and emerging data localization regulations require AWS data centers to keep customer data within specific jurisdictions, constraining how infrastructure can be shared or consolidated across borders. U.S.-China trade tensions affect cross-border fulfillment operations and cloud service availability in that market. Rural broadband limitations restrict deployment of AWS edge computing — infrastructure placed close to end users to reduce response-time delays — for latency-sensitive applications.
Where is this company structurally vulnerable?
Internal e-commerce order processing and external enterprise customer applications run on shared physical infrastructure, so an outage in the us-east-1 Virginia region disables both streams from a single failure point. The geographic concentration that makes dual-use cost amortization possible is the exact mechanism by which one infrastructure event cascades into simultaneous retail and cloud failures.