How Amazon Built a Global Logistics Network from Scratch

In November 2015, I sat in a strategy meeting where Amazon quietly announced it would stop being a customer of logistics companies and start becoming one. What happened over the next decade redefined what a supply chain could look like — and what it means to compete on delivery.

It Started with Warehouses

Amazon's original supply chain leadership was entirely warehouse-centric — what insiders called "FC Centric." The company's first generation of executives came up through fulfillment centers in the American Midwest. They cared deeply about what happened inside four walls, and less about what happened before goods arrived at the inbound dock.

That made sense for a while. The thing Amazon genuinely owned — its advantage — was the customer promise. Two-day delivery, reliably kept, was something no other online retailer could match. Everything else was contracted out: UPS and FedEx handled delivery, and for a time that worked well for everyone.

The Breaking Point: A Christmas Blizzard

Around 2010, during one of the peak holiday seasons, UPS hit a wall. Carrier capacity could not absorb the surge. Across large parts of the US, Christmas gifts did not arrive in time. The customer experience — Amazon's most sacred asset — had been compromised by something outside Amazon's control.

That was the inflection point. Amazon responded on three tracks at once: it began working with the near-bankrupt US Postal Service to expand carrier capacity; it leveraged its volume to force UPS and FedEx into a more competitive arrangement; and it began quietly planning its own delivery infrastructure.

The 2015 Pivot: Nodes, Arcs, and a Real Network

At the 2015 North American supply chain strategy session, the direction became clear: Amazon would move from a model centered on fulfillment centers to a model centered on a true end-to-end network. Warehouses were nodes. Transportation was the arc connecting them. The goal was to control every segment of both.

Four new divisions launched in relatively rapid succession:

• Amazon Global Logistics (AGL) — extending the supply chain upstream to the origin port, giving Amazon visibility and control over cross-border shipments from Asia.
• Amazon Transportation System (ATS) — an owned trucking operation for inter-facility freight, replacing reliance on third-party carriers for domestic line haul.
• Amazon Air — a dedicated air cargo network for urgent inventory rebalancing, enabling same-day redistribution across the country.
• Amazon Last Mile (AMZL) — a proprietary delivery network of delivery stations and independent contractor drivers (the Delivery Service Partner program), reducing dependence on UPS, FedEx, and USPS for the final mile.

The eBay Lesson

eBay was Amazon's closest peer in the early 2000s — and was actually more profitable in those years. It made a deliberate choice not to invest heavily in physical fulfillment infrastructure. The consequences played out slowly but inexorably. eBay is still a significant marketplace, but it is permanently constrained to the parts of e-commerce that do not require a logistics backbone. Amazon, by contrast, used its retail operation as the foundation for a logistics business that now serves not just its own marketplace but an expanding range of external customers.

Later entrants like Shopify have found it impossible to close the gap. The window for building a competing physical network has closed. The capital required, the time required, and the operational learning curve required are simply no longer accessible at the scale Amazon occupies.

The Technology Behind It All: SCOT

Amazon's supply chain optimization technology — internally called SCOT (Supply Chain Optimization Technology) — is what makes the physical network intelligent. Rather than a single integrated system like SAP, SCOT is an ecosystem of thousands of specialized modules, each handling a specific part of the planning, allocation, execution, or analytics stack.

Amazon Fulfillment Technology (AFT) reaches into every scan, every indicator light, every picker's path, every conveyor endpoint inside a warehouse. Amazon Transportation Systems (ATS) directs every driver, every truck stop, every route. Every execution outcome is recorded and fed back into the planning layer.

When I was at Amazon, even Walmart and JD.com — both world-class supply chain operators — were running algorithms that were meaningfully behind Amazon's. That gap has probably narrowed. But Amazon's head start, measured in data volume and algorithmic iteration cycles, remains substantial.

The Planning Discipline Behind the Technology

Technology alone does not produce results. What gave SCOT its power was the organizational discipline that surrounded it. The North American S&OP team I worked on ran on a strict weekly and annual cadence:

• Sunday: prior week business data enters the system.
• Monday: a demand plan covering 50 product lines, 13 weeks forward, at roughly 200–300 million units per week, is pushed from product teams to the S&OP team.
• Tuesday: the planning team and warehouse GMs meet with the North American Operations VP for the Weekly Business Review — last week's warehouse performance reviewed, next 1–2 weeks adjusted.
• Wednesday: the 13-week inbound and outbound plan is finalized and distributed to all fulfillment centers.
• Thursday and Friday: urgent and important issues are resolved.

The annual rhythm was equally disciplined. Q1 was for reviewing peak season lessons and locking current-year initiatives. Q2 was for negotiation and development. Q3 was for stress testing — and by mid-September, large-scale development stopped entirely. The company entered "Black Out": no transfers, no vacations, all hands on peak preparation. The inbound peak arrived in mid-October. The outbound peak ran from early November through Christmas Eve noon.

What This Means for the AI Era

The people who built Amazon's supply chain — the quantitative specialists, the operations managers, the data analysts — did so over 20 years of compounding effort. The irony is not lost on anyone inside or outside the company: the AI systems those teams helped design and train are now systematically replacing them. The recent waves of layoffs at Amazon are not cost-cutting in the conventional sense. They are the operational readout of a system that was always designed to reduce its own dependence on human intervention.

For those who built it — and for anyone starting fresh today — the more useful observation is this: the skills that made Amazon's supply chain possible (systems thinking, data fluency, comfort with ambiguity, relentless focus on execution) are precisely the skills that remain valuable when AI handles the routine. The network Amazon built is a model. Understanding it is part of how you build anything that operates at scale alongside it.