Amazon is reshaping its fulfillment network to shorten distance and time per order, using more local inventory, automation, and ultra-fast delivery windows to lift service and margins simultaneously.
In Brief
- Amazon is redesigning inventory placement and facility design so that unit growth outpaces growth in shipping and fulfillment cost.
- Ultra-fast delivery promises of 30 minutes to 3 hours are being operationalised through dense local nodes, stricter service thresholds, and heavy automation.
- This network model shifts planning focus from moving parcels across long linehaul legs to managing near-real-time local capacity, inventory risk, and cost-to-serve.
A Structural Pivot: From Long-haul Network To Local Minutes
Amazon’s latest disclosures describe a clear operating shift: the fulfillment network is being rebuilt to minimise distance travelled per item and compress delivery windows to hours or minutes, while still improving cost-to-serve.
In the first quarter of 2026, worldwide units grew 15 percent year over year while outbound shipping cost grew 12 percent and fulfillment expense grew 9 percent on a currency-neutral basis. That gap signals a network where productivity gains are now outpacing volume growth.
At the same time, Amazon reports more than one billion items delivered same day or overnight so far in 2026, millions of items available for same-day delivery, and a 1-to-3-hour service on over 90,000 items across thousands of cities and towns. A separate 30-minute service, launched in 2025 and now live in nine countries, is delivering thousands of items and seeing orders in India increase 25 percent month over month.
This is not just more speed on the same footprint. It is a different way of allocating inventory, designing facilities, and setting service thresholds.
How The New Network Works In Operational Terms
Amazon summarises the logic in a few lines: it will ‘keep optimizing inventory placement to shorten distance traveled, reduce touches per package and improve consolidation rates’ while deploying its latest generation of robotics and automation across new and existing facilities.
In operational terms, this kind of shift typically requires:
- A more granular network zoning model that breaks large regions into local catchments for same-day and 1-to-3-hour coverage.
- Inventory policies that move high-velocity, ultra-fast-eligible items into smaller, closer nodes, while holding slower movers in more central locations.
- Order allocation logic that prioritises lowest-distance, lowest-touch fulfillment options that still meet the promised delivery window.
- Facility design that supports higher turns, more automation, and fast cross-docking rather than long dwell times.
Amazon will equip all large-format U.S. fulfillment centres launched in 2026 with its latest automation. Early results are described as improved site safety, higher productivity, and lower cost-to-serve.
At network level, this means more of the unit pick, pack, and sort work is being handled by robotics, with humans focused on exception handling and value-added tasks. The evidence that unit growth now exceeds cost growth suggests that these automation layers are scaling beyond pilot stages into the core operating model.
Grocery and Perishables: Minutes as a Growth Engine
The clearest test bed for this ‘miles for minutes’ model sits in grocery and perishables. Amazon reports more than 150 billion dollars in gross grocery sales in 2025 and is now the second-largest grocer in the U.S. Perishables are delivered same day in more than 2,300 U.S. cities and towns.
Perishable sales for same-day delivery grew more than 40 times year over year and now account for nine of the top ten items where the service is available. Customers using same-day perishables add nearly three times as many items and spend over 80 percent more than customers who do not use that service.
From a supply chain perspective, this reframes ultra-fast fulfillment. Instead of treating minutes as a pure cost premium, Amazon is using the same network investment to increase average order value and item count. That improves vehicle utilisation and spreads fixed facility cost across larger baskets.
To support this, the network must synchronise cold chain capacity, store replenishment, and last-mile routing. Service thresholds for freshness and on-time delivery are tighter than for general merchandise, so planning cycles for demand, labour, and transport become more frequent and more local.
Benchmark Context: How Peers are Moving In The Same Direction
Other large retailers are on similar trajectories, though at different scales. Recent disclosures show:
- Walmart has automated more than half of e-commerce fulfillment centre volume and supplies over 60 percent of its U.S. stores with some freight from automated distribution centres, which has contributed to shipping-cost reductions in the 30 percent range over multiple quarters.
- Value-focused networks such as Dollar General report that more than 80 percent of delivery orders are now fulfilled in under an hour across thousands of locations by combining local store inventory with third-party and proprietary drivers.
Amazon’s difference is the combination of very high assortment breadth, tightly defined delivery windows down to 30 minutes, and an explicit network design goal of reducing distance and touches per package while holding or improving margin.
Planning and Governance Move Closer To The Edge
A minutes-based network requires a different planning cadence and governance model.
First, demand planning and inventory allocation must operate at a finer geographic and temporal resolution. When a company offers same-day or 30-minute delivery on specific assortments, allocation decisions for those items are effectively pre-committing near-term fulfillment capacity within very small radii.
Second, labour and transport capacity planning need to respond to short-term swings without overspending. Short lead times limit the usefulness of traditional weekly planning cycles. Instead, systems must ingest order patterns and adjust staffing, routing, and batching in near real time.
Third, governance over service promises becomes a structural discipline. Once a 30-minute or 1-hour promise is published for a given location and item set, any failure to deliver has immediate cost and brand consequences. That pushes operating teams to treat service-level definitions and change control as core master data tasks, rather than marketing levers.
Amazon’s own language links these pieces directly, stating that as network efficiency improves, it can ‘deliver items faster and improve the customer experience while at the same time lowering our cost to serve’ and that there remain ‘meaningful opportunities to further enhance productivity across our global fulfillment network, all while continuing to raise the bar in delivery speed’.
The Constraint: Capital, Complexity, and Variable Cost Pressure
Trading miles for minutes is not costless. Several constraints are clear in the disclosures.
First, capital intensity is significant. Amazon’s cash capital expenditure in the first quarter of 2026 was 43.2 billion dollars, primarily related to cloud infrastructure and generative AI but also including new fulfillment and logistics capacity. Data centres are built with useful lives of more than 30 years, and chips, servers, and networking gear with lives of five to six years. Fulfillment infrastructure will follow similar cycles. In periods of high growth, capital expenditure will outpace revenue, and free cash flow will be pressured until utilisation catches up.
Second, operating cost shocks such as fuel inflation remain material. Amazon expects higher transportation costs related to fuel and is introducing a fuel and logistics-related surcharge for marketplace sellers to partially offset this. Sharing these costs across network participants protects margin but can influence seller behaviour and assortment.
Third, execution complexity rises with each new service tier. A network that supports standard delivery, same day, 1-to-3-hour windows, and 30-minute promises across multiple geographies must maintain multiple operating modes in parallel. Inventory segmentation, routing rules, and capacity buffers all become more complex, which increases the burden on systems and process discipline.
Finally, volatility in component supply and tariffs still matters. Amazon explicitly notes risks from tariff and trade policies, resource and supply volatility, and regional labour constraints as material drivers of results. Even in a localised fulfillment network, reliance on external inputs for automation, vehicles, and infrastructure continues to shape risk posture.
What This Operating Model Now Enables and Constrains
Amazon’s fulfillment disclosures point to a network that is gradually converting geographic advantage and automation into a structural position where faster service does not automatically mean higher unit cost.
By placing more of the right inventory closer to demand, reducing touches per package, and embedding automation into new large-format sites, the company is creating a configuration where volume growth can run ahead of shipping and fulfillment cost growth for sustained periods.
The trade-off is a higher baseline of capital expenditure, more operational complexity at the edge of the network, and increased sensitivity to labour, fuel, and component shocks. The model demands tighter planning cadence, stronger control over service promises, and continued investment in automation and data.
For networks operating at smaller scales or in other industries, the lesson is less about matching specific delivery speeds and more about the operating logic: shifting from long-haul optimisation to local, time-bound capacity management, and treating minutes not only as a service metric but as a design constraint across inventory, facilities, and cost-to-serve.
