Walmart’s latest disclosures show an inventory engine rebuilt around automation, AI and store-based fulfilment, with operating income now growing faster than sales and e-commerce running profitably at scale.
In Brief
- Automation in distribution and fulfilment centers, coupled with handheld-supported stores, is compressing delivery times while easing pressure on labour and inventory growth.
- AI assistants are raising basket size and shaping how orders flow through the network, tightening the link between digital demand and physical capacity.
- Profit mix from advertising, membership and marketplace activity is underwriting heavy investment in automation without eroding overall margins.
The Operating Break: An Inventory Engine Built For Speed and Cost Control
Recent results position Walmart as a live case of what an AI-enabled inventory engine looks like at scale. Revenue grew 4.9% in constant currency in the fourth quarter, e-commerce grew 24%, and adjusted operating income rose 10.5%. At the same time, inventory increased 2.6%, about half the rate of sales growth for the full year.
The structural change is not a single technology but a different way of running inventory and fulfilment:
- automated distribution and fulfilment centers now handle a large share of inbound and online flows;
- stores and clubs are used as forward nodes, supplying both walk-in and online demand;
- AI tools sit on top of this footprint, shaping demand and routing orders.
For any large network, the notable point is that these elements are described as one portfolio, not separate tracks. Automation and AI are no longer side projects; they are the logic behind how stock is positioned, how orders are served, and how costs are managed.
How The Engine Works In Operational Terms
Walmart reports that about 60% of its U.S. stores now receive some freight from automated distribution centers and around 50% of e-commerce fulfilment center volume is automated. That level of penetration suggests a phased replacement of legacy conveyor systems with higher-automation solutions that can ship store-ready loads. In practice, this means:
- more product arriving at stores in a configuration that reduces in-store handling;
- more consistent flow into fulfilment centers, supporting predictable online picking.
More than one million associates in the U.S. now use handheld devices with computer vision to map inventory. That detail matters because it converts in-store stock from a static retail layer into addressable capacity for online orders. The company states that roughly 35% of U.S. store-fulfilled digital orders are delivered in under three hours. To reach that, inventory in stores must be both visible in systems and pickable within tight time windows.
On top of this physical capacity sits a digital layer. Agentic AI, branded internally as Sparky, is used to build baskets from customer intent. The assistant is used by roughly half of app users and is associated with about 35% higher average order value versus non-assisted orders. That higher density changes the economics of picking and delivery: more units per order and more revenue per stop across the same network and labour.
At system level, this kind of configuration usually requires:
- common item and location data so that stock in any node can be treated as part of one pool;
- an order-management layer that allocates each line to a node based on service promise, local stock, and available capacity;
- service thresholds that are channel-agnostic, so it is possible to choose between store, distribution center or fulfilment center dynamically.
Walmart’s disclosures about building platforms once and deploying them across banners and countries indicate that these enabling layers are being standardised, not rebuilt market by market.
Inventory Discipline Designed Into The Model
Inventory growth at roughly half the rate of sales, combined with e-commerce growth of nearly 25% to more than 150 billion dollars, signals deliberate constraints around owned stock. Two levers stand out.
First, marketplace expansion changes the balance between owned and third-party inventory. Selected marketplace categories grew above 40% year on year, and 52% of marketplace sellers use Walmart Fulfillment Services. Instead of carrying this stock, Walmart provides fulfilment and customer access, earning fees while shifting working-capital and markdown risk to sellers. For other companies, the principle is a clearer separation between capacity provision and inventory ownership.
Second, AI and better measures show up in planning behaviour. Management has described entering the year with ‘clean’ inventories after buying cautiously in seasonal ranges and accepting that some categories may have been bought light. That posture only works if the network can replenish quickly when demand overshoots the conservative plan. Automated nodes and store visibility are the enablers.
In operational terms, this typically implies:
- shorter planning cycles, with more frequent updates to allocations based on realised demand;
- tighter thresholds for seasonal and promotion buys, backed by a plan to refill from upstream nodes rather than front-loading;
- clear accountability for markdown exposure by node and category.
The combination of marketplace mix and constrained owned stock makes it possible to grow sales and e-commerce volume without allowing inventory to expand at the same rate.
Speed as a Network Design Parameter
Fast delivery is not being treated as a side benefit but as a primary service level. The company reports that it can serve 95% of the U.S. population within three hours and that many express orders are delivered in under 30 minutes, with average express delivery under one hour. Fast-delivery usage, defined as orders under three hours, grew about 60% year on year.
Network design choices flow from that promise:
- inventory has to sit physically close to demand, in stores and clubs as well as in fulfilment centers;
- last-mile capacity must be dense enough to make sub-three-hour routes viable;
- upstream automation must keep stock flowing without manual bottlenecks.
Sam’s Club is following the same pattern on a smaller scale, with around 60% of members able to get delivery in three hours. That suggests that using club locations as fulfilment points is now a standard role, not an experiment.
From a planning perspective, this means treating stores and clubs as multi-role assets with explicit capacity allocation between walk-in traffic, click-and-collect, and delivery. It also requires labour scheduling and pick-path design that can absorb sharp peaks in express orders without undermining the base shop.
Funding The Model: Profit Mix and Capital Peak
Building this inventory engine carries cost. Tariff-related costs and higher claims created about a 300 basis point headwind to adjusted operating income in the last fiscal year. At the same time, management indicates that capex for FY 2027 will be around 3.5% of sales and that the company is at the peak of annual spending on supply chain automation and store remodels.
Walmart is offsetting these pressures through a shift in profit mix. Advertising businesses grew 46% in the last fiscal year to 6.4 billion dollars, and membership fees exceeded 4.3 billion dollars. In the most recent quarter, around one-third of operating income came from advertising and membership. These revenue streams carry little or no inventory and minimal physical overhead relative to core retail.
On the cost side, SG&A leverage in the fourth quarter is attributed partly to supply chain automation and productivity benefits that have been ‘implementing for years. U.S. e-commerce has been profitable in each of the last four quarters, with roughly double-digit incremental margins. Operating income for FY 2027 is guided to grow 6–8% on constant currency sales growth of 3.5–4.5%.
Across industries, the implication is that high fixed investment in automation and fast service can be sustained if there are parallel efforts to:
- build asset-light profit pools that do not require more inventory or labour;
- capture measurable productivity improvements once automation is rolled out at scale;
- cap capex as a percentage of sales after the initial build-out period, relying on reinvestment of growing cash flows rather than constant expansion of the asset base.
Constraints That Still Apply
The model still faces external constraints. Drug pricing legislation in the U.S. is expected to create a headwind of about 100 basis points to full-year sales in health-related categories, with roughly 30 basis points in the first quarter of effect, even if margin is protected through manufacturer rebates. Tariff environments remain ‘bumpy’, with management acknowledging the need to manage commodities as they move up and down. Government benefit disruptions have created short-term demand swings that need to be absorbed through flexible inventory and labour planning.
These factors reinforce a simple point: automation and AI reduce unit handling and improve accuracy, but they do not remove volume volatility or regulatory exposure. Planning systems still need to accommodate sudden shifts in basket composition and traffic by channel.
What The AI-enabled Inventory Engine Enables Across Sectors
Taken together, Walmart’s disclosures outline a supply chain that can support:
- profitable e-commerce growth at scale, with positive incremental margins;
- sub-three-hour delivery to most of a large national population, using existing physical assets;
- inventory growth below sales growth, supported by marketplace mix and cautious seasonal buying;
- margin protection despite pricing investments and regulatory headwinds, financed by high-margin services.
For other organisations, the cross-industry implication is less about copying specific tools and more about the sequence: automate core flows, turn local assets into fulfilment nodes, use AI to link demand to capacity, and rebalance profit mix so that fast service and heavy capital spend do not permanently depress returns.
