As of 8 November 2026, Walmart has signed a multiyear agreement with Zipline to add 300 drone delivery sites across the United States, building on a programme that has completed more than 2 million deliveries since the two companies started working together in 2021. Deliveries can arrive in as little as 30 minutes where available, with tens of thousands of eligible items, and Walmart says 90 percent of Americans live within 10 miles of one of its stores. For teams working on on-demand delivery, the interesting part is not the aircraft. It is that the store network is being turned into a minute-scale fulfilment grid.
What the 300 sites actually change
Supermarket News reported that the agreement builds on a partnership that began in 2021 and that the expansion could extend beyond the 300 sites. Greg Cathey, Walmart senior vice president of ecommerce fulfilment transformation, framed it as proximity: the stores put the retailer close to the communities it serves, and Zipline extends that proximity to more customers.
Read carefully, this is an assortment decision dressed as a delivery decision. A drone carries a short basket, roughly a handful of items under eight pounds, and delivers it in minutes. That rules out a weekly stock-up and rules in the forgotten ingredient, the missing frosting, the new pack of tennis balls. The store is no longer only a place where shoppers walk aisles. It becomes a picking node with a very short radius and a very narrow eligible list.
The five-tier ladder Walmart described makes the segmentation explicit: drone delivery measured in minutes, 30-minute-or-less, Express within an hour, On-demand in as soon as three hours, and scheduled windows. Each tier has a different basket size, a different cost structure and a different set of items that can realistically be promised.
Store density is the cost advantage
Two fulfilment models, one constraint
There are two ways to make a minutes-level promise. One is to build dedicated nodes close to demand, which is the quick-commerce model: small urban sites, narrow assortment, rent per node and inventory depth as the main cost drivers. The other is to use an existing supermarket footprint, which is Walmart's route: picking labour and drop distance dominate, and the failure that shows up first is substitution and fill rate rather than assortment gaps.
Both models converge on the same constraint. The promise is only as good as the inventory positioned near the shopper. A drone that arrives in 19 minutes with the wrong item, or with a substitution the customer rejects, is not a fast delivery. It is a fast complaint.
FreightWaves noted that Walmart's US ecommerce sales grew 24 percent in the second quarter, with deliveries fulfilled from local stores up 43 percent. Those two numbers belong together. Growth in store-fulfilled delivery is what makes a store-based drone network economically legible: the picking infrastructure, the staff and the inventory are already paid for by the ground business.
Reading the 37 percent
Fee-based fast deliveries reached a record 37 percent of store-fulfilled deliveries in the quarter to 31 July, according to figures reported earlier. That number matters because it shows customers will pay for speed when the promise is credible. It also sets a ceiling question: if a meaningful share of fast deliveries carries a fee, the economics of a drone tier depend less on delivery cost per order and more on how many orders the same store can absorb in a peak hour.
Which stores qualify for minute-scale delivery
Not every store in a chain can host a minute-scale tier, which is why store selection has become the bottleneck rather than aircraft supply. Three filters do most of the work, and they are ordinary trade area filters rather than aviation ones. Each one can be answered from data a retailer already holds.
The first is catchment density, the core question in any trade area review. A store whose ten-mile ring contains enough households with the right basket profile can justify the fixed cost of a launch site. The second is backroom capacity: drone orders are picked by people, so the store needs slack in picking during the same hours when walk-in traffic peaks. The third is item overlap, meaning how much of the drone-eligible list is already a fast mover in that specific store, because slow movers in a narrow assortment produce substitutions rather than satisfaction.
| Filter | What to measure | Failure it prevents |
|---|---|---|
| Catchment density | Households in the delivery ring with matching basket profile | Fixed cost with too few orders |
| Backroom capacity | Picking slack during peak walk-in hours | Late picks and cancelled orders |
| Item overlap | Share of eligible list that is already a fast mover | Substitution and refunds |
| Weight profile | Share of baskets under the payload cap | Orders that cannot be flown |
The weight profile is the filter most often skipped. A store can look ideal on density and still fail if its typical basket sits above the payload limit, because the drone tier then serves only a thin slice of demand and never reaches the order volume that justifies the site. Screening for high-potential stores therefore means screening baskets, not postcodes alone.
A checklist for store-level delivery capacity
The practical work is to convert these filters into a repeatable store selection review, so that site selection stops being a negotiation and becomes a calculation. The review below is the sequence we use when a retailer asks which stores should carry a new delivery tier.
FAQ
Is this a drone story or an assortment story?
Mostly assortment. The aircraft sets the payload and the time window, but what actually sells is a narrow list of fast-moving items that customers need within the hour.
Why does store density matter more than aircraft range?
Because cost per drop is driven by distance and by whether the item is already on a nearby shelf. A dense network shortens both, and it avoids building new nodes from scratch.
What should a brand track first?
Fill rate and substitution rate inside delivery orders, then whether the brand holds the default reorder slot for its replenishment items. Installs and app downloads say little about either.
Does a minute-scale tier cannibalise scheduled delivery?
It can, which is why the tiers are priced and merchandised differently. The risk is not cannibalisation but promising minutes for baskets that should have been scheduled.
