As of November 7, 2026, Amazon has restarted its own fresh grocery delivery in Germany, three cities at a time. The service went live in parts of Frankfurt, Cologne and Munich, with fruit, vegetables, meat, poultry, fish, dairy, eggs, bakery and frozen food slotted into the ordinary same-day order rather than a separate app. Prime members pay nothing above 29 euros inside a four-hour window and 3.99 euros below it, while non-members pay 9.99 euros. The relaunch is a logistics bet, and the number that will decide it is not the delivery fee but whether the item a shopper wants is actually pickable in the warehouse that serves their postcode.
Fresh food returns to the same order, not a separate app
The detail that matters operationally is who picks the order. Amazon is storing, picking and packing perishables itself, in refrigerated zones added to existing distribution centres, rather than listing a grocer's shopfront on its marketplace. That is a different business from the partner model it ran after shutting Amazon Fresh in December 2024: the platform now owns the cold chain and the pick, which means it also owns the failure modes. The range sits in the standard search results and the shopping app, so a shopper looking for a phone case can add milk without noticing a channel change.
Why the basket threshold decides the model
A 29 euro free-delivery threshold is not a pricing gimmick; it is a statement about the basket the operation is built to serve. Below that threshold the fee ladder climbs steeply, from 3.99 euros for Prime members to 9.99 euros for everyone else, and a two-hour slot adds 2.90 or 4.90 euros for members and 10.99 euros for non-members. Those numbers push the shopper toward a larger, planned basket rather than a single missing ingredient, which is the opposite of how quick commerce trained the market.
| Order type | Fee for Prime members | Fee for non-members | What the operation has to get right |
|---|---|---|---|
| 29 euros or more, four-hour window | No charge | 9.99 euros | Pick accuracy across a wide basket |
| Under 29 euros, four-hour window | 3.99 euros | 9.99 euros | Cost per order on small baskets |
| Two-hour slot, 29 euros or more | 2.90 euros | 10.99 euros | Slot capacity and cold-chain timing |
| Two-hour slot, under 29 euros | 4.90 euros | 10.99 euros | Labour scheduled against a narrow window |
Cold chain, pick accuracy and the assortment gap
Refrigerated space is a capacity decision
Adding chilled zones to an existing fulfilment centre is cheaper than building a dedicated dark store, but it ties the fresh range to the geometry of a building designed for ambient parcels. Chilled capacity per site becomes the real constraint on how many fresh orders a postcode can absorb in a day, and it is invisible to the customer until the slot disappears from the app. That is why the expansion plan to further metropolitan areas in the coming weeks will be decided by refrigerated square metres rather than by marketing.
An incomplete range breaks the trip
Fresh food is high-frequency and low-margin, and the economics only work when a single order covers most of a household's list. A range that carries milk, eggs and vegetables but misses the specific brand or the organic line a shopper wants converts a planned basket into a top-up order, which is exactly the order the fee ladder punishes. Assortment coverage therefore matters more here than in general merchandise: a missing line does not cost one item, it costs the whole trip.
What to measure before the next city opens
The first weeks of a three-city launch produce more useful data than any forecast, provided the measurement is set at the right level. The unit has to be the fulfilment site and the SKU, not the country, because a national availability figure averages away the chilled zone that ran short and the postcode that never had the line listed. A country-level number can look healthy while two of the three launch cities cannot complete a weekly basket.
The second field is the one most often skipped, because a substitution looks like good service in the short run. In fresh food it is not: a swapped yoghurt or a different cut of meat turns a convenience order into a complaint, and the customer who complains about a substitution rarely reorders the category from the same channel. Substitution acceptance is a better early signal than delivery time, because delivery time can be bought with capacity while acceptance has to be earned line by line.
The lesson from the first attempt
Amazon Fresh ran in Germany from 2017 and never grew beyond Berlin, Potsdam, Hamburg and Munich before it was discontinued in December 2024. The new attempt differs in one structural way: fresh items now travel with non-food parcels, so the last mile is shared and the delivery cost per order falls. That is a genuine advantage, and it is also why the relaunch should be read as a test of assortment depth rather than of demand. Demand for online groceries in Germany is not in question; the question is whether a warehouse built for parcels can keep a chilled shelf deep enough to complete a weekly basket. Until that is answered in three cities, the fee table is the least interesting number in the release.
FAQ
What changed compared with the Amazon Fresh years?
Fresh items now travel inside the ordinary same-day order alongside non-food parcels, and Amazon stores, picks and packs perishables itself in refrigerated zones added to existing distribution centres. The shared last mile is the structural difference from the 2017 to 2024 attempt.
Why does the 29 euro threshold matter so much?
It defines the basket the operation is built for. Below it the fee ladder rises steeply, which pushes shoppers toward planned larger baskets rather than single-item top-ups, and that changes picking, slot capacity and cost per order.
What is the biggest operational risk in fresh delivery?
Assortment coverage and pick accuracy. A missing line costs the whole trip rather than one item, and a substitution in a chilled order turns a convenience purchase into a complaint that rarely repeats.
How should the launch be measured?
By site and by SKU: listed lines against the intended list, availability by site, pick accuracy and substitution acceptance per shift, and basket value against the free-delivery threshold.
