The last mile is the expensive one

Delivery cost is dominated by the last mile. Moving a container across an ocean costs cents per item. Getting that item from a local depot to one door costs euros, because it involves a vehicle, a person, traffic and a building with a door code. Every automation effort in logistics eventually points at this.

Two very different machines

Drones ignore the road network entirely, which is their whole advantage. A flight of eight kilometres takes minutes regardless of traffic, and the economics work best exactly where road delivery is worst: rural areas, islands, hospitals moving blood and samples between sites, and dense cities at rush hour. Payload is the limit. Most systems carry under two and a half kilograms, which covers a prescription, a meal or a small part and excludes most parcels.

Sidewalk robots travel at walking pace on pavements, carry more, cost far less and are limited to roughly a kilometre or two around a depot. They work in campuses, suburbs with good pavements and dense urban cores with slow traffic, and they fail where kerbs are broken, crossings are wide or snow is on the ground.

Why regulation is the story

Flying beyond visual line of sight over people is the regulatory hurdle that defines drone delivery. The rules require detect-and-avoid capability, reliability evidence and, in most jurisdictions, an approval per operation or per corridor. Progress has come as regulators granted broader permissions in defined areas, and the geography of drone delivery is a map of those permissions rather than a map of demand.

Sidewalk robots face a different question: who owns the pavement. Municipalities have variously welcomed them, capped their numbers, required a remote operator per robot or banned them after complaints about blocked crossings and wheelchair access. That is a local political decision, so the rollout is city by city and reversible.

The economics

A drone delivery makes sense when the alternative is a van driving a long way for one item. Medical logistics is the clearest case: a blood sample that must reach a lab within an hour is worth far more than the flight costs. Retail delivery of a coffee is not, unless the customer pays a premium that survives the novelty wearing off.

Sidewalk robots are cheaper per delivery but slower, which means each unit completes few deliveries per hour. The model only works at high density with short trips, and it competes not with a van but with a person on a bicycle, who is faster and more flexible.

The honest position in 2026 is that both are viable in specific niches and neither has displaced a delivery van at scale.

The noise problem

Drone delivery has a public acceptance issue that engineering has partly, not fully, solved. Multirotor aircraft are loud in a frequency band people find irritating, and a neighbourhood with regular overflights complains. Quieter propellers and higher cruise altitudes help. A city with hundreds of daily flights over houses would be a genuinely new kind of urban noise, and that debate has not really started.

What to watch

Watch for a regulator granting a broad beyond-visual-line-of-sight permission across a whole metropolitan area rather than a corridor. Watch for a delivery operator publishing cost per drop rather than flight counts. And watch medical logistics, where the value per delivery is high enough that the economics never needed a breakthrough.