Ground robots win on reliability
Zach Rash & Daniel Singer, CEO & CBO of Coco Robotics, on why ground delivery beats drones
This is the core moat in robot food delivery, because the winner is the company that can make thousands of small things go right fast enough that refunds stay rare and merchants keep trust. A hot meal has maybe 15 to 20 minutes to arrive in good condition, so reliability depends on routing, curb cuts, crossings, battery life, teleoperator handoffs, merchant loading speed, and city specific street behavior all working together on every trip.
-
Coco’s system is built for dense city runs, not simple campus loops. Its robots move on sidewalks, bike lanes, and road shoulders, run about 1 to 2 miles, switch to a remote operator in under 300 milliseconds when needed, and are meant to preserve a tight delivery window for prepared food.
-
The economics are unusually sensitive to failure. In the interview, management frames human delivery as roughly 95% to 98% reliable, while autonomous service has to push toward 99.9% because a few late, missing, or cold orders can wipe out contribution profit through refunds and customer churn.
-
That is also why urban deployment matters more than headline robot count. Starship has scale, with 2,000 plus robots and 8 million plus deliveries, but much of that density is on campuses. Coco has leaned into harder city terrain, where reliability is tougher to earn but more valuable to DoorDash, Uber Eats, and large restaurant chains.
The next phase is turning this operational reliability into software leverage. As fleets grow and more edge cases are learned, teleoperators can oversee more robots, cost per delivery falls, and the strongest urban operators can expand from hot food into groceries, pharmacy, and small parcel delivery without losing the service level that makes the model work.