Operations First Ground Robot Delivery

This reveals that Coco is treating robot delivery as a logistics service first, not a science project. The hard part is not merely getting a robot to move, it is getting hot food from thousands of busy merchants to customers on tight time windows with few refunds and low labor cost. That pushes the company toward dense city routes, teleoperations software, fleet maintenance, and merchant workflows that work before full autonomy does.

• Coco built around real merchant operations in Los Angeles, Chicago, Miami, and Helsinki, where staff load the robot like a driver and the fleet updates live maps from every trip. That operating loop creates the data and edge cases needed to improve autonomy later, instead of starting in simplified environments with limited commercial value.
  1 sacra 2 doordash 4 doordash
• The contrast with Starship is concrete. Starship has strong autonomy and large scale, but much of its footprint is on campuses, where routes are simpler and universities subsidize the service. Coco is optimizing for dense urban food delivery, where bike lanes, sidewalks, road shoulders, merchant congestion, and customer wait times matter more than pure autonomous miles.
  1 sacra 3 sacra
• Serve shows the other side of the market, a well funded urban operator pushing harder on autonomy technology, including Level 4 certification, Gen 3 robots, and teleoperation software from Phantom Auto. That makes the category a race to combine software and fleet economics, not just a race to prove driverless movement.
  2 doordash 5 serverobotics 6 serverobotics

The category is heading toward hybrid systems where human oversight keeps service levels high while autonomy steadily removes labor from each trip. Companies that already run dense urban fleets with real merchant demand will have the best training data, the best unit economics, and the strongest position as robot delivery expands from meals into grocery, retail, and local logistics.