Coco Robotics has launched Coco 2, its next-generation autonomous delivery robot designed for urban logistics. The Los Angeles-based company says the platform marks its shift from human-guided operations to full autonomy. Coco 2 expands beyond sidewalks into bike lanes and roads where permitted, targeting city-wide goods movement for restaurants, grocers, pharmacies and retailers.
Highlights
- Coco 2 reduces delivery times by up to 50% over its predecessor, with three times longer uptime and improved weather resilience.
- The platform runs on NVIDIA Jetson Orin NX, processing navigation data on-device without relying on cloud connections.
- Coco currently serves over 3,000 merchants through Uber Eats, DoorDash and Wolt, with plans to scale to thousands of robots globally by year-end.
- The fleet’s AI draws from millions of miles of real-world sidewalk data, collected across cities including Miami, Chicago and Los Angeles.
From Sidewalk Delivery to Multi-Lane Operations
Coco originally focused on sidewalk-based food delivery. However, the Coco 2 platform extends operations into bike lanes and roadways where regulations allow. The company says this broader operating range cuts delivery times in half compared to the previous generation.
Additionally, increased uptime and durability raise daily order capacity while lowering cost per mile. Coco positions the platform as a general-purpose urban robotics system. The goal is to make instant delivery affordable at scale for any business, on any delivery platform and across a wide range of order sizes.
NVIDIA-Powered AI Stack Drives Navigation
Coco 2 relies on a full NVIDIA robotics stack for its autonomous capabilities. The robots use NVIDIA Omniverse libraries along with Isaac Sim and Isaac Lab frameworks to generate physically accurate simulations. These synthetic environments mimic real streets, pedestrians, vehicles and obstacles.
As a result, the AI can practice complex navigation scenarios before deployment. On the hardware side, the NVIDIA Jetson Orin NX processes sensor data locally on each robot. This enables real-time perception, route planning and maneuver execution without cloud dependency.
Amit Goel, head of strategic partnerships at NVIDIA, noted that scaling physical AI requires a loop between real-world data and high-performance edge computing. He said Coco is using NVIDIA’s full robotics stack to accelerate deployment of autonomous systems in complex urban settings.
Real-World Data Fuels Continuous Improvement
The Coco 2 platform draws on what the company calls the industry’s largest dataset of sidewalk robot operations. That data comes from millions of miles navigating diverse conditions, including the following:
- Flooded streets in Miami
- Heavy snow and freezing temperatures in Chicago
- Dense urban traffic in Los Angeles
CEO and co-founder Zach Rash explained that every mile driven makes the entire fleet smarter. He described a feedback loop between deployment, data collection and model advancement. This process enables the fleet to adapt in real time when entering new cities.
Scaling Toward Global Fleet Expansion
Coco currently powers autonomous delivery through three major platforms: Uber Eats, DoorDash and Wolt. The company serves more than 3,000 merchants and restaurants ranging from local businesses to national brands.
Building on that momentum, Coco plans to scale its fleet to thousands of robots globally by the end of 2026. Founded in 2020, the company has completed over 500,000 zero-emission deliveries across the U.S. and Europe. For more information, visit www.cocodelivery.com.
Sign up for our weekly emails to stay up to date on the latest news!
