Chipmaking giant Nvidia is investing $10 million in Uber spinout Serve Robotics, funds the startup will use to further expand its sidewalk delivery robot service outside Los Angeles and San Francisco.
The investment – Nvidia's first in the sidewalk delivery space – is part of a long-term collaboration that will see the two companies work together toward advancing their own wheelhouses of robotic technology.
"Nvidia's investment is rooted in a longtime partnership with Serve using various technologies from us, from the edge to cloud-based technologies," Gerard Andrews, senior product marketing manager of robotics at Nvidia, told TechCrunch. "What we're excited about is the prospect of working closely with Serve to push the limits of what is possible on last-mile delivery problems."
Serve's robots, which the startup says are capable of operating in specific geofenced areas without a remote operator for safety, currently rely on Nvidia's Jetson edge AI platform, the hardware, or compute module, that sits inside the robot and powers autonomous movement. The startup also plans to use Nvidia's perception and mapping tools, which will help its robots understand where they are in the real-world environment and where they need to go.
Like most autonomous vehicle companies, Serve tests its models in simulation before they hit the roads, which, like testing in the real world, requires scores of data and photorealistic maps of cities. For that, Nvidia offers synthetic data generation tools for training perception models.
Offerings such as these come under a suite of tools from Nvidia, dubbed Isaac, that provide robotics developers with a range of software technologies from simulation to robotic fleet management. Nvidia hopes to use lessons from its partnership with Serve to improve its technology in the nascent robotic space, said Andrews.
"We see ourselves as a company that's leading with autonomy and scaling real autonomous robots out in the real world," Ali Kashani, co-founder and CEO of Serve, told TechCrunch. "Nvidia is one of the most critical companies to the robotics space as a whole, and they're also investing in the tools, so it just makes sense for us to work close together given this is a developing space."
There are different approaches to commercialization within sidewalk delivery being trialed now. Companies like Coco and, until recently, Tortoise, have leveraged remote operators to drive robots to their destinations, giving them a quicker and easier path to market than aiming for full autonomy.
Serve is choosing to go the more technically challenging route to full self-driving from the get-go, which means it needs serious compute power to handle the processing of data in real time.
"The sidewalk is more chaotic than the street," said Kashani. "We have even interacted with goats. The randomness of what you see on a sidewalk is an order of magnitude higher than the street. If you think about it, on the street cars have a distinct set of actions – they change lanes, there's braking and accelerating. ... When you come to the sidewalk, anything can happen anytime, and you need to be ready for it. So in an interesting way, it's actually more challenging. The advantage, of course, is that things happen slower and you have more time to react."
This isn't the first strategic investment Serve has seen in recent months. In December, the company raised $13 million in an expanded seed round, bringing on investors that are poised to help the company on its path toward commercialization, such Delivery Hero-backed DX Ventures, 7-Eleven's corporate VC arm 7-Ventures, and Uber, which will be tapping Serve as a delivery service partner for Uber Eats.