Serve Robotics, a developer of autonomous sidewalk delivery robots, is expanding beyond its own delivery operations to supply technology to other companies building robots.
The company, which emerged from Uber’s 2021 acquisition of Postmates and went public in 2024, announced Thursday the launch of a new division called Physical AI Services.
Specifically, Serve Robotics will provide customers with its connectivity and data tools. Bedrock Robotics, a startup building autonomous systems for construction equipment, is already a customer, according to Serve. The company also announced a partnership with T-Mobile to deliver AI computing power closer to robots, alongside its third-quarter earnings release on Thursday.
“We are solving the problem of robots in human spaces,” said Ali Kashani, CEO and co-founder of Serve Robotics, in an interview with TechCrunch.
Serve Robotics currently manufactures two types of robots: a sidewalk delivery robot with googly eyes and a hospital robot acquired through its purchase of Diligent Robotics.
Those operations will continue, Kashani emphasized in a recent interview. Instead, he sees an opportunity to sell the technology developed for Serve’s own robots to other companies creating robots for farms, warehouses, and other uncontrolled environments.
“I think of it as an AWS moment,” he said, “where we take the technology we built for ourselves and make it available for others to build on,” adding that this would not have been possible without developing its own robot fleet.
The expansion comes as Serve’s sidewalk delivery business grows but shifts away from Uber toward other partners, including DoorDash and Wonder.
The new Physical AI Services division will offer three products. Serve Link connects robots in the field to AI systems running in data centers, allowing companies to access more powerful AI models without placing all that computing power onboard.
“Any robotic application generally needs to connect robots to some offline intelligence, whether for data collection, exception handling, or otherwise,” said Kashani. “The connectivity piece must be high performance. You need to manage highly unreliable networks and still maintain reliable connections, and that is what we solve.”
The division is also launching a data curator tool that helps companies identify edge cases in the vast amounts of real-world data a robot collects, which can then be fed back into training models to improve performance. Its third product is a real-world, foundation-model data catalog built from Serve’s own sidewalk robots, designed to teach AI models how to understand complex human environments.
Kashani said he has been considering this idea—a playbook he learned from Amazon—for years.
“When technology is still nascent, you have to build the end-to-end application because that’s the only way to know what to build,” he said. “But as the technology matures, more people will want to build similar things. You want to open your technology so others can build with you and on your platform.”
