Serve Robotics, a company known for its self-driving sidewalk delivery robots and a former Uber spinout, is broadening its business by launching a new division called Physical AI Services. Announced alongside its third-quarter 2026 earnings, this venture aims to become a technology provider for other robotics firms by offering connectivity and data tools developed from Serve's own robotics operations. The company has already signed clients such as Bedrock Robotics, which focuses on autonomous construction equipment, and is collaborating with T-Mobile to enhance AI computing accessibility for robots.

The core idea behind Physical AI Services is to solve operational challenges for robots working in dynamic human environments. Serve Robotics CEO Ali Kashani likens the new business to an "AWS moment," where the company opens up its proprietary technology stack for others to build upon. This new approach complements Serve’s existing delivery robot business without replacing it. Currently, Serve produces two robot models, including a hospital-assistant robot acquired through its purchase of Diligent Robotics.

Physical AI Services will initially market three products designed to improve robot connectivity and intelligence. Serve Link ensures high-performance, reliable communication between robots and remote AI systems, even over unreliable networks. A data curator tool helps customers identify rare but important data scenarios for improving machine learning models, while a foundation-model data catalog, built from Serve’s delivery robots’ real-world experience, aids AI in interpreting complex human settings. This strategic shift stems from years of building integrated robotic applications and aims to fuel further innovation across industries like agriculture and warehousing.

Serve Robotics is also transitioning its sidewalk delivery business from solely working with Uber to partnerships with companies like DoorDash and Wonder, signaling a broader client focus. The Physical AI Services launch represents a move to monetize and share the advanced technologies developed internally, supporting a growing robotics ecosystem. By offering connectivity and data infrastructure similar to how Amazon Web Services powers cloud computing, Serve intends to become a foundational platform for physical AI development.