Horizon Robotics / 地平线

Overview

Horizon Robotics is presented in the source as a physical-AI platform company supplying chips, algorithms, and system capabilities for intelligent vehicles and robots rather than manufacturing complete robot bodies.

Current Profile

The episode frames Horizon’s automotive concentration as the first commercial stage of a longer robotics strategy. Mass-produced cars provide a mature supply chain, large deployment volume, engineering feedback, data, and cash flow; the company then aims to reuse chip design, perception, decision, control, and data-loop experience in robot platforms while acknowledging that multi-joint bodies, balance, contact, and manipulation require new control data and hardware capabilities.

Key Characteristics

• Combines algorithm and chip development to bridge fast model change and multi-year silicon cycles.
• Treats vehicles as an early, scaled physical-AI product and robots as a broader later platform opportunity.
• Supplies “brain,” “small-brain,” development-tool, and system capabilities rather than complete robot bodies.
• Positions itself as a horizontal platform serving multiple automakers and robot developers.
• Uses developer tools, sample projects, university collaboration, and competitions as an ecosystem-seeding strategy.
• Emphasizes long-cycle product definition and advance trend judgment over reacting to current hype.

Evidence

Hardware–software co-design

• 特番｜从会开车到会做事：AI如何进入真实世界 attributes to Horizon guests the claim that chip definition must anticipate model needs three to four years ahead while respecting power, heat, space, and real-time constraints.

Vehicle-first platform route

• 特番｜从会开车到会做事：AI如何进入真实世界 presents cars as the scaled entry point that can build shipments, cash flow, data loops, and production experience before expansion into robots.

Robot and developer platform

• 特番｜从会开车到会做事：AI如何进入真实世界 says the robot offering covers high-level decision, low-level control, feedback, tools, examples, and support for varied bodies, with universities and young developers treated as early ecosystem participants.

Qualifications

• The profile is based on a sponsored or company-centered live podcast without automaker, regulator, customer, or independent researcher cross-examination.
• The claimed platform advantages, future market split, transfer from cars to robots, and commercial timing are strategic statements rather than independently validated outcomes.
• Vehicle intelligence does not automatically solve robot manipulation, balance, body reliability, action-data, or use-case economics.

What Changed

• Created the profile from the episode’s account of Horizon’s vehicle-first physical-AI platform strategy.
• Kept market-share, timing, and technical-advantage claims explicitly company-attributed.

Relationships

• Physical AI - broader category Horizon uses to connect vehicles and robots.
• Physical Intelligence System Stack - system-stack model matching Horizon’s chip, algorithm, data, control, and tooling position.
• Embodied Intelligence Vehicle Path - cars provide Horizon’s first scaled physical-world deployment route.
• Autonomous Driving Responsibility Boundary - responsibility framing shapes the current assisted-driving market Horizon serves.
• Robot Control Data Scarcity - robot expansion introduces action-data and embodiment constraints beyond vehicle experience.
• Small Brain Action Layer / 小脑（Action Policy） - low-level motion-control layer corresponding to Horizon’s “small-brain” platform description.