D-Robotics and GigaAI Partner to Bring World Models to Edge Chips

In the robotics industry, many partnerships end with a handshake and a photo. The strategic cooperation between D-Robotics and GigaAI on September 14 is different: what matters is not the signing itself, but which assets each side brings to the table.

GigaAI contributes its world model, embodied foundation model, robot bodies, and experience in real-world deployment. D-Robotics provides its edge AI computing platform, algorithm toolchain, and robotics ecosystem.

The first integrated model is GigaBrain-0.7, running on the Sunrise S600 hardware base and incorporating GigaWorld's world generation capabilities. The goal is a deliverable, affordable edge embodied intelligence solution that integrates software and hardware.

D-Robotics and GigaAI have defined a clear division of labor across five links.

GigaWorld handles scene generation, action consequence prediction, policy evaluation, and failure sample retraining, providing an interactive, evaluable, and iterable virtual training environment.

GigaBrain acts as the edge embodied brain, responsible for natural language task understanding, visual spatial perception, task decomposition, skill routing, and edge decision-making. GigaWorld-Policy generates action candidates, outputs policies, and enables failure recovery.

The Sunrise S600 takes on edge multimodal inference, task scheduling, trajectory planning, state monitoring, and data collection. Robot bodies like Maker and Shiguang execute tasks in the physical world and feed data back into the model iteration loop.

Together, these five links form a closed loop from visual perception and language understanding to action generation, body execution, data feedback, and world model iteration. The partnership aims to fix the broken chain that has historically scattered models, chips, bodies, scene data, and application delivery across different players, making it hard to transfer model capabilities to real robots.