GigaAI Raises $350M for Embodied AI & World Models

Chinese AI innovator GigaAI has successfully raised approximately $350 million USD across three funding rounds in just six months, underscoring significant investor confidence in its pursuit of advanced embodied artificial intelligence. The company, established in 2023, is rapidly distinguishing itself in the competitive physical AI sector by focusing on foundational models that bridge the gap between digital understanding and real-world robotic action.

The substantial capital infusion, comprising Pre-Series B, Series B1, and Series B2 rounds concluded by June 2026, highlights a pivotal moment for GigaAI. The latest tranche, Series B2, alone accounted for roughly $84 million USD, signaling strong momentum. This funding is earmarked to accelerate the development and deployment of its proprietary 'world model' platform, GigaWorld, and its core 'brain' model, GigaBrain, which are designed to imbue robots with enhanced decision-making capabilities and environmental awareness.

At the recent World Robot Conference (WRC2026), GigaAI unveiled GigaBrain-0.7, a next-generation foundational model for embodied AI. Unlike conventional models that directly translate perception into action, GigaBrain-0.7 integrates 'understanding,' 'prediction,' and 'action' systems. Its predictive engine leverages world models to anticipate post-action scenarios, enabling more sophisticated and adaptive robotic planning. The model's extensive pre-training, exceeding 37,000 hours of embodied AI data, emphasizes its adaptability across diverse tasks, environments, and robotic hardware, a critical factor in the rapidly expanding robotics market, which is projected to reach hundreds of billions in the coming decade.

Further demonstrating its technical prowess, GigaAI, in collaboration with Tsinghua University, introduced GigaBrain-WBC-0.5. This advanced whole-body control model not only generates subsequent movements but also forecasts the impact of environmental factors and the robot's own state on its actions. This allows for robust performance even under challenging conditions, such as uneven terrain or external interference, mitigating risks of falls or erratic behavior. The company also showcased the portability of its models by successfully transferring control functions from Unitree's humanoid robot 'G1' to GigaAI's own 'Maker L01,' proving the potential for shared foundational motor control capabilities across different robotic platforms.

The company's capabilities were vividly illustrated through a table tennis demonstration, 'GigaBrain-PingPong,' where a robot engaged in a lengthy rally against a human opponent. This high-speed task demands rapid object recognition, trajectory prediction, and intricate whole-body coordination, serving as a rigorous testbed for GigaAI's world model integration and real-time control systems. Beyond providing foundational models to external partners, GigaAI is also venturing into direct product deployment.

GigaAI has already launched consumer and industrial robots, including the 'ShiGuang S1' household robot, capable of autonomous domestic chores, which secured an order for 100 units from Hubei Science Technology Investment Group. The industrial variant, 'Maker H01,' is designed for manufacturing tasks like material handling and assembly. The company's strategic approach, combining foundational model development with practical application, positions it as a key player in the global push towards more intelligent and autonomous robotic systems.