Cutting-edge large-scale models are gradually becoming the cognitive "brain" of embodied robots. Improved zero-shot and few-shot generalization capabilities are expected to accelerate the transition of robots from demonstrations to practical deployment. However, the competitive barriers in the industry are also changing: as model capabilities advance, the importance of rapidly deploying edge models, optimizing computational efficiency, and co-designing hardware and software is increasing. Goldman Sachs believes that the scaling up of physical AI will drive the growth in demand for multimodal real-world data. Advanced 3D spatial sensors, high-resolution haptic arrays, and data acquisition processes are expected to benefit first, while data quality remains a key factor restricting model training. Application opportunities are not limited to humanoid robots; collaborative robots and the automation transformation of existing factories may see practical applications emerge earlier. In terms of stock selection, Goldman Sachs prefers supply chain companies with reasonable valuations, solid fundamentals, and growth potential in AI robotics.
