[Goldman Sachs: World Models Could Become the New Engine for Future AI Infrastructure, Adding New Computational Demands] According to a report by Jinse Finance on July 6, Goldman Sachs emphasized that 'world models' could become the second engine driving future AI infrastructure demand. Unlike large language models that primarily process text and images, world models aim to understand causal relationships within physical and social systems, such as simulating friction, material behavior, supply chain reactions, policy impacts, or corporate competitive strategies. Physical world models will support robotics, logistics, autonomous driving, and industrial design, while social world models could be applied to strategic simulations, investment decisions, governance stress testing, and policy scenario analysis. Goldman Sachs believes that world models are not a replacement for large language models but rather add new computational demands. If their development progresses faster than expected, current investment forecasts surrounding computing power and electricity may still be underestimated.