Nvidia and Hitachi are expanding the use of physical AI beyond robots to social infrastructure such as trains, power grids and factories, citing the need for higher precision and stronger verification than generative AI. Hitachi is strengthening its infrastructure AI business around HMAX by Hitachi. Executives described accuracy requirements in terms of “nines” and stressed safety risks from malfunctions. The companies highlighted phased development using training, simulation-based validation and deployment, as well as the growing use of synthetic data and the need for traceability and human override.

NC AI said on Monday it demonstrated a World Foundation Model, or WFM, for robot intelligence as it seeks to enter the global physical AI market. The company said it trained and validated the model using its own research infrastructure and recorded success rates indicating practical applicability. It said its approach generates actions from latent-space information before video generation, improving efficiency and accuracy. NC AI said it trained the WFM using 25 percent of GPU resources needed to fine-tune global state-of-the-art models and achieved up to an 80 percent success rate on key tasks.