“The adoption of physical AI is slower and more linear compared to large language models because it involves safety considerations and real-world physical machines.”