Generative AI holds significant potential to advance the Internet of Things (IoT) by enabling new applications and improving existing ones. This article explores the vision, applications, challenges, and opportunities of integrating Generative AI into IoT systems. Generative AI can generate high-quality data, perform complex tasks with human-level performance, and generalize well to new tasks, which are essential for IoT applications such as mobile networks, autonomous vehicles, the Metaverse, robotics, healthcare, and cybersecurity. In mobile networks, Generative AI can simulate network data to predict bottlenecks and optimize resource allocation. In autonomous vehicles, it can enhance user interfaces and improve communication between vehicles and occupants. In the Metaverse, it can create immersive virtual environments. In robotics, it can improve interaction capabilities and adaptability. In healthcare, it can assist in processing medical data and generating reports. In cybersecurity, it can enhance security measures and predict threats. However, several challenges hinder the deployment of Generative AI in IoT. These include the lack of generative models for IoT data, the need to deploy large models on memory-constrained devices, the challenge of generating content with low latency, model adaptation and personalization, the need for IoT-based AI agents, privacy and security threats, and the lack of development tools and benchmarks. Opportunities to address these challenges include developing generative models for IoT data, model compression, runtime optimization, edge-cloud collaboration, efficient fine-tuning, designing IoT-based AI agents, federated learning, and creating IoT-oriented development tools and benchmarks. These opportunities aim to enable Generative AI for IoT, making it more efficient, secure, and adaptable to diverse IoT environments. The article concludes that Generative AI has the potential to revolutionize IoT applications, and further research is needed to fully realize this potential.Generative AI holds significant potential to advance the Internet of Things (IoT) by enabling new applications and improving existing ones. This article explores the vision, applications, challenges, and opportunities of integrating Generative AI into IoT systems. Generative AI can generate high-quality data, perform complex tasks with human-level performance, and generalize well to new tasks, which are essential for IoT applications such as mobile networks, autonomous vehicles, the Metaverse, robotics, healthcare, and cybersecurity. In mobile networks, Generative AI can simulate network data to predict bottlenecks and optimize resource allocation. In autonomous vehicles, it can enhance user interfaces and improve communication between vehicles and occupants. In the Metaverse, it can create immersive virtual environments. In robotics, it can improve interaction capabilities and adaptability. In healthcare, it can assist in processing medical data and generating reports. In cybersecurity, it can enhance security measures and predict threats. However, several challenges hinder the deployment of Generative AI in IoT. These include the lack of generative models for IoT data, the need to deploy large models on memory-constrained devices, the challenge of generating content with low latency, model adaptation and personalization, the need for IoT-based AI agents, privacy and security threats, and the lack of development tools and benchmarks. Opportunities to address these challenges include developing generative models for IoT data, model compression, runtime optimization, edge-cloud collaboration, efficient fine-tuning, designing IoT-based AI agents, federated learning, and creating IoT-oriented development tools and benchmarks. These opportunities aim to enable Generative AI for IoT, making it more efficient, secure, and adaptable to diverse IoT environments. The article concludes that Generative AI has the potential to revolutionize IoT applications, and further research is needed to fully realize this potential.