This chapter investigates the application of Federated Learning (FL) in decentralized video streaming systems, focusing on how FL enhances scalability while preserving user privacy. Traditional centralized video streaming architectures face challenges such as bandwidth constraints, privacy risks, and server-side processing limitations as demand continues to grow. FL addresses these issues by enabling local devices to collaboratively train machine learning models without exchanging raw data. Key topics include data partitioning strategies, model aggregation techniques, and the trade-offs between scalability, efficiency, and privacy in FL-based streaming systems. Additionally, we present real-world case studies that demonstrate how FL can be effectively deployed to optimize video streaming performance, particularly in areas like personalized content recommendations and adaptive bitrate streaming. These examples highlight the benefits of decentralized learning in achieving high-quality, scalable, and privacy-respecting video streaming services.