Decentralized Machine Learning (DML) is a complete infrastructure facilitated for on-device machine learning. By blockchain and federated learning technologies, it decentralized the three integral elements in machine learning, which are data, processing power and algorithms. Without extraction of private data and thus protecting data owners’ privacy inherently, DML unleashes the untapped data usage in individual smart devices (smartphones, tablets, PC and IoT sensors) for various business intelligence, market insights and sentiment predictions consumed by different commercial and organizational clients. DML will use the idle processing power of individual devices for machine learning through a mobile application, and the machine learning algorithms will be crowdsourced in a digital marketplace.