The Industrial Internet of Things' aims to connect industrial assets with one another and subsequently bene t from the data that is generated, and shared, among these assets. In recent years, the extensive instrumentation of machines and the advancements in Information Communication Technologies are re-shaping the role of assets in our industrial systems. An emerging paradigm here is the concept of social assets': assets that collaborate with each other in order to improve system performance. Cyber-Physical Systems (CPS) are formed by embedding the assets with computing capabilities and linking them with their cyber models. These are known as the Digital Twins' of the assets, and form the backbone of social assets. Collaboration among assets, by allowing them to share and analyse data from other assets can make embedded computing algorithms more accurate, robust and reliable. This paper proposes a Multi Agent System (MAS) architecture for collaborative learning, and presents the fi ndings of an implementation of this architecture for a prognostics problem. Collaboration among assets is performed by calculating inter-asset similarity during operating condition to identify friends' and sharing operational data within these clusters of friends. The architecture described in this paper also presents a generic model for the Digital Twins of assets. Prognostics is demonstrated for the C-MAPSS turbofan engine degradation simulated data-set (Saxena and Goebel (2008)).