Electric power systems are prone to various kinds of transient disturbances which exist only for a fraction of second and often trigger cascading failures. Hence it is important to detect and prevent them from spreading in time. Conventionally these events are prevented by deploying costly special protection systems (SPS). Unfortunately, in many cases SPSs mis-operate as they could not predict the stability well ahead and are designed to operate based on past experiences and extensive off-line simulations. This paper proposes an online transient stability prediction scheme based on live synchrophasor data. The novelty of the proposed method is that it accurately predicts the transient stability based on only few (10 to 12) sample fault data without solving computationally extensive electromechanical dynamics. Synchrophasor data from geographically distributed Phasor Measurement Units (PMUs) are collected, synchronized, aggregated (if required) and analyzed on a stream computing platform to predict the trajectories of the generators which are then used to predict the transient stability of the grid. Performance of the proposed scheme is evaluated on the benchmark systems and evaluation results are presented in this paper. © 2012 IEEE.