Online hate speech, particularly over microblogging platforms like Twitter,\nhas emerged as arguably the most severe issue of the past decade. Several\ncountries have reported a steep rise in hate crimes infuriated by malicious\nhate campaigns. While the detection of hate speech is one of the emerging\nresearch areas, the generation and spread of topic-dependent hate in the\ninformation network remain under-explored. In this work, we focus on exploring\nuser behaviour, which triggers the genesis of hate speech on Twitter and how it\ndiffuses via retweets. We crawl a large-scale dataset of tweets, retweets, user\nactivity history, and follower networks, comprising over 161 million tweets\nfrom more than $41$ million unique users. We also collect over 600k\ncontemporary news articles published online. We characterize different signals\nof information that govern these dynamics. Our analyses differentiate the\ndiffusion dynamics in the presence of hate from usual information diffusion.\nThis motivates us to formulate the modelling problem in a topic-aware setting\nwith real-world knowledge. For predicting the initiation of hate speech for any\ngiven hashtag, we propose multiple feature-rich models, with the best\nperforming one achieving a macro F1 score of 0.65. Meanwhile, to predict the\nretweet dynamics on Twitter, we propose RETINA, a novel neural architecture\nthat incorporates exogenous influence using scaled dot-product attention.\nRETINA achieves a macro F1-score of 0.85, outperforming multiple\nstate-of-the-art models. Our analysis reveals the superlative power of RETINA\nto predict the retweet dynamics of hateful content compared to the existing\ndiffusion models.\n