In the current paper, we review the potential to predict the creation of unicorn start-ups, which are businesses with a valuation exceeding one billion dollars, using machine-learning classification algorithms and training them on all the global data available. In spite of the fact that such entities represent an uncommon, yet powerful part of the entrepreneurial ecosystem and trigger much interest of investors, policymakers, and scholars, there is little systematic identification of such firms beyond highly speculative and intuitive findings. Using structured information provided by various global websites and applying Logistic Regression, Random Forest, XG Boost, and Neural Network models, the present research unveils that a number of startup variables are strongly interrelated with the achievement of unicorn status. Such are paths of funding, founder experience, the size of the team, the geographic origin, and the sector of interest. The given evidence shows that the combination of machine-learning and interpretability tools, in particular, SHAP values, would create a resilient and clear structure, which could be used to respond to the strategic investment and policy choices.