While all donations to fundraising institutions help contribute to their causes, it is major gifts that have the largest impact. For many of these organizations, major gifts make up 80% of their donation dollars, making major gifts vital for their very survival. Fundraising institutions have thus always been building lists of prospective major donors, until recently doing so by hand while combing through spreadsheets of donor data, typically. We show that machine learning can be used to accurately predict which of a fundraising institution’s constituents is most likely to make a major gift, making use of 11 shallow and deep learning algorithms. We seek ‘useful’ models - models that help fundraising institutions discover major giving prospects, which involves the tolerance of some false positives while minimizing false negatives. Our research found models able to achieve 92% accuracy with about a 2:1 ratio of false positives to false negatives. LSTM-TRU and Extra Trees classifiers tended to produce the best results in terms of accuracy and false positives. In addition to predicting who will give a major gift, we generate models that predict how much major donors will donate, with an RMSE in the $1,000 range for educational foundations.