We propose a framework for constructing factor models for alpha streams. Our motivation is threefold. Firstly, when the number of alphas is large, the sample covariance matrix is singular. Secondly, out-of-sample stability of the sample covariance matrix is challenging. We discuss various risk factors for alphas, such as style risk factors, cluster risk factors based on alpha taxonomy, principal components and using the underlying tradables (stocks) as alpha risk factors, for which computing the factor loadings and factor covariance matrixes does not involve any correlations with alphas, and their number is much larger than that of the relevant principal components. We draw insight from stock factor models, but also point out substantial differences.