We derive novel matrix factorizations, leading to an efficient streaming algorithm for privately estimating partial sums even when the size of the input is a priori unknown.