If you’re interested, I found https://arxiv.org/abs/2505.09814v1 to beat Strassen for medium-sized and larger covariance matrices. YMMV of course. Takes a little adjustment for XX^H but it’s not so bad.