The author introduces a block variant of the well known Householder factorization procedure. Numerical examples illustrate the methodology, and it is shown that the new method is faster than the corresponding LAPACK algorithm.