The Death of Denormalization?

Craig Mullins has a provocative article entitled 'The Death of Denormalization'. His argument lies in the fact that RDBMS technology now has better internal performance features, query optimizations, use of MQTs and ASTs, etc. These make the process of (de)normalization of databases almost redundant. While he may have some credence in his argument, I am skeptical - it pushes the design process too close to the system implementation. I have yet to utilize his new view for systems development but will certainly keep it in mind.