This keynote will provide a reflection on the evolution of super computing technology with respects to the needs. Looking at the PRACE “Scientific Case for Computing in Europe 2018–2026”, ETP4HPC “Strategic Research Agenda”(1), and Europe EURO-HPC(4) plans we will try to establish a matching between emerging computer architectures and the software stacks. Major actors such as US, China, Japan, Europe, ASEAN compete in the Exaflop race driven by the top 500, the famous ranking of supercomputers. Everybody now a days admits that the under lying benchmark (Linpack) does not reflect the diversity of the needs, nor the different dimensions of the problematic that include number crunching, but also memory size and communication speed. These two last dimensions become crucial when dealing with big data analytics fed by ubiquitous sensor networks. Lately Artificial Intelligence (AI) techniques such as machine learning and in particular artificial neural networks (e.g. deep learning) were largely advertised in this context. How will next generation of supercomputers respond to the needs of computational science and AI? What are the consequences for research and development in computer science?