While half the German cabinet is travelling to Jülich, Germany, to inaugurate a new supercomputer Jupiter, which ranks fourth on the Exascale Top 500 list with a power consumption of 17MW, the hyperscalers are getting down to business when it comes to monopoly policy. In Jülich, 24,000 GH200 Grace Hopper superchips from NVIDIA are being installed, naturally also for AI and LLMs. All these ideas make sense, but one of the most nonsensical applications of graphics cards and their derivatives is AI. Weather forecasts, CO2-free power supply and energy savings in air traffic are already areas of research that sound applicable and whose significance is obvious. Astrophysics and quark-gluon plasmas are interesting, but I’m more interested in whether I can get through the winter with solar energy, battery storage and a heat pump.