I like new cars, but hate buying new cars because they decrease in value so quickly - hence my trusty Corolla. Similarly - I like new PCs. but hate buying new PCs, as they get slow and clunky very quickly. As data sizes increase, we are soon left staring at progress bars for hours. Ok - you can move everything onto a server like we have with some old hardware in our computer room for the slick mapping/Sentinel 1 analysis and leave things processing in the background - but these are still computers in effect, sitting in our lab, ageing. So one thing I am keeping an eye on is cloud processing - which although costly at present has great potential to cut down on hardware costs and always gives you the processing power you need. Google to the rescue once again - with Google Earth Engine (python and java friendly) can be used directly with QGIS for processing and also hosts most if not all of the Landsat and Sentinel archives already.
Dáire recommended it to me a few months back but I was too distracted to sign up - however i have remedied that now and I've signed up to be a 'trusted tester'. I will start bringing this into my research and lecturing and see how I get on if I get approved.