I've had experience writing distributed, decentralized software (mobile agents, distributed wireless sensor networks, animal wireless sensor networks) and the biggest impediments to building decentralized systems is that there is very little prior work to rely on. There are few libraries, examples, or clear explanations of the different approach that is needed. There are also some difficult trust problems with decentralized systems (are you OK with trusting your data is secure stored on several of your neighbors devices? Are you sure any random computer is secure to run your app on or use your data? Are you OK with letting some unknown vendor run tasks on your device and your data? There are solutions to these but they are not widely known or used.) In particular, decentralized systems can be very difficult to troubleshoot, especially if you are moving code from machine to machine and executing it wherever it lands. These problems can be solved but right now everyone has to solve them themselves from the ground up instead of relying on known common solutions. Some types of decentralized technologies are essentially the same as computer viruses and figuring out how to keep them under control is essential. Centralization became widespread because it was cheaper and offered centralized control (which is attractive to primates with hierarchical societal tendencies). Decentralization is possible, it's just going to require a similar amount of work to make it competitive and available.