They have, it’s just difficult to have it catch on. Bitcoin being the prime example… but the compute is basically burned for no useful purpose beyond security of the network itself. Because that is the main concern - security of the blockchain. Many have tried to use that same shape and instead have the compute be directed at something useful, with varying levels of success but nothing at that scale. There are however very successful long running projects like folding@home or the Mersenne prime search which have lots of people running a program locally that directs idle compute power on volunteers PCs to work on scientific pursuits.

There’s lots of cool examples of interesting things people have tried, (training models, etc) but it’s not easy to get traction.

Something like vast.ai is much easier to make fly because people who offer their resources for rent get directly paid by time used.

In short, it’s a cool idea that lots of people want but it’s hard to make fly.

Edit: also I guess we could consider malware / botnets successful implementations as well - for nefarious purposes.