I think one reason this might be the case is because modern cloud-computing makes scaling a single centralized service super easy and straightforward.

In the early 2000s if you wanted to build something "at scale", the only real way forward (besides raising tons of money upfront) was to build a decentralized protocol. Offload most of the compute, bandwidth and storage to the end nodes.

Nowadays, you can just an AWS auto-scaling group behind a load-balancer and a CDN, and you can pretty much handle infinite traffic.