100 tiny coins that nobody seriously trades moving to PoS doesn’t mean anything, ETH has been moving to PoS “real soon now” for literal years. Bitcoin and it’s variants can hardly decide on the max block size, so a fundamental shift like PoS doesn’t seem likely.
Also, PoS doesn’t solve resource shortage pressure - it reduces some power consumption, but it still puts artificial pressure on chips and drives.
Today they set into motion the first public testnet PoS migration, Ropsten, which will occur at a specific TTD within the next three weeks or so. This is the first time a long-lived Ethereum network will be migrated to PoS. There are only a handful of public testnets, after that the only thing left to do will be to merge mainnet using the exact same process.
If you don't follow the development process, it's easy to say "any day now I'm sure." But for people who choose to run the software and personally participate in the merge rehearsals, the proof is very much in the pudding.
> Also, PoS doesn’t solve resource shortage pressure - it reduces some power consumption, but it still puts artificial pressure on chips and drives.
PoS doesn't really use more resources than any other typical data center application. Without resource-intensive PoW miners, a blockchain is just a piece of software like any other, really. A copy of the entirety of Ethereum fits on a single consumer drive.
Also this is an active area of development!
The most widely used Ethereum client Geth (Go-Ethereum, because its written in Go) has 3 or 4 modes. a pruned+fast synced node absolutely fitting on any consumer drive. while its archival node with all the prior states of the blockchain (not necessary for consensus) takes 9 terabytes, and is nearly impossible to sync up with without specialized drives and computational resources.
Other clients improve on that. Erigon is also go-based but with a different data structure, it only does archival nodes and takes 2 tb. Rust based ones like Akula and Silkworm also only do archival nodes, and they take 2 tb as well. Roughly 400% improvements in performance, and the same gains are seen in how much space they take.
There is probably more room for improvement as well, such that an archival node is a viable thing to do.
That 9TB (or 2TB) is only 400GB when it's stored without all of the states unpacked to disk (i.e. non-archive node).