It's true that PoS could lead to centralization and in a way encourages centralization - being a technical improvement over PoW in the context of energy use - PoS nodes run the same software with the same energy requirements regardless of the number of coins staked.
I would like to point out firstly that blockchain attempted to take a centralized world and decentralize it without much of a transition period, which has led to some pushback from traditional institutions. If PoS leads to a slightly less than centralized network, it is still fundentally better than a fully centralized network. This could be a necessary transition period in the long run.
There are ways to mitigate the problem of PoS centralization:
- Limit the amount of ETH per node
- Require smaller stakeholders to validate the chains of "super-nodes" before they are confirmed, such as rotating nodes in the bottom percentile of stakeholders to be validators with a bonus reward.
- Mining algorithms provide diminishing returns. Ex. Regardless of the amount of ETH, the stake reward maxes out at a set amount so that once a node reaches the Nth percentile in size, it's reward is equal to that of the largest node.
- Profitability is inversely proportional to the number of nodes. Ex. If you control 80% of the network, your max profit is capped at 20% of what it should be.
Thought Experiment: Nothing is done to prevent centralization, worst case scenario happens, and just a few nodes control the network. This increases the likelihood of the network losing value because of changes to the network and loss of faith. Network either a) Forks into a system with these previous checks or b) breaks the network up in order to restore faith and value in the system.
Regardless, we're in the early days. Trial and error is going to be helpful in figuring it out, and there's a good chance that the main network now will not be the main network 10, 20 years from now.