Computing power is just a proxy for capital/resources. Why not be more efficient and use the capital directly and save power in the meanwhile.
Current market cap of ETH is ~$324B, thus getting 50.1% of ETH would require $162.3B in capital. However, as soon as you start acquiring ETH the price will increase, especially at those large volumes.
It would be insanely hard to come up with enough resources to buy enough ETH in a POS world to take over the network. Never mind the fact that as soon as it's become evident you've taken over the network the value of the network is essentially worthless and you've just destroyed billions of dollars worth of capital in the process.
I wonder if a state actor could pull it off more cheaply. Start buying large amounts while letting it leak that you're going to take over the network. See if enough people will panic-sell on the leak to drop the price of your takeover.
It's kinda like taking over a condo building on a much larger scale: the people you buy out first can charge a premium; by the end, you set the terms.
Why would people sell rather than fork to a version of the network where those ETH did not exist? One of the benefits of POS is that when you fork away from a malicious actor, they have to start over from the beginning while in POW, they can just point their hardware to your new chain unless you change the mining algorithm and screw over all the other miners.
This is exactly what happened with Ethereum in the early days when a bad actor was able to exploit a third-party contract to the tune of 5% of all ETH.
The Ethereum everyone talks about today is the fork, due to the Ethereum Foundation which owns the trademark leading the fork.
The Ethereum blockchain with the unaltered history is called Ethereum Classic
Note that the way this fork was pulled off was very ad-hoc.
Ethereum devs were unable to create a legitimate transaction reverting the DAO funds because they do not have access to the hackers' private key. The reversion was done with a "surgical state change" hardcoded into the client itself.
Think of it this way, if another company announces they're planning to buy a publicly traded company, what happens to the value of the shares?
The price goes up, you've just made it more expensive for yourself to take over the network.
If you were to attempt to take over you'd be better to do so in silence. However, it would be hard to hide that kind of control and wealth when every information on the network is publicly available.
In a public company, though, having been taken over doesn't defeat the purpose of the company.
If you're intentionally trying to take things down, sellers have a huge incentive to not be left in the 49% who hold something that's now lost its value - as you say, "as it's become evident you've taken over the network the value of the network is essentially worthless." I think you could get the value to go to worthless well before actually hitting 51% on intent alone, if you're a big enough power.
Why assume a state actor? Given the sorry state of DeFi contract security, it's far more likely that an enterprising hacker can gain a dominant staking position by pillaging and then staking ETH from buggy contracts.
Current market cap of ETH is ~$324B, thus getting 50.1% of ETH would require $162.3B in capital. However, as soon as you start acquiring ETH the price will increase, especially at those large volumes.
It would be insanely hard to come up with enough resources to buy enough ETH in a POS world to take over the network. Never mind the fact that as soon as it's become evident you've taken over the network the value of the network is essentially worthless and you've just destroyed billions of dollars worth of capital in the process.