It just short circuits the ewaste step really. Those with the most money could buy the most mining hardware so the result was the same

Not exactly though.

With POW you need to make a sizeable investment, and hope you'll get a good return eventually.

With POS you can keep your money and do it almost for free (as long as you run a node, not that big of a deal).

So exchanges will completely dominate staking, but they likely not even touch POW mining.

“Fungible” capex? What does that mean?

Capex in PoW is non-fungible by virtue of being spent in hardware, physical equipment, whereas staking means 32 eth is 32 eth regardless of hardware

The opex is the part that tethers on-chain security to reality.

Can you elaborate on why you think this is the case?

Stealing private keys (through whatever means) is considerably easier than stealing a comparable amount of mining hardware then running it. It's not that big of a risk in reality of course, but there is still a considerable amount of security provided by the fact that obtaining enough compute for a 50% attack is harder than obtaining enough eth to do the same.

Key phrase to search - "nothing at stake problem"

For the most part ethereum doesn't suffer from this issue.

Eth's work-arounds all involve centralization. If you want decentralized security, you need real physical opex.

What centralization? Adherence to protocol rules? How requirement of locking down 32ETH is different from requirement to run ASIC that does two rounds of SHA256?

Please don't throw around buzzwords.

The slashing enforcement mechanism introduces centralization; it has to, to work around nothing-at-stake.

I'm not throwing around buzzwords. Please don't accuse people of that without sufficient understanding to be reasonably confident you're correct.

> locking down 32ETH is different from requirement to run ASIC

This is precisely the opex-vs-capex distinction we are discussing in this thread.

32eth is strict capex. ASIC capex <<< power opex.

> The slashing enforcement mechanism introduces centralization; it has to, to work around nothing-at-stake.

It is other way around. Slashing is the solution for nothing-at-stake. By requiring multiple validators to sign each block height you prime them for being slashed if they ever double sign those heights (or sell key to someone who does). Where is centralization here? Are you speaking slashing condition detection? Or maybe about slashing tx inclusion into the chain?

> Are you speaking slashing condition detection

Correct - the decision about when to slash

It is as easy as finding a duplicate signature. And while it takes some computational power (pennies comparing to PoW), you need just one watcher to guarantee that chain will remain honest. So, 1 honest person out of n. Almost as good as it gets.

Whole spiel about "only EF will run those, whole thing is centralized" is just noise. You as an attacker can't distinguish between situation when I'm running a slasher from one when I'm not. Now go and risk the attack.

Yes, I'm not denying that.

The question was why "opex is the part that tethers on-chain security to reality", and opex specifically.

Opex specifically because we don't have a predicative algorithm for proving capex. The only predicative value-proof algorithm we have is PoW, which works on opex. PoS is completely impredicative.