@franky1, you make some very good points. In particular, it is true that the Ethereum stakeholders wouldn't just be able to buy the ASICs from one day to the other. Nor will they likely be able to bribe the existing miners unless they also pay for their ASICs as well, since these will most likely be very hard to sell after a fall of Bitcoin.
This cost is considered, however, in the referenced $6B–$20B estimate by Nuzzi et al.
I also want to contest you on the statement that Ethereum would collapse as a result of buying enough ASICs or bribing enough miners. $6B–$20B is only 1.3%–4.4% of Ethereum's market cap, so it will only require 1.3%–4.4% of all Ether in order to pay for a 51% attack. Not something that is at all likely to cause a crash of Ethereum, will you agree?
you think 1-4% is not enough to crash a market.. but here is the thing.. not all coins circulating are on the market.. only a small proportion of coins crculating are on market orders so a sudden increase of coins hitting the market supply would cause change
take for instance bitcoin, there are over 19m coins in circulation.. but there are not 19 m coins on the market. most market orders are 0.0X coins per order.. so if there was a sudden period where people were colluding to sell for instance 190,000(just 1%) coin the market orders would flood with sell orders compared to the norm.. now translate that to the numbers of ethereum circulation amount and market supply and see some scenarios of then dumping 1-4% of that circulation on the market supply.. you will soon see how it affects the markets
we seen it many times in the bitcoin market when the market orders were <1btc each. but then whales created walls and orders of just 1000 coin order lumps.. it had enough impact on the markets
To your point about the attackers having to 'wait around to collude' (assuming that they choose to buy ASICs rather than bribe miners and buy into existing infrastructure), it is worth pointing out that they don't have to be idle as miners in that time. They can behave as honest miners right up until the attack. Also, since this will mean that rewards will become more thinly spread out, some of the actual honest miners, who aren't backed by Ethereum stakeholders, will likely fall off in that time, making it easier for the attackers to gain a majority.
so we agree that they would mostly want to act as honest miners until some colluded attack date when they had enough power to attack.. however they would then while running honest, realise that now they have become bitcoin investors. their own economic position is now in ROI of bitcoin. and wanting to shoot self in the foot to attack bitcoin.. would lose them income
And to your point about the consequences of a 51% attack, it is worth underlining that the majority of the costs of a 51% attack are from buying the equipment. The operational costs are small in comparison. Once the attackers have gained a majority (and with the backing of Ethereum's stakeholders), they will thus be able to rewrite many more than just a few blocks of the history. They could potentially steal a lot of bitcoin in doing so. And it is therefore unlikely, as far as I can see, that Bitcoin's value wouldn't be severely damaged by this. Would you agree?
if you play out hashrate speeds.. there is only so far they can go back and then be able to then re-hash old block heights to catch up with the honest mining pools.. to achieve an ability to go back and then catch up and overtake requires more then just 51%
imagine it like olympic runners (400m relay)
2 runners, one can do 100m in 10seconds. the other can do it in 9.8seconds
now if the fastest runner can just about beat the other runner each race.. how far can he run backwards before then running forwards to overtake the other runner
can the fastest runner honestly run the first 100m side by side other. then run backwards then start from the start and run the entire 400m and still beat the other, where the other only had to run 300m in the time the dishonest racer had to redo the entire 400m
think about that
the dishonest runner cant just go back as many xxxm portions and redo the race like 2km back and then overtake the other honest runner
run the math