If you think it needs be be more than $100 M then post here with a number. But to assume the sky is the limit is to avoid all rational economic calculation.
With a $40 trillion a year global economy, 5% in PoW debasement (the USA wages and GDP both grew by 5% nominally per annum from 1790 to current, so inflation goes into wages) is $2000 billion a year. That is how much I think we need to prevent the scenario I am worried about below.
Also this $2000 billion ($2 trillion) will go to the owners of PCs, if we perfect the PoW algorithm, so they get both the rise in wages and the debasement back as mining profits. Win-win.
What exactly the attack that one could profitably make against $10 million dollars worth of electricity invested in hash power?
This (world war 3 to bring in the SDR system)! And it is underway now. Most people are naive to
the reality ongoing now. Go watch the video of General Wesley Clark explaining that the
USA had a plan for over a decade to overthrow 7 countries in the Middle East.
Obama is going to war with or without the vote of Congress and the public, just as he did in Libya.
Conclusion: don't rely solely on proof of work to define the best chain. If someone pops up with a new chain that is more than 6 to 10 blocks forked they are ignored until manual intervention is taken after community consensus can be reached about whether or not that 6-block fork was mined in private or the result of a legitimate chain split caused by a network outage.
Forget the profit motive. What we have to protect against is the international bankers are moving us to a NWO with SDRs as explained in my first link above.
Network outage and packet filtering is something they can do.
So we have to think about what happens to the blockchain in that case.
This is why I am not enamored with BitAssets right now. I want to focus on designing a currency that can survive what we will be going through in the next few years. I'd rather give up exact timing on the arrival of the blocks of the blockchain under adverse conditions.