I'm curious about this point. Is this true? Does the energy cost increase in proportion to the hardware cost? It seems like bitcoin mining hardware becomes much more energy efficient with each new generation.
Could you link to a source?
There's some background data in here: https://medium.com/@interdome/how-much-electricity-does-bitc...
See my response to your first comment - you can not simply prevent any 51 % attack by changing the protocol. By the way, I am not downvoting your comments, I am (usually) not voting on any responses to me but let the other users judge.
It's actually a lot cheaper to attack the network than you are stating, almost all mining uses Stratum which has been successfully redirected using BGP route hijacking in the past. It has no methods of authentication and does no way of integrity checking what it is mining against. A simple "mistake" in routing could have all of a countries hash power redirected to a malicious actor. You don't need power, datacenter or miners, a $5 wrench applied to the right head would do the trick just fine.
They don't have to be proportional but you want them to be. You neither want to transfer billions of Dollars only protected by mining hardware and energy worth a couple thousand Dollars, nor do you want to protect transfers worth a million Dollars with mining hardware and energy worth billions of Dollars.
The lack of possibility for a pricing feedback loop makes this all a little tricky too, as the system can't make any changes based on external influences.
Someone with 51% of the network (as you say, costing $100M+ of mining equipment currently) could execute an arbitrary double spend attack. If they are financially motivated, they might try an attack like this:
1) Buy a large asset (say a publicly traded company, a huge tract of real estate, a closet full of gold) with bitcoin. 2) Wait for the transaction to clear. For a large asset purchase such as the above, this could take weeks. 3) Once the asset is in hand, use your mining power to reverse the transaction.
There is a critical flaw in this plan: you are obviously stealing, and the justice system will treat your actions as theft. If you try this for a reversible asset transfer, like real estate, they will reverse the transfer in addition to punitive measures. For gold, you might be able to hide the gold somewhere, but again at that point it is just common theft.
Attacks that are not financially motivated are a different story; a nation state or other party could try to 51% the network simply to destroy bitcoin. Of note here is the fact that nation states have more straightforward ways of hurting bitcoin, starting with outright bans.
Why cant bitcoin exist alongside them?
The article lines out how to neutralize an attacker mindlessly building a chain of useless transaction but this obviously does not solve the problem. What if the attacker forks and rebuilds the chain of the last week but randomly leaves out every tenth transaction while slowly poisoning the genuine chain with useless transactions until the algorithm decides to switch? You just undid thousands of transaction originally confirmed a week ago. If you could just make up a couple of rules to prevent all 51 % attacks they would not be an issue in the first place. But this again would imply that it does not matter how the computing power is distribute between the good guys and the bad guys which obviously completely contradicts the fundamental design of Bitcoin.
[1] http://gavintech.blogspot.de/2012/05/neutralizing-51-attack....
You are quoting the wrong part. This only holds if the attacker does not try to build a side-chain.
It's much more difficult to change historical blocks, and it becomes exponentially more difficult the further back you go. As above, changing historical blocks only allows you to exclude and change the ordering of transactions. It's impossible to change blocks created before the last checkpoint.
[1] https://en.bitcoin.it/wiki/Attacks#Attacker_has_a_lot_of_com...
If Bitcoin would gain widespread adoption and see transaction worth billions of Dollars every day, it could easily become a worthwhile attack target even at the mentioned attack costs. The only protection would be to use a lot more mining hardware worth billions of Dollars to make an attack to expensive but this in turn would also raise the energy consumption proportionally and in the end the costs per transaction. I think that in this regard Bitcoin is fundamentally flawed because - up to some limit - you want the transaction volume protected by a proportional amount of mining hardware causing proportional energy and transaction costs which then imposes a lower bound on the costs per transaction.
[1] https://news.ycombinator.com/item?id=8782877