I suppose that vulnerability ratio computation really should reflect the true cost. But for Bitcoin, I don’t know the cost to design your own chips and build 5,000,000 tH/s of capacity (before difficulty rises even further).
Let’s say it adds $50M to the $500M I had estimated previously. That’s still leaves bitcoin with a ratio showing it to be more vulnerable relative to Ethereum, etc.
Incidentally, since first posting this article the total Bitcoin hashrate rose, but not as fast as the exchange rate — causing the vulnerability ratio to rise (i.e., weaken) even further.