Suppose I wanted to conduct a 51% attack against Bitcoin. If I succeed, I stand to benefit monetarily, for example by shorting BTC on the Chicago Mercantile Exchange -- since a successful 51% would massively damage trust in bitcoin. Alternatively I could double-spend, though reaping the rewards without getting caught might be harder. Let's put the details aside and just agree that if you had the power to 51% bitcoin, you could make a lot of money off it.

What stops me from doing this is that the energy required for 51% of the mining power would be ludicrously expensive, even for just the half-hour[1] or so it would take to execute a damaging attack.

At least, it must be more expensive than the amount of money I could gain via my short position and/or double-spends. Otherwise it would be worth doing, right?

But for that to be the true, it must mean that the whole rest of the network is currently expending at least that much energy, every half hour or so, to constantly stop somebody from doing this.

So doesn't that put a dollar-amount floor on Bitcoin energy consumption, equal to [maximum amount one could stand to gain from crashing Bitcoin] repeated every [minimum duration of 51% attack to destroy confidence in bitcoin]?

Whether the mining is occurring with cheap energy in Iceland or expensive energy in California doesn't change that enough of that energy must be used to make it too expensive for an adversary (who could also be operating in Iceland or wherever) to attack.

[1]I say half-hour because 3 confirmations are considered sufficient for Coinbase to accept a BTC deposit, so I think it would make big waves if 3 blocks got reversed by a 51% attack.