Why do you think the huge energy waste wont be a deal breaker?
The energy used needn't increase as the network is used more, so the current situation needn't get worse. As the block reward halves, it will become easier to manage the total cost of mining by controlling transaction fees.
(I think the opposite will happen. The energy cost of Bitcoin will likely reduce over time, for the reason given below.)
What will happen when all BTC are mined?
The hope is that the total transaction fees per block will rise to compensate. Partly by the number of transactions increasing, but if necessary the fee per transaction can be increased. In the short term, the real-world value of Bitcoin may increase, too. If Bitcoin doubles in value every 4 years, that will nicely counter-act the block reward halving every 4 years, even if transaction fees and number of transactions stays the same. (I say this is a short-term effect because I don't believe any economy can sustain that growth indefinitely.)
Alternatively, the number of miners will reduce until it balances.
Another way of putting it is that no-one really knows. It may end up in a downward spiral of higher fees, hence fewer transactions, hence lower mining income, hence fewer miners, hence network less secure, hence less trust in the currency, hence less demand for Bitcoins, hence lower real-world value. One reason I think Nxt is important is that I think Bitcoin may (slowly) die in this way. Not "will", but it's a high enough risk to be worth hedging against.
The question is: How high will the transaction rate have to be to incentivise enough miners to keep the network secure?
Or: How low can the difficulty fall for the network to be considered secure?
It won't be a problem because the number of minors won't change suddenly. While a drop of 50% in the block reward seems like a lot, it really isn't, because it's the some order of magnitude. It would be a big drop if it happened without warning, but minors plan their purchases will in advance.
It won't be a problem for the miners because they can drop out in good time. However, that doesn't answer the question. As the miners stop investing, the network becomes less secure.
The security of the network against an outside attack is roughly what it would cost to mount such an attack, which is roughly what it would cost to acquire 51% of the hashing power, which for reasons of economics is directly proportional to the real-world value that miners get paid for mining one block. So if the block reward halves, and Bitcoin value, transaction volume and fee-per-transaction stay the same, then the security of the network halves. So it's really a question of how low can it get before it matters.
That's a matter of opinion, and your threat model. Some say the network is already vulnerable. Some say Bitcoin it is actually over-secured now.
Personally I am in the former group. An attack might cost someone like a bank $100m. HSBC got fined $1.9bn for money laundering. UBS got fined $1.5bn for rigging Libor. If those guys felt seriously threatened by Bitcoin, $100m is easily affordable to them. Since a lot of ASICs are made in China, the Chinese government could probably do it with a phone call. A terrorist might manage it by kidnapping the ASIC manufacturer's daughter. Not to mention we've already seen hash pools hovering around 51%, that we're basically just trusting not to mount an attack. Many people in the Bitcoin community are complacent, in my view.
