[SebRollen]

You're right that the energy cost is incurred at the block-level, but since each block can hold a certain number of transactions, you can then infer what the effective cost is per transaction. The source for kWh per transaction comes from Digiconomist's bitcoin energy consumption index, and you can find their methodology here: https://digiconomist.net/bitcoin-energy-consumption#assumpti...

[thebean11]

It's just an incorrect attribution. If there are no transactions, blocks will still be mined. If there are more transactions than can fit in a block, no extra blocks will be mined. If the transaction format changes so that more transactions can be included in a block, the energy usage doesn't change.

It's just a very obfuscating way to describe what's happening. Not only is it not a useful abstraction, but it actively leads you to wrong conclusions to talk about it in terms of energy use per transaction.

[SebRollen]

I think this commenter put it pretty well, so I'll just link to it: https://news.ycombinator.com/item?id=31596781

[thebean11]

All of those things are directly attributable to miles on the car. More miles means more repairs, more frequent oil changes, faster depreciation of your purchase price etc. On the other hand, number of transactions is purely a function of block size and transaction size (which again has no influence on energy usage).