Bitcoin uses a proof of work system as a form of security in a decentralised system. At the end of the day, there's a certain cost per block, which approaches the reward for mining that block. Today you can fit hundreds of transactions into one such block, imagine a block like a mail package that holds all the world's transactions of the last 10 minutes or so, which is sent to everyone else who record those transactions, and imagine posting that package has a price. It's technically an incorrect analogy but it's just that, an analogy. But here's the thing, the block is digital and as bandwidth, storage and CPU inevitably scales up, you can eventually fit millions of transactions in that block, making each individual transaction much cheaper (despite the fact block rewards go up, too). And that's been the trend so far, every year transaction costs drop and electricity costs per value of money sent drops. It's quite likely that if bitcoin does become a popularly used worldwide digital ledger that the cost of electricity would be pretty trivial, to register any entry on the ledger (which could represent say 0.0001 bitcoin, but it could also represent a barrel of oil, or a million barrels of oil or a billion dollars, if you wanted to. Once you have a ledger you can write to you can make it represent anything you want really).

In short, the claim isn't incorrect today, it uses a lot of energy relative to its utility as a value network, but it's one of those things where the theory on paper, and the current trends in practice, show that the problem will solve itself over time. It reminds me of the endless stream of journalists decades ago who kept saying solar had no future because its embodied energy was more than it could generate, which was a partial truth in the 1980s and one that was rarely mentioned alongside the fact in theory it didn't have to be this way and in practice the trend showed it was getting solved over time. Today payback times of embodied energy went down from 50 years (for panels with a 20 year lifespan) to a fraction of a single year. Funny enough, you can still find people saying these things to this very day, ah well.

If you want to know the rough numbers... well money spent on getting x reward will approach x. Assuming 75% of the money miners spend on getting the block reward (the other 25% going to manufacturing or procuring the hardware, management etc etc), and knowing the price of a bitcoin is $240 and every day roughly 3600 coins are mined, then electricity expenditure is about $650k per day worldwide. Worldwide electricity consumption is about $3 billion (very much a back of the envelope calculation) a day at a hypothetical 10c a kwh, for the US alone total energy expenditures exceed a trillion a year. Bitcoin today is relatively wasteful for the utility it provides today that's for certain, but it's looking pretty good for the future, and in the grand scheme of things it's a really tiny operation in absolute terms.

What's interesting is that a lot of the bitcoin mining has sought out cheap electricity, e.g. Iceland where renewables are between 99 and 100% of electricity, there's a limited environmental impact to mining there and as prices are so cheap, multiple big miners have set up shop there in the past.

Finally it's important to realise that the cost of validating a transaction is pretty much near zero. The electricity waste is in the proof of work, which is hashing and it's just as hard whether you hash 1 or 1 trillion transactions. Further, the amount wasted is based on the block reward which is ridiculously high in the first years of bitcoin because of its distribution system. When the block reward drops, the amount of waste drops, and every day this inches closer to a more market-driven price of a transaction, thereby reward for mining and thereby the cost of waste invested to reap that reward. In short we're seeing, by design, a few years of ridiculously high waste-per-value transmitted to essentially kickstart the bitcoin ecosystem and distribute bitcoins, which drops every year. So again, the waste isn't necessarily tied to a transaction, whether a block has 0 transactions or a million, is no different for hashing and 99% of the reward today is still the block reward (i.e., irrespective of the amount of transactions). So anyone citing a cost per transaction figure is very misleading even though it's not incorrect in some way, because that cost is not marginal but fixed and exists even if there are 0 transactions and is therefore not a 'per transaction' cost. A bit like if you maintain a bridge for $100 a day and 1 person walks on it, doesn't mean the cost of walking on the bridge is $100 per person, it's actually 0 because if you walk 0 or 100 people over the bridge, the fixed cost is always $100 and the marginal cost is always $0. Obviously a bridge has capacity limitations, but bitcoin as a digital ledger has much fewer. The total bandwidth of bitcoin today, no joke, is 1 mb per 10 minutes, or about 1.7 kb per second. We could increase that by orders of magnitude in the coming years and be okay, meaning whatever 'cost per transaction' anyone cites would also drop by orders of magnitude. Anyway I'm rambling but the cost/waste argument always feels a bit disingenuous given a medium to long-term context in which it gets less wasteful every year, and given in the short-term it's pretty minor in absolute terms anyway.