Now there are some important differences between this mining profit and credit card transaction fees. One is that this isn't a transaction fee, in that it isn't paid by the buyers or sellers. Which is true; it's paid by the users of Bitcoin generally, I guess. It's like if you could pay with your Visa card, and it didn't cost you anything, and it didn't cost the merchant anything either, but at the end of the year the Federal Reserve printed $500 billion and handed it to Visa. That would compensate Visa for running a payments system, and the incidence of that cost would be on, like, "users of the dollar system," though it wouldn't be tied to transactions exactly. You'd bear that cost whether you used or hoarded your dollars. That's certainly true of Bitcoin, too. Presumably you use Bitcoins because you want to transact in them but we all know that's not true.4
The second is that this isn't a transaction fee, in that it's not paid to miners to process transactions. It's paid to miners for mining. Put another way, the miners are mainly paid for their cryptographic makework, not really for processing transactions, which is computationally a trivial side effect of what they do with their fancy computers in Iceland. Put another 'nother way, this cost is "seignorage," not a transaction fee. I do not find these responses very interesting but you might. The "mining" is not in itself a valuable activity. The reason miners are part of the Bitcoin ecosystem is to reward them for confirming that transactions are valid. That is the service they provide, so it stands to reason that that is what is being rewarded with their 3.5% profit.
The third is that this fee is high only because of the relatively low volume of Bitcoin transactions, and will eventually get much lower. (See the prior point, it doesn't actually cost the miners much to do the transaction-processing work.) And, sure, as a percentage it's come down; in 2009, daily miners' revenues were frequently several thousand percent of daily transaction volumes. On the other hand, that trend is not obviously continuing; see that graph above. And here is the dollar cost of a Bitcoin transaction over time:
Which does not quite tell the story of "mining transaction price will come down over time."5
But, I don't know, like I said I'm not much of a Bitcoin connoisseur. I'm mostly interested in the way that Bitcoin has found a way to compensate transaction processors without charging a fee directly to the buyers or sellers. That's neat, but I'm not sure it means that there's no cost to the buyers or sellers. The claim that Bitcoin could one day provide a low- transaction-cost alternative (or currently provides an anonymous or accessible alternative) to the U.S. dollar payments system is separate from the claim that Bitcoin is currently a cheaper, or free, payments alternative. Bitcoin, as a transaction mechanism, is actually pretty expensive! It's just that it's clever at hiding its costs.
1 This generally describes a lot of structured transactions where the dealer has a better model than the client and says, "nonono you pay us nothing up front and we give you this valuable transaction for free!" "Valuable" there means "to us." I enjoyed this story about how Bank of New York Mellon provided "free" currency conversion at the cost of ... just converting clients at the worst price of the day. More broadly there is this Thomas Philippon paper finding that "the finance industry of 1900 was just as able as the finance industry of 2010 to produce loans, bonds and stocks, and it was certainly doing it more cheaply." Whatever is being competed on, it isn't efficiency.
2 There's some provision for transaction fees for, essentially, odd-lot transactions. As mining rewards decline these are projected to become more common, I guess.
3 All the data is from blockchain.info, which is I guess a quasi-official source of Bitcoin information.
4 This is sort of a rabbit hole. The mining rewards look like inflation, since they're money-printing. So in some sense they're a cost imposed on Bitcoin hoarders and not on Bitcoin transacters. But of course Bitcoin itself is broadly deflationary. And I guess you can view that -- or its price volatility for that matter -- as a transaction cost of its own. Anyway, I think "Bitcoins actually paid to the network to facilitate transactions" is a reasonable proxy for "transaction fees," but you can disagree!
5 Which is not to say that it won't. The interesting thing is though that the point of the miners is to securely and correctly confirm transactions that actually happened. The mining reward incentivizes lots of people to mine, which makes the network secure and robust. If the total transaction value -- mining rewards plus transaction fees, which are ultimately "envisioned" to be worth more than the mining rewards in the future as the mining rewards come down -- were zero, or some trivial amount, then no one would work for them, and one person with a lot of computer hardware could hack Bitcoin to basically transfer all of the Bitcoins in the world to herself. (Again, see this explainer, which says that this is impossible "unless the hacker has more computing power at her disposal than all other Bitcoin miners combined.")
So there needs to be some reward to the miners, and that reward has to in some loose sense be related to the value that you'd get from hacking Bitcoin: If the reward is too low, there won't be enough people checking transactions to ensure that the transactions are valid. Here is a skeptical view. It's not obvious that the reward needs to be 2% to 4% of transaction value, instead of say 20-40 basis points, or 2-4 basis points, or whatever, but I'm not sure that there's an a priori reason it needs to go (close) to zero either. I mean, I guess, like, "efficiency" and "competition" and so forth sound like a priori reasons, but then Visa's fees aren't zero.