This is a follow up to my earlier post on Crypto.

I sent the article to RS, and he came back with some thoughtful comments. I thought I'd record some of my replies here.

1 Lower Cost

The main reason you point to for the lower cost for crypto seems to be scale – is that right?  And is there an intuitive reason why crypto should be more cost effective than traditional forms of finance?

I think the starting point for my analysis is the observation that Crypto appears to be cheaper than traditional finance for basic account to account transactions right now, regardless of the "why". This is made even more puzzling by the fact that Crypto is actually sub-scale at present, and certainly lower scale than traditional finance right now. The reason I'm "scared" is because I expect the cost of Crypto transactions to drop even further as it achieves scale.

It's not obvious why Crypto should be cheaper to be honest. An unsatisfying explanation might actually point to the fact that traditional financial institutions are just bad/inefficient at doing at IT upgrades (this is definitely empirically observable), while Crypto companies, being Tech-adjacent, are just more efficient at building out IT systems. (A fun case study on this would be why NAB bought Ubank rather than building a competitor in-house).

The relevant scale advantage for Crypto might actually be on the workforce/supplier side rather than in the network itself (i.e. global Crypto standards might mean that a bank can hire a very competent team of contractors in Eastern Europe/India/South-East Asia to build the integration into your interbank payments network rather than having to hire from a narrow pool of domestic specialists)

2 Cost of verification on Crypto

On the cost of Crypto verification (vs govt fiat) you're correct only as far as traditional blockchains are concerned. (Eric Budish has a recently published paper in QJE on precisely this - specifically the cost of traditional "Nakamoto Consensus"). 

The newer/layer-2/more-abstract blockchains are all experimenting with new and different modes of verification that are much cheaper. I would say that a few things that we've learnt in the past 5 years or so that have helped: 

• Blockchains are harder to attack in practice than in theory, and so more relaxed verification methods seem to work fine
• Mining on a blockchain usually ends up being done by a specialised entity, which means that you can demand more of the entities doing the mining (e.g. relying on miners to be online ~100% of the time and punishing them if they're not)

If you ask ChatGPT to explain to you the security model of say Solana vs Bitcoin, I think a lot of the differences/efficiencies start to become more obvious.

3 Cost of regulation

The cost advantage here is unrelated to the cost of regulation.

This is the cost of pure messaging (i.e. sending lots of messages in a structured way, with expectations on availability and confirmation). 

Stripe actually sells you AML/KYC as part of their package of crypto services, and will expect you to complete AML/KYC to register for crypto withdrawals. Similarly, all the major crypto exchanges involved here will do AML/KYC. 

4a International Transfers Generally

My suspicion is that you (and I) are probably too sophisticated/wealthy to want to be doing stuff directly in Crypto right now.

The question I'm posing is more like: Actually, would Wise save money if it used Crypto on its backend?

4b Wise Fees

Also, Wise's fees don't look so great to me. I've used it to send money to a friend in the US, and I was not impressed.

I just tried to see how much it'd cost to send $1,000 AUD to Europe - Wise is quoting $4.86 AUD in fees. That seems very high for a bare message. 

By comparison, I could shop in the US with my Mastercard card and pay the end of day FX rate and no additional fees. It isn't obvious why "sending money" shouldn't be similarly priced.

Maybe the comparison here is to long distance phone call rates vs Skype (when it first came out). 

Looking at the price of $4.86 for an international transfer, feels to me a bit like how long-distance calls used to be priced. Not so high that you could never make one, but not so small that you'd think of it as "effectively free". VOIP/Crypto should then take that cost down to something you'd sell via freemium models.