People always think of these companies as if all they were doing was simple credit card processing, delegating a lot to a bank. If that was the case, then yes, they are commodities with no moat. Reality is far more complicated.
It's not that a large merchant needs to support one payment type for one country: They need a collection of payment types that work on each country, without having to dedicate engineering to every single one. For companies that are marketplaces, like Amazon or Uber, you need to have payments, both ways, including onboarding. And for online companies, there's the shadow of fraud: Not just fraudulent purchases, or card testing, but, as a marketplace, fake merchants. And the larger your payment company is, the easier it is to do good fraud detection.
When you put it all together, the amount of engineering effort those companies are saving you vs 'use bitcoin LOL' is enormous. And instead of shrinking every year, it's growing. So, for instance, see how Amazon decided that maybe they wanted to do fewer payments in house, and have Stripe manage it for them. It's not as if Amazon lacks engineering capacity, or if they started without a payment system: It's that there's a bunch of value there, and they thought that it'd be more profitable to delegate more and more of the system to one of those not-so-commodity businesses.
But yes, if all you needed was the simplest payments, in a lone jurisdiction, it's a low margin business. The moat is elsewhere.
I'm not saying that what Stripe, Paypal, Adyen, etc do isn't a lot of work, I just don't see how there's a moat there. For example I'd be shocked if the Amazon deal wasn't on terms that basically give Stripe very little profit - like you said, Amazon is capable of doing it themselves, it would only cost them XX million worth of engineering salaries, and Jeff Bezos isn't known for giving away money.
Fraud detection on the other hand I'd believe could be a very high margin business, the "moat" being proprietary techniques that perform better than the competition, better data (as you mentioned), etc. However I've never heard anybody sing the praises of any of these platforms' fraud detection products - is that really where they are making their money?
It's not that a large merchant needs to support one payment type for one country: They need a collection of payment types that work on each country, without having to dedicate engineering to every single one. For companies that are marketplaces, like Amazon or Uber, you need to have payments, both ways, including onboarding. And for online companies, there's the shadow of fraud: Not just fraudulent purchases, or card testing, but, as a marketplace, fake merchants. And the larger your payment company is, the easier it is to do good fraud detection.
When you put it all together, the amount of engineering effort those companies are saving you vs 'use bitcoin LOL' is enormous. And instead of shrinking every year, it's growing. So, for instance, see how Amazon decided that maybe they wanted to do fewer payments in house, and have Stripe manage it for them. It's not as if Amazon lacks engineering capacity, or if they started without a payment system: It's that there's a bunch of value there, and they thought that it'd be more profitable to delegate more and more of the system to one of those not-so-commodity businesses.
But yes, if all you needed was the simplest payments, in a lone jurisdiction, it's a low margin business. The moat is elsewhere.