Security only through obscurity - sure.
But obscurity as an additional layer, as part of a defence in depth strategy, still has some value.
It’s rare for any large org to publicly discuss any details of its security design, let alone a bank. Monzo must be supremely confident in their system to go public with this information, or judge that the marketing/recruitment benefit outweighs any potential risk.
What if AWS gets cracked/hacked/compromised?
I know it's not happened yet, but it's not impossible.
* For individual banks and their customers, is it more likely that an AWS-wide exploit will compromise an AWS-hosted bank, or is it more likely that a self-hosting-specific exploit will compromise a self-hosted bank?
* For society, is it better that security efforts are concentrated in on centralised providers like AWS, or is it better that security efforts are distributed, on individual hosting entities?
In reality it's always about tradeoffs: who to delegate to and who to trust.
The difference is that cloud relies on public services, which once compromised (e.g. via social engineering), allow for lateral attacks resulting in much bigger impact (e.g. Lapsus$) across the complete customer base. This makes social engineering much more attractive in cost vs impact. The resulting monoculture in not only the software, but the infrastructure and configuation also increase the impact on technical attacks on specific exploits.
What are the public services that AWS relies on, and how are they different from a bank's server farm, or a bank renting out space in a datacenter?
The same, really, applies to all other concerns.
All of these services are hosted by AWS in a multi-tenant fashion, sharing not only the code, but infrastructure and configuration patterns.
Is not advertising your security architecture "security through obscurity"?
As to your second point, security through obscurity is generally believed to not be worthwhile.