If any of the candidates they are refusing are capable of doing the daily work at a FANG but can't pass the interview, they're artificially constraining their supply of candidates and increasing the cost they need to pay for developers for absolutely no reasons.
The type of work in a FANG is mostly not that dissimilar to other companies (the exception being teams working with machine learning, performance optimization, dealing with tricky scale). I understand hiring specialists for specialist jobs (and I still wouldn't test them on leetcode; ask domain specific questions).
It's just ego driven over spending on developers.
It's worth considering that engineering interviews are largely conducted by other engineers, and engineering hiring bars are largely set by senior engineering staff. They have a pretty vested interest in maintaining their own status and (relative) scarcity. Constantly raising the technical interview bar keeps engineers a supply-constrained resource...
On the contrary, rejecting a high number of candidates pushes market rates down.
The incorrect hidden assumption is that by decreasing your own supply the value goes up.
It's well known that turning down candidates or refusing to interview them sends the message that their skill are less valuable. It's HR management 101.
The same goes for "anti-poaching" agreement. If FAANGs don't hire from each other they reduce the number of high-paying employers willing to hire the average FAANG employee, effectively reducing average salaries.
There was a big scandal about this.
This doesn't fit the "we can't find candidates, so we need more H1B's" narrative.
Supposedly, there are many open slots, unfilled, so purposefully failing people which can do the job, which meet the minimum requirements, should not be the outcome.
Instead of raising te bar, as you say, due to candidates o'plenty.
I disagree. In fact, this validates it.
On the basis of a world-wide talent pool vs a domestic. If you know there are stronger candidates elsewhere then you absolutely want to test away the domestic candidates.
Thinking about it from a hiring manager's perspective. You don't lower your standards just because a specific pool of candidates can't meet your criteria when you can widen the pool - as far as you are able.
If the H1B candidates were tested on a 'easier/weaker' test the I would 100% agree with you, but I'm assuming things equal here. And, ignoring any thing related to domestic vs foreigner workforce, "people taking our jobs" debate.
(but I'm not from the US so don't actually care)
This is okay and to be frustrated at it, but it's important to recognize why it's happening.
FAANGs go through an incredible number of candidates with only a few slots (comparatively) and the point is to simply make the bar higher and higher. Just like a pragmatic engineering - you have a 'good enough' candidate set.
For me, the problem is when smaller startups copy the format - expecting candidates to jump through all the same hoops. If a FAANG has 1:100 position to interview candidates ratio, a startup will be lucky to have 1:5 (it's incredibly expensive/time-consuming for a startup to interview).
Running the same test likely means you're rejecting a lot of potentially good candidates who didn't want to go through a 'FAANG' interview.