Doesn't matter (much). High capability attackers can, and do, place agents within major infrastructure companies to figure out stuff like this. If there is a backdoor, it is virtually guaranteed to be compromised by more than one actor eventually, and they're likely to share the data (if not the access) with their intelligence allies.
Sometimes, companies willingly provide the access (e.g. NSA closet at AT&T), other times, it is more sneakily obtained. But, a system with a backdoor is much more likely to be compromised than a system without one.
Edit: I should be clear that any closed source infrastructure is potentially subject to the problem of infiltration. But, a company intentionally putting a hole in the system makes the job even easier. The opportunities for an infiltrator to poke a secret hole into the system are much smaller than the opportunities for obtaining the key to an already existent hole. A hole that is protected from discovery by the company itself is much less likely to be detected by other teams, etc. I mean, to really make a secret hole, you need things like process and network activity statuses to ignore you, logging to not see you, you can't show up in an obvious way on the filesystem, etc. You probably have to have cooperation across at least a few subsystems. It's much easier to exploit an already exploited system, is what I'm trying to say, because you don't actually have to exploit the system, anymore, just the people/organization who builds the system and people are easier to crack than encryption keys.
