It's weird to say, but there's a genuine lack of headcount all over the company while Pichai (and probably Zukerberg) also does a correct assessment on the situation. This seems a contradiction, but if you take a deeper look into mid level managements there are a good number of teams responsible for billions of users/revenue and driving their growth but screaming for more headcounts. (Yeah, I'm in one of those teams) # increases to hundreds (or thousands?) at smaller scale. IMO, they deserve more headcounts for their scope. But it's also clear that the overall productivity of those companies begins to plateau.
Why does this happen? Of course I don't know. I've seen some clues on bigger structural issues but cannot say for sure. But the famous "I just want to serve 5TB" video gives us some hints... Most of the particular issues mentioned in the video have been solved but its spirit hasn't gone away. And now back with a good reason. Which makes it much harder to solve.
Think about launching non-trivial but small features in their major products. At a small company, a competent junior engineer can usually do that within a quarter. In Google it's not that simple. There are so many stakeholders. Privacy and security. Legal. Downstream dependencies. Infrastructure team. PA wide modeling and quality review. They're also busy and might not like your launch. At least PM will likely be your side but they may have a different priority than yours. To navigate this organizational complexity, you probably want to have a good manager/tech lead. If you don't care? You're going to piss off them for sure and if the things go very wrong then you could get indivisible attention from the VP level...
And you're now dealing with several hundreds of millions of users so a minimum level of engineering quality should be ensured. You gotta deal with resource planners who also need to allocate finite hardware resources among unlimited demands. The service should have some level of reliability, scalability and redundancy. Thanks to all the works done by core and technical infrastructure team, this is easier than other places but the inherent complexities don't go away. Oh, did I mention that most of the complex infrastructures have integration tests that run over 1~2 hours with a good level of flakiness? If the build dashboard doesn't go green, you might miss your launch by 1 week. It's just a tip of iceberg for productionization, multiply the work by 10x. This is a death by thousand cuts and I don't see a silver bullet to solve everything at once.