PCC is a whole different level. For example, you still have to trust that Google is doing what it says to control access. PCC makes it auditable and verifiable by clients when connecting to a node.
You can also audit that the binaries don’t leak any data in, say, debug logs, which is definitely possible on GCP/Borg. PCC nodes are “cryptographically airtight.”
https://archive.is/HFlx1 is one example of the lengths they are going to - third parties running entire Google cloud software stack (i.e. the GCP stuff we all know, and the underlying infrastructure too) in their own data center that is entirely air-gapped from Google itself. This is a huge undertaking.
There's a huge difference in what GDC-Airgapped runs on and what is actually running in GCP.
Airgapped is based on top of Kubernetes, and it's using mostly off the shelf components for networking and compute.
GCP is based on top of Borg, using custom networking and computer hardware (though manufacturers by a partner). As a note, not only is access without a support token alertable (goes to your skip level), there's a distinct level of "I'm not even going to build the tool to enable this". Which makes being in support a b**.
If you want access to something, it's significantly easier to just ask the customer to do it themselves.
I don't know what details I would be allowed to share, so I'd better not share any. You can try looking it up on the Internet.
But Google does a lot of work to protect against insider threats, because everyone understands that in an organization of this size there will always be bad apples, spies, etc. Google's systems are designed to protect customer data from malicious employees using technical measures; it's much more than just "if we catch you, you're fired" as was asserted upthread.
This is not true for Google, at least. I know because I work at Google.
So I wonder how accurate your knowledge of Meta, Microsoft etc is.