They run a decent amount of their own compute/bare metal server for customer workloads. But likely still had some critical dependencies on GCP.

> Surprising for a company of their scale with all this VC money.

Not sure too many VCs would be cool with deep redundancy when there's more features to build to bring in more customers instead.