Cloud infrastructure, the Go programming language, Kubernetes. Now of course your definition of "change the world" comes into play here, but it certainly hasn't just been advertising and acquisitions.
Indeed.
Cloud Infrastructure = playing catch up to a player that set the stage (and still leads), Go I don't think "changes the world" quite as much as consolidates and improves things that already were available in everything from C to Python, but suited to a specific set of use cases that Google particularly cared about for their own business first and the "world" later. K8s came out in roughly the same window as other orchestration tech (e.g. Mesos) but had/s the benefit of the 'Google' name attached to it. It simplifies some things certainly, but is it really a "world changer"?
TBH, more seems to be about changing Googles world more than the broader one, but that goes to your original quote.
Borg's real win was the insane colocation and oversubscription it enabled, both of which K8s and all major public clouds have failed to replicate because you can't securely share bare metal between customers.
and K8s ended up being too little too late, despite beating Mesos and being quite successful, because VMs had won by that point.