Completely agree about the billions.
A five year old human is a General Intelligence. That human has not had all of the world's knowledge handed to them. This suggests AGI will be structure, not compute.
So throwing billions to do a lot more compute won't break through the AGI ice. As you say, it doesn't mean it won't be useful, but it won't do AGI.
I think the agentic approach will work in the end, or accomplish roughly what they are trying to do with AGI.
There is certainly a great deal of AI washing, greenwashing and quantum washing in the media being shared by the big companies. Again as you say, a lot of cash has been spent and they don't want it to all wash down a drain. If they can convince us all it is the next big thing then they hope for large scale adoption.
Even if they can't give a particularly compelling set of use cases yet. They've made a groovy tool, but they're not sure exactly what to use it for.