That makes sense in the abstract but of the dozens, maybe hundreds, of projects Google is currently working on, how many outside of Ads make a sizable amount of profit?

A mildly successful cloud provider would probably make FAR more money long term than the median Google product. Pulling the plug on GCP to build a couple dozen projects that will mostly get killed themselves seems like a huge waste, even without considering all of the negative externalities of burning even more customers by cancelling a major service.