What good is “building software systems” if you can’t produce products that people and businesses want? After all of these years and all of the initiatives, still almost all of their profits come from ad sales.

It's entirely possible that this state of things will continue. Google will continue to be a/the leader in online ad sales, and periodically they will try to expand into some other lines of business, never with any particular success. That's not the best possible place to be, but it is a good one, particularly since ad sales are so very lucrative.

As long as the trend of blocking ads doesn’t increase and as long ad buyers don’t see Facebook as being better at targeting.

Ads on the FB app are almost impossible to block.

I think it goes even a little deeper than that: "help yourself" can be scaled to a level of profitability that is objectively impossible to reach once you put humans in the loop. I don't think that Google is interested in diluting their existing high profitability activities with lower profitability, so whenever they enter a new market like cloud computing they effectively limit themselves to the subset of that market that is open to absence of human support. Just like on the other end of the scale, Oracle is limiting themselves to the subset of their market that will gladly pay extra for the privilege of getting overrun by salespersons.

edit: in other words, I think Google would rather have a highly profitable slice of the market than dominate it and have all the gains eaten up by a greedy sales org.

I think all companies would want to "have a highly profitable slice of the market than dominate it", since it ensures higher profits and frees you from monopoly accusations, but only few have managed to work it in a good way (e.g, Apple). In most markets where Google succeeded, they ended up being dominant.

Apple didn’t do that by reducing cost. They have hundreds of physical retail stores where they hire people directly (not contractors). They did it by offering a product that people were willing to pay a premium for.

Does anyone think that GCP is successful by any metric compared to Azure and AWS?

The real question is, does Google really need to dominate to be successful? We should be rating success by the quality of a product and market share is a relatively poor proxy for that.

The quality of the product especially for enterprise products has a lot to do with the quality of support. Google also lags there.