It was a major source of licensing revenue for Sun, and one of the main reasons why historically they had been reluctant to open source Java. They were making around $200M revenue from it in 2007, which was the major source of income for their software business.

I thought this part of an Ars Technica article[0] on the trial was pretty telling:

"On cross-exam, a Google attorney brought up a graph from an internal presentation by Brenner showing "aggressive" and "conservative" estimates of what would happen to Java licensing revenue from 2007 to 2010. The graph's "aggressive" line showed a decline from around $140 million per year to about $105 million, and the "conservative" line showed a decline from the same starting point to around $50 million.

The graph was created before the launch of Android. Google's point was clear: Java was in decline, Android or no Android—and its executives and salespeople knew it."

It doesn't really matter what the numbers were though. They would have been higher had Google paid Sun for a Java license.

[0] http://arstechnica.com/tech-policy/2016/05/suns-head-of-java...

That's not relevant to fair use analysis. What's relevant to fair use analysis is if Java's market would have been bigger if Google neither bought a license nor produced Android. Obviously, any unlicensed use of a copyright protected work is going to reduce the caller of the protected work compared to the exact same behavior coupled with paying for a license, but that's not what the market effect factor is about.

You miss the case where Google had produced Android with a different platform (LLVM, Objective-C, or their own language). Since Java was in decline on mobile, it should have just entirely ripped Java from the mobile market.

I don't miss that case, I just don't believe it's the relevant one for fair use analysis. Though, if it were, that would weigh even more heavily in Google's favor.