The AI has several specific weaknesses that you might be able to exploit. It relies on general strategic algorithms but it’s lacking in tactics. In particular, it can’t see any plan that you may draw against it when it’s two turns away, and there are some specific tricks that it will fall for, too. I’ve played it a few times, and while I’d like to emphasize that I greatly respect the impressive work of the programmer who created it, it’s not even close to a somewhat skilled human player. The probable reason why you’re having some trouble defeating it outright, could be that you “believe” it too much because its moves make sense when you look at it superficially.
Having said that, in your actual game, since it looks like Europe is pretty much under control, I wouldn’t bother too much about that theater right now, because like you said, Japan may win the game if you do. You say that Japan will take Calcutta and then be 1 VC away from the 6 it needs, so I suppose that Hawaii and Sydney are still in Allied hands. Without additional information it’s hard to say whether either of those is likely to fall, but against the AI and knowing what it does, I’d go for knocking Japan straight out of the game. Typically, if you equip a strong US fleet with a bunch of transports (probably a few more than the 10 you’re using for Spain right now) and it’s in SZ10 (near Western US), then any human player would start to make sure that Japan itself is properly defended. But the AI won’t - it will only see you coming once you actually move that force to SZ26 (near Hawaii). So I wouldn’t send those units to Spain, but rather, in the other direction. Even if you can’t take Japan, you can reclaim the Philippines and the Dutch East Indies, which is a huge money boost to the US.
As a disclaimer, I’m saying this without any knowledge of the actual situation in the Pacific in your game. It’s based on my observation that the AI will happily leave Japan almost undefended if it can send its units far away to conquer territories.
Disclaimer: