How to force AI to make the absolute best move every time?

I don't know, I may be using the wrong terminology. I meant the "computer" or chess engine, yes.

Just boot up Stockfish and run it on AWS.
Getting depth 32-35 doesn't take more than ~30-40 seconds in the cloud analysis, I think that is not too long, and that depth should be sufficient for the vast majority of best moves?
Sometimes the cloud analysis stops at depth 28 or 32 etc and does not progress any further, so even playing manually in the analysis mode doesn't fully provide a solution.
AWS is amazon web services? Sure, thanks, I can try that out. I still would love to do it on chess.com though, for the game review etc and just doing it all in one place.

Playing against AI even at maximum 3500 it still often makes "Excellent", "Good", or even mistakes at times. If you run the moves through analysis with maximum depth then like half of the moves "maximum 3500" AI makes are not the best.
One interesting question: which one is stronger? The maximum AI or the chess.com engine at maximum depth?
I have a feeling that both of those are quite weak compared to a real engine running on a good computer.
I also think that the Maximum engine is weak, despite the name. 3200 rating on chess.com is human super-GM strength, so pathetic for an engine.
But lets say I you actually want the AI to make the best move everytime, how do setup that up?
AI will not do that except by chance. If AI (or any system) could do that then "chess" would be solved. Just ask for the best move in the game starting position with the 32 units. Then it would have to know the final score of chess.
But theoretically the traditional brute force engines (given sufficient quantum power) would solve chess before any AI-system as their search strategy guarantees that all relevant positions are visited. All alternatives are more efficient and consume less computing time but are also less certain.

The chess.com engine at maximum depth is stronger than the maximum AI. The chess.com analysis uses Stockfish by default, while the maximum AI uses Komodo, and will only analyze for about a second each move.
I have played Stockfish against the maximum AI and Stockfish won. Here is the game: https://www.chess.com/game/computer/73770009
If you want to play against Stockfish, just download an interface and Stockfish. I reccomend the nibbler interface.
https://github.com/official-stockfish/Stockfish/releases
https://github.com/rooklift/nibbler/releases
Playing against AI even at maximum 3500 it still often makes "Excellent", "Good", or even mistakes at times. If you run the moves through analysis with maximum depth then like half of the moves "maximum 3500" AI makes are not the best. Of course this is useful to train yourself to take advantage of the mistakes a human player would, and also to get variety to each game.
But lets say I you actually want the AI to make the best move everytime, how do setup that up?
One way is to play in the analysis mode, set to max depth, or cloud analysis, and then analyze every move up to 40 depth and then manually move the pieces. It's a bit tedious though, can you not make the AI do it by itself?