Shouldn't the engine evaluation be either equal or winning at all times?

Lol. Is this for real?
"If both sides play the absolute best moves" - what are the absolute best moves? No one knows and may never truly know.
"... it would be a draw" - this is an assumption, we don't know for sure

Chess may have roughly 10^120 games possible. This is far larger than the number of atoms in the known universe. 

If you had an infinitely powerful computer which could compute all 10^120 chess games, it would either say mate in X or draw in X at the start of a chess game. We may never see that happen because it may be too many to ever solve completely.

Well, that's just not true.  To begin with, consider positions where one player just sacrificed a rook to achieve a forced mate.   Next, consider positions where the player that is up a rook needs to give it up to avoid a forced mate and stay in the game.    Then consider the previous situation, but where the rook has to be given up after so many moves of 'perfect play' that not even the best computers of today can calculate it.

It's not even really that crazy to imagine such a position.  Imagine a player failing to develop a rook and knight out of the corner and ending up with their king out in the open with all the opponent's pieces near it. 

There are billions of positions where it takes just  one glance to determine win  lose or draw.

Maybe that's how we can finally solve 10^120 chess games - design an engine wife, to nag the engine for being lazy.

The OP seems to be under the impression that any advantage, however small, is sufficient to win. There is absolutely no evidence to back up this point of view, and experience indicates the opposite: that it requires a fairly large advantage to win, and that intermediate cases will simply give one player the ability to apply unpleasant pressure to the opponent for many moves, in an effort to induce another mistake that will finally yield a winning advantage.

That would be impossible. Even making the first move would require an modern supercomputer 10 to the power of 90 years. There are just wayyy to many chess combinations to count for. We don't even know the number of possible chess positions.

Ah, but if we ask a legendarily powerful supercomputer we might get the answer to the number of possible chess positions - maybe 42 (well, it might be 10 followed by 42 zeros but 42 would need to be in there somewhere).

Bc if engine tried to calculate all of that even with 7,000,000 espressos and energy drinks within, lets say, a day, it wouldn't be able to calculate all the variations.

And if you tried to force it, it would just explode.

(Also why would you give energy drinks to a computer? It can't even drink.)

It's not a robot.

It's a computer.

Batteries then

I understand your point.

In a way, I think you are right. When you have a forced mate in 3 is shows M3. I you have a mate in 10 (M10) it will show it as a definite win. As computers become more powerful it might show M24 or M37 more often.  So, in principle you have a point, computer engines becoming stronger will lead to wins being predicted earlier and earlier.

But where I think this is a misrepresentation, and that's also what I see in the comments, you are invoking a super computer that might or might not ever exist. If your super computer would exist you would be right. 

If you would let AlphaZero play a +5 position against another AlphaZero a million times, the first AlphaZero would probably win a million out of a million. So, if your statement is, given a significant enough advantage, will that result in a win? Yes. If your question is, can a hypothetical engine that does not exist, guarantee any winning position as a definite win. Yes, probably. If your question is will engines ever will be powerful enough to mark said position as a guaranteed win, in the mathematical sense? Maybe, but perhaps not...

this is confusing. 

Maybe the perfect chess game requires 42 moves to complete...

i dont think your premise is correct. if both sides play the best first move the computer should (and will) show white with a slight advantage, not equal.

but the first move only shouldnt be that hard to figure out. white has only 20 options. one of them is the best. black has only 20 responses. total possibilities after one complete move (by both white and black) is only 400

confusing confusing confusing

LOL