Not even sure there's much difference between 3000, 4000 and 16000 in practical terms. They all just mean the computer crushes you and laughs in your face.
I was running through one of my games on Fritz a couple of weeks ago, and out of the blue it announced forced mate in 25 for me, starting with a move that I had dismissed as being bad. I guess the only difference might be that the 4000-rated engine would announce mate in 50 or something like that 
To the OP- it is theoretically possible to "solve" chess, but current technology makes it completly impractical. As you pointed out it requires storing a vast amount of data. There is also the issue that you would need to perform a huge number of calculations. A few years ago a computer scientiest managed to solve checkers, which is a vastly simpler game, and the calculations on that took something like 20 years to finish. We would need a whole new level of computer technology to do it.
Stop > Go
4 > 3
<3
it may be solved sometime in the far far far far away future.
I believe so.
Some day, we can create a computer that is able to analyze all positions in chess. The computer can come anytime, when the teq is here. 100-5000 years?? But im sure it will come, if not all humans die.
Who is winning is the question. Some thinks White is winning because he has the first move. Some think it will be draw, because both players play perfect. And then some (including myself) think black will win, because white will be in zugzwang.
But of course, chess will never be boring, because noone will of course remember all the lines.
Really? You believe that the opening position is zugzwang? I find this to be the most unlikely of the three possibilities and lean very heavily towards it being dead drawn.
Probably one of the reasons the game had been played for centuries by countless people is because it is universally perceived that a perfectly played game has to be a draw.
Who ever solves chess it will happen to him what happen to a last guy that solved it.
Measuring the atoms in the universe must assume a finite universe. I don't understand such claims. Best to stick with an infinite universe, it gets the pesky, cartoonish cosmological gods out of the way
These estimates are based on the known universe. Whether the universe is infinite or not is irrelevant.
"Maybe you don't need to store every position to solve chess. Maybe you just need a formula or something, that can calculate the best move for every position."
He is totally right. An algorithm, to be exact. which is already being used, isn't it? Furthermore, let's say you want to store all the positions, you would only play 1 move for every answer from your opponent, right?
This already means that you have to store only 1 move, instead of 16 (all your pieces) in the beginning, since you could have calculated before what that perfect move would be for that given position. This effect ofcourse declines as you get less and less pieces, but still allows for significantly less moves.
But I would use the algorithm though =P
I read this somewhere, and apparently the "perfect" chess computer would have an elo rating of approximately 16,000. Since the highest current rating any man/machine has now (the year 2010) is about 3400 from Rybka 4.0, we still have a long way to go.
That doesn't make much sense, elo ratings are not absolute measurements of skill, they just show how well you perform against others in the same rating poll. So a 16000 rating would be pretty impossible to get by playing against 3000ish computers, even with a 100% win ratio-and even if the computer played enough games to rise to 16000, all we'd know is that he has a perfect score against 3000ish engines, but that could be a skill level of "just" 4000.