From watching the world championship match and many of the really high level tournaments I get the impression that 1.e4 is not too popular at the very highest levels in the last few years. It seems like a much higher percentage of games open with 1.d4 or eventually transpose into a 1.d4 opening.
Do you agree? Are there any statistics somewhere about the top 20, top 50, etc. players that would prove or disprove this thought?
A lot of high level players play 1. e4 e5
Even Carlsen
I think the Petroff and that Sicilian Gelfand plays with the backward d-pawn (I could never spell it, begins with “S”) have been so drawish as to take the wind out of the sails of e4. However, things always change (like 4.Nc3 getting popular in the Petroff). Hopefully e4 isn’t dead!
Wasn’t there a line in the King’s Gambit, that for years, made it look like that one side (forget if it was white or black), that was practically a forced win?
Then many years later, someone found a defense against the line that brought new life to the King’s Gambit?
Of course at anything other than the stratosphere of chess players, the King’s Gambit does make for some interesting games.
Can’t speak for the GM level of play though. 
As a side note: My most memorable game lately (if I can ever find that game on my computer again), I played a VERY interesting Smith-Morra Gambit. I was black. The opponent let his gambit pawn hang too long and after a few moves I took it… practically inviting my opponent to go into an overdrive offence. He ended up giving away his pieces in a vain attempt to keep his attack going. 
I lost a couple pawns in front of my king. But I had a 3 piece fortress (plus one pawn) around my king he couldn’t break through. Eventually, his attack ran out of steam. I could practically feel his frustration… even though it was an internet game.
This is probably confirmation bias (I play the French), but it seems like every time I see the French defense used in top games, Black loses. So e4 may not be unpopular, but e4 e6 may well be.
I looked at all of the TWIC (The Week In Chess) databases released since the beginning of this year. That includes some games in late 2011 and all of 2012 so far.
The total number of games was 116,779 - of these 56,044 started with 1 e4 (defined as Encyclopedia of Chess Openings Codes B00 - C99), or 48%. TWIC is generally publishing reasonably high-level games, so I would say that this indicates that yes - 1 e4 is popular at the highest levels.
In this group, the number of games that began or transposed to a 1 d4 opening (defined as ECO A40-A99 & D00-E99) was 44,190 or 37.8%. So 1 e4 was more popular.
I then took just the same TWICs and limited to both players above 2500 - this produced 8,539 games of which 3,257 began with 1 e4. This is 38.14% Obviously a drop, but still a sizable portion of the games, so again 1 e4 is popular at the highest levels.
In this group the number of games that began or transposed to a 1 d4 opening were 4,027 or 47.2%. So in this case 1 d4 edged out 1 e4, although both were obviously popular.
I used to get the NIC Yearbooks, and they would generally have a comment in each volume on the breakdown of opening in their classification system, between 1.e4, 1.d4 and alternative first moves that did not transpose into one of these (since many 1.c4/1.Nf3/1.g3 openings end up transposing into 1.d4 systems). My recollection was that 1.e4 was played more often, but that the 1.d4 systems scored better for white, and NIC attributed this primarily to the Sicilian Defense, which scored better for black than other main-line defenses. I’ve stopped getting the Yearbooks but wouldn’t be surprised if this was still the case.
What is striking to me is how the repertoires of the top players have expanded. More of the top players have made themselves harder to prepare for by not playing a narrow range of openings. They play 1.d4 more often, but they also do not hesitate to play 1.e4, 1.c4, or 1.Nf3 in order to capitalize on their opponents’ lines or to avoid certain move orders. The days when a top player could play only a few openings an entire career are a thing of the past. If Kasparov is correct that there was a “revolution” in openings, it was in the willingness of the top GMs to use computers to expand their openings horizons.
Perhaps there is a lesson in there for the rest of us?
That once we make 2700, we should broaden our repertoire?
I started phasing out of the Najdorf 20+ years ago when I realized that many lower-rated players knew more theory than I did in their pet anti-Najdorf line. This is still true for their pet lines after 1…e5, but I feel somewhat less likely to lose in the opening.
1.e4 is more forcing while 1.d4 allows for more choices (and creativity) by both players. Many 1.e4 openings require precise knowledge of theory to move 20 and beyond, while even Grandmasters may find themselves out of “book” by move 6 or 8 after 1.d4. Of course, there are exceptions in either case.
For example, I remember Karjakin - Anand (2006) where black beautifully sacrificed two pieces for a mating attack in an ultra-sharp Najdorf. Alas, Anand later admitted that the entire concept was preparation that he had done with his computer. The young Karjakin got schooled by the veteran.
[Event “Corus”]
[Site “Wijk aan Zee”]
[Date “2006.01.14”]
[Round “1”]
[Result “0-1”]
[White “Karjakin”]
[Black “Anand”]
[ECO “B90”]
Michael Aigner