We report on a survey devoted to the search of exoplanets around young and nearby stars carried out with NACO at the VLT. The detection limit for 28 among the best available targets versus the angular separation from the star is presented. The nondetection of any planetary mass companion in our survey is used to derive, for the first time, the frequency of the upper limit of the projected planet-star separation. In particular, we find that in 50% of the cases, no 5M J (or more massive) planet has been detected at projected separations larger than 14 AU, and no 10M J (or more massive) planet has been detected at projected separations larger than 8.5 AU. In 100% of the cases, these values increase to 36 and 65 AU, respectively. The excellent sensitivity reached by our study leads to a much lower upper limit of the projected planet-star separation compared with previous studies. For example, for the β Pictoris group (~ 12 Myr), we did not detect any 10M J planet at distances larger than 15 AU. A previous study carried out with 4 m class telescopes put an upper limit for 10M J planets at~ 60 AU. For our closest target (V2306 Oph; d= 4.3 pc), it is shown that it would be possible to detect a 10M J planet at a minimum projected separation from the star of 1 AU and a 5M J planet at a minimum projected separation of 3.7 AU. Our results are discussed with respect to mechanisms explaining planet formation and migration and forthcoming observational strategies and future planet-finder observations from the ground.