By means of computer simulations of a coarse-grained DNA model we show that the DNA hairpin zippering dynamics is anomalous, i.e. the characteristic time τ scales non-linearly with N , the hairpin length: τ ∼ N α with α > 1. This is in sharp contrast with the prediction of the zipper model for which τ ∼ N . We show that the anomalous dynamics originates from an increase in the friction during zippering due to the tension built in the closing strands. From a simple polymer model we get α = 1 + ν ≈ 1.59 with ν the Flory exponent, a result which is in agreement with the simulations. We discuss transition path times data where such effects should be detected. The folding dynamics of DNA (or RNA) hairpins, which are single stranded molecules forming a stem-loop structure, has been a topic of broad interest within the biophysics community for a long time [1][2][3][4][5][6][7]. Hairpin folding is a prototype example of secondary structure formation [8] and shares common features with the more complex case of protein folding [9]. In both cases the folding process is described by a one-dimensional reaction coordinate performing a diffusive motion across a free energy potential barrier (see e.g. [10]). Recent advances in experimental single molecule techniques allow to monitor the folding of hairpins [7] and of proteins [11] with an unprecedented time resolution. These and future experiments are expected to elucidate many aspects of the folding dynamics [12], the reason being that the actual conformational changes occur on timescales which can be typically a few orders of magnitudes smaller than the total folding time [11].The aim of this letter is to investigate the folding dynamics of DNA hairpins, focusing in particular on the rapid zippering which follows the formation of a stable nucleus of a few base pairs. The latter process is generally much slower as initially the hairpin undergoes a large number of failed nucleation attempts. We show here that the zippering time τ scales with the hairpin length N as τ ∼ N α with α > 1. This conclusion is based on extensive simulations of coarse-grained model of DNA and on scaling arguments for polymer dynamics. Our results are at odds with the zipper model [13] which assumes that the hairpin closes like a zipper following a biased random walk dynamics, which implies α = 1. The results give insights on the forces involved in the folding process and in particular in the role of frictional forces. In addition, as argued at the end of this letter, recent experiments on transition path times [7] appear to be better described by a non-linear dependence of zippering time vs. N , supporting the results reported here.