As part of our deep multi-colour survey of two gravitational lensing clusters, Abell 1835 (z=0.252) and AC114 (z=0.312) used to search for very high redshift galaxies we have detected a number of extremely red galaxies (EROs, R-K[s]> 5.6). Following up on our deep optical and near-IR imaging (Richard et al. 2006) we used deep HST/ACS imaging in the z[850] band and Spitzer IRAC and MIPS photometry to further constrain the photometric redshifts and to classify the nature of this objects. New Chandra observations are also being completed. Based on different classification schemes, e.g R-J vs. J-K colour or a combination of near-and mid-infrared colours, we try to discriminate the two most popular EROs populations, elliptical galaxies and dusty starbursts. As result we would classify most of the EROs as starbursts. The spectral energy distribution of the majority of EROs with a detection in two or more IRAC bands reveal starbursts at z ˜ 1.5 to 3 with a considerably extinction (A[v]>3). The strong field to field variations in surface density δ=1.1 arcmin-2 for Abell 1835 and δ=0.16 arcmin-2 for AC114 indicates that the numbers and sky distributions of EROs are very different for each field of view. This is consistent with the results by other wide field EROs surveys, which suggest a strong clustering of EROs. The average surface density (δ=0.63 arcmin-2) is by a factor 2 lower than in other surveys, which possibly results both from the field to field variations as well as from the requirement of these EROs being optical drop-outs. We will also discuss parameters like star formation rate and age estimates for these EROs and provide a comparison with other EROs samples, such as IRAC-selected EROs, sub-mm galaxies, and optically faint galaxies found in other searches for high redshift (z>4-6 galaxies).