We present cosmological constraints from the analysis of angular power spectra of cosmic shear maps based on data from the first three years of observations by the Dark Energy Survey (DES Y3). Our measurements are based on the pseudo-Ct method and complement the analysis of the two-point correlation functions in real space, as the two estimators are known to compress and select Gaussian information in different ways, due to scale cuts. They may also be differently affected by systematic effects and theoretical uncertainties, making this analysis an important cross-check. Using the same fiducial Lambda cold dark matter model as in the DES Y3 real-space analysis, we find S-8 = sigma(s)( root)Omega m/0.3 = 0.7931(-0.025)(+0.038), which further improves to S-8 = 0.784 +/- 0.026 when including shear ratios. This result is within expected statistical fluctuations from the real-space constraint, and in agreement with DES Y3 analyses of non-Gaussian statistics, but favours a slightly higher value of S-8, which reduces the tension with the Planck 2018 constraints from 2.3 sigma in the real space analysis to 1.5 sigma here. We explore less conservative intrinsic alignments models than the one adopted in our fiducial analysis, finding no clear preference for a more complex model. We also include small scales, using an increased Fourier mode cut-off up to k(max) = 5 h Mpc(-1), which allows to constrain baryonic feedback while leaving cosmological constraints essentially unchanged. Finally, we present an approximate reconstruction of the linear matter power spectrum at present time, found to be about 20 per cent lower than predicted by Planck 2018, as reflected by the lower S-8 value.
