We use the final catalogue of the VIMOS Public Extragalactic Redshift Survey (VIPERS) to measure the power spectrum of the galaxy distribution at high redshift, presenting results that extend beyond $z=1$ for the first time. We apply an FFT technique to four independent sub-volumes comprising a total of $51,728$ galaxies at $0.6<z<1.1$ (out of the nearly $90,000$ included in the whole survey). We concentrate here on the shape of the direction-averaged power spectrum in redshift space, explaining the level of modelling of redshift-space anisotropies and the anisotropic survey window function that are needed to deduce this in a robust fashion. We then use covariance matrices derived from a large ensemble of mock datasets in order to fit the spectral data. The results are well matched by a standard $\Lambda$CDM model, with density parameter $\Omega_M h =\smash{0.227^{+0.063}_{-0.050}}$ and baryon fraction $\smash{f_B=\Omega_B/\Omega_M=0.220^{+0.058}_{-0.072}}$. These inferences from the high-$z$ galaxy distribution are consistent with results from local galaxy surveys, and also with the Cosmic Microwave Background. Thus the $\Lambda$CDM model gives a good match to cosmic structure at all redshifts so far accessible to observational study.