In recent years cosmology has undergone a revolution, with precise measurements of the microwave background radiation, large galaxy redshift surveys, and the discovery of the recent accelerated expansion of the Universe using observations of distant supernovae. All these groundbreaking observations have boosted our understanding of the Cosmos and its evolution. Because of this detailed understanding, more detailed tests of cosmological models require unprecedented precision that is only available with the next generation of astronomical observatories. Radio observations in particular will be able to access more independent modes than optical, infrared or X-ray facility and will show very different systematics compared to these other wavebands. The SKA enables us to do an ultimate test in cosmology by measuring the expansion rate of the Universe in real time. This can be done by a rather simple experiment of observing the neutral hydrogen (HI) signal of galaxies at two different epochs. The signal will encounter a change in frequency imprinted as the Universe expands over time and thus monitoring the drift in frequencies will provide a real time measure of the cosmic acceleration. Over a period of 12 years one would expected a frequency shift of the order of 0.1 Hz assuming a standard LCDM cosmology. However, monitoring such changes would require some modifications to the current baseline design of the SKA. In particular, the design needs to be adapted to achieve higher spectral resolution, at least within sub-bands (strong requirement), and to allow for a well monitored distribution of the local oscillator signal, preferably at milli-Hz accuracy over a period of 12 years (weaker requirement, which could be circumvented by pulsar observations). Based on the sensitivity estimates of the SKA and the number counts of the expected HI galaxies, it is shown that the number counts are sufficiently high to compensate for the observational uncertainties of the measurements and hence allow a statistical detection of the frequency shift. In addition, depending on the observational setup, it is shown that the evolution of the frequency shift in redshift space can be estimated to a precision of a percent. Although technically challenging, the direct measurement of the frequency shift and hence the cosmic acceleration can provide a model independent confirmation of dark energy. At highest precision it can distinguish between some competing cosmological models and combined with probes at other wavelength can break degeneracies and improving the figure of merit of cosmological parameters.