Context. Dust formation in stellar outflows is initiated by the formation of some seed particles that form the growth centres for macroscopic dust grains. The nature of the seed particles for silicate dust in stellar outflows with an oxygen-rich element mixture is still an open question. Clustering of the abundant SiO molecules has been discussed several times as a possible mechanism and investigated both theoretically and by laboratory experiments. The initial results seemed to indicate, however, that condensation temperatures obtained by model calculations based on this mechanism are significant lower than what is really observed, which renders SiO nucleation unlikely. Aims. This negative result strongly rests on experimental data on the vapour pressure of SiO. The case for SiO nucleation may be not as bad as it previously seemed and needs to be discussed again because new determinations of the vapour pressure of SiO molecules over solid SiO have shown the older data on SiO vapour pressure to be seriously in error. Here we aim to check again the possibility that SiO nucleation triggers the cosmic silicate dust formation in light of improved new data. Methods. First we present results of our measurements of vapour pressure of solid SiO. Second, we use the improved vapour pressure data to recalibrate existing experimental data on SiO nucleation from the literature. Third, we use the recalibrated data on SiO nucleation in a simple model program for dust-driven winds to determine the condensation temperature of silicate in stellar outflows from AGB stars. Results. Our measurements extend the temperature range of measurements for the vapour pressure to lower temperatures and pressures than ever before. This improves the reliability of the required extrapolation from the temperature range where laboratory data can be obtained to the temperature range where circumstellar dust condensation is observed. We determine an analytical fit for the nucleation rate of SiO from recalibrated literature data and show that the onset of nucleation under circumstellar conditions commences at a higher temperature than was previously found. This brings calculated condensation temperatures of silicate dust much closer to the observed condensation temperatures derived from analysis of infrared spectra from dust-enshrouded M stars. Calculated condensation temperatures are still by about 100 K lower than observed ones, but this may be due to the greenhouse effect of silicate dust temperatures, which is not considered in our model calculation. Conclusions. The assumption that the onset of dust formation in late-type stars with oxygen-rich element mixtures is triggered by the cluster formation of SiO is compatible with dust condensation temperatures derived from infrared observations.