We use nearly 20 years of photometry obtained by the OGLE survey to measure the occurrence rate of wide-orbit (or ice giant) microlensing planets, i.e., with separations from ~5 AU to ~15 AU and mass-ratios from $10^{-4}$ to 0.033. In a sample of 3112 events we find six previously known wide-orbit planets and a new microlensing planet or brown dwarf OGLE-2017-BLG-0114Lb, for which close and wide orbits are possible and close orbit is preferred. We run extensive simulations of the planet detection efficiency, robustly taking into account the finite-source effects. We find that the extrapolation of the previously measured rate of microlensing planets significantly underpredicts the number of wide-orbit planets. On average, every microlensing star hosts $1.4^{+0.9}_{-0.6}$ ice giant planets.