Observations of high-redshift supernovae (SNe) open a novel opportunity to study the massive star population in the early Universe. We study the detectability of superluminous SNe with upcoming optical and near-infrared (NIR) surveys. Our calculations are based on the cosmic star-formation history, the SN occurrence rate, the characteristic colour and the light curve of the SNe, which are all calibrated using available observations. We show that 15–150 SNe up to z∼ 4 will be discovered by the proposed Subaru/Hyper Suprime-Cam deep survey, a 30-deg2 survey with 24.5 AB mag depth in the z band for 3 months. With its ultradeep layer (3.5 deg2 with 25.6 AB mag depth in the z band for 4 months), the highest redshift can be extended to z∼ 5. We further explore the detectability by upcoming NIR surveys utilizing future satellites such as Euclid, WFIRST and WISH. The wide-field NIR surveys are very efficient at detecting high-redshift SNe. With a hypothetical deep NIR survey for 100 deg2 with 26 AB mag depth, at least ∼50 SNe will be discovered at z > 3 in half a year. The number of detected SNe can place a strong constraint on the stellar initial mass function or its slope, especially at the high-mass end. Superluminous SNe at high redshifts can be distinguished from other types of SNe by the long time-scale of their light curves in the observer’s frame, optical colours redder than other core-collapse SNe and NIR colours redder than any other type of SNe.