Crop water productivity is a key element of water and food security in the world and can be quantified by the water footprint (WF). Previous studies have looked at the spatially explicit distribution of crop WFs but little is known about the temporal dynamics. We develop a new global gridded crop model – AquaCrop-Earth@lternatives (ACEA) – that can simulate three consumptive WF components: green (WFg), blue from irrigation (WFbi), and blue from capillary rise (WFbc) at high temporal and spatial resolutions. The model is applied to analyse global maize production during 1986–2016 at 5 × 5 arc minute grid. Our results show that in 2012–2016, the global average unit WF of maize is 723.2 m3 t−1 y−1 (89.5 % WFg, 8.3 % WFbi, 2.2 % WFbc) with values varying greatly around the world. Regions characterised by high agricultural development generally show a small unit WF and its interannual variation, such as Western Europe and Northern America (WF < 500 m3 t−1 y−1, CV < 15 %). On the contrary, regions with low agricultural development show opposite outcomes, such as Middle and Eastern Africa (WF > 2500 m3 t−1 y−1, CV > 40 %). Since 1986, the global unit WF of maize has reduced by 34.6 % mainly due to the historical decrease in yield gaps. However, due to the rapid expansion of rainfed and irrigated cropland, the global WF of maize production has increased by 48.8 % peaking at 762.9 × 109 m3 y−1 in 2016. As many regions still have a high potential in decreasing yield gaps, the unit WF of maize is likely to continue reducing, whereas the WF of maize production is likely to continue growing as humanity’s rising appetite can lead to further cropland expansion. The simulation of other crops with ACEA is necessary to assess the pressure of overall crop production on ecosystems and freshwater resources worldwide.