Understanding the subnational dynamics of food demand, while accounting for the evolution of the agrifood system at the global level, remains a challenge for designing food policy to ensure food security. To link global agrifood systems with subnational food policy planning, we downscale per capita food demand for 62 food commodities at the subnational level across all world regions over the next 5 years, based on downscaled projections from a global economic model. The novelty of our approach lies in being the first to downscale global food demand projections to the subnational level, thereby enabling the assessment of within-country heterogeneity in future food demand while explicitly linking it to global drivers and the broader agrifood system. The results show considerable within-country heterogeneity: subnational changes in food demand often move in the opposite direction of national projections. Using Kenya as a case study, we reveal nuanced subnational demand patterns and discuss them in the context of Kenya's Food Security Policy Framework. The downscaled estimates confirm that national averages can be misleading—while national per capita maize demand is projected to decline, several counties may experience increases, and national gains in dietary diversity are largely driven by already medium–high diversity areas. Different global scenarios result in heterogeneous subnational dietary patterns, showing how global drivers affect local demand. For sparsely populated areas, model predictions are less reliable and should be interpreted with caution. Overall, our results highlight the importance of accounting for subnational heterogeneity when designing forward-looking food policies to ensure food security at the subnational level.