Food is a fundamental right that deserves attention but is usually dealt with from the supply side in aggregated models that use macroeconomic variables to forecast the demand and the required supply. This study challenges this paradigm by developing a simulator to analyze food consumption from the demand side and estimate the evolution of disparity in food consumption over time with respect to region, sex, ethnicity, education, and income. This novel approach was applied to Brazil using household expenditure surveys to feed serial neural networks. Results show that the ‘poorer’ north and northeast of Brazil encounter the lowest consumption of food and are therefore the most food vulnerable regions. This trend continues to 2040. The ‘richer’ south and southeast regions have higher food consumption, which varies according to sex, ethnicity, education, and income. Brazil has contrasting issues with some groups having considerably higher food consumption, while other groups still have less than the threshold for healthy consumption. Now, the country not only has to deal with the food access by the most vulnerable due to the latest economic declines but also to deal with excess consumption, the so-called “double burden of malnutrition”.