Introduction: Malaria, childhood acute respiratory infection, and child undernutrition1 are leading causes of preventable mortality in children under five, concentrated in low and middleincome countries where climate variability directly modulates transmission, exposure, and nutritional outcomes [1–3].. Routine health surveillance in these settings remains sparse, and the utility of satellite-derived representations of the Earth’s surface as predictors of population health outcomes is poorly characterised. Methods: We evaluate AlphaEarth Foundations 64-dimensional satellite embeddings as predictors of population health outcomes using LSTM and Transformer ensembles across three pathologies: malaria, acute respiratory infection, and stunting. Results: Embeddings provide meaningful predictive value when merged at sufficient spatial granularity: (i) in malaria case prediction, they raise test R2 from 0.623 to 0.777 in Nigeria and from 0.867 to 0.881 in India; (ii) in childhood ARI prediction across 11 DHS countries, pooled R2 rises from 0.098 to 0.164, with XGBoost reaching R2 = 0.210; (iii) in child weightfor-height z-score prediction across 35 DHS countries, gains are within seed variance, consistent with country fixed effects already absorbing the between-country variance the static embedding can express. Interpretation and conclusion: Together, these data show that AlphaEarth satellite embeddings consistently add a predictive value when applied at a location or cluster level across distinct population health outcomes. We close with a request: direct access to the Google Earth AI foundation-model suite, including Population Dynamics embeddings that incorporate health indicators, would substantially accelerate this line of work across all three outcomes.