Google Research has introduced the Planetary Prediction Engine (PPE), an autonomous AI system that dramatically accelerates the creation of planetary-scale geospatial prediction models. The experimental capability handles the entire modeling workflow autonomously—from discovering relevant data across global repositories to feature engineering, model training, and report generation—all triggered by natural-language queries. The system addresses a critical bottleneck in humanitarian response, where weeks of manual data curation and specialized expertise have historically delayed critical insights during crises. The PPE orchestrates this workflow through three modular stages coordinated by large language models. First, it intelligently translates natural-language queries into geographic constraints and discovers relevant signals from established repositories like Data Commons and Google Earth Engine, supplemented by live open-web searches of government and academic data sources. Second, it curates multimodal datasets by fusing geospatial foundation models with satellite imagery and demographic embeddings while implementing automated safeguards against data leakage. Third, it searches across multiple model families with custom overfitting prevention protocols to produce optimized predictions. Benchmark results demonstrate substantial improvements across diverse domains. On CDC health indicators, PPE achieved a mean R² of 76.8% compared to 60% for manually-curated expert pipelines. For food security prediction in data-scarce Nigeria, the system doubled accuracy when downscaling from regional to local government area levels. These gains position PPE as a potentially transformative tool for humanitarian organizations, environmental agencies, and public health officials who need rapid, high-fidelity geospatial intelligence during crises.