NVIDIA and IIT Madras' AI4Bharat lab have released Nemotron-pretraining-Indic-v1, a comprehensive dataset of 2.8 trillion tokens across 22 Indian constitutional languages alongside an open-source GPU-accelerated data generation pipeline. The 22 languages represent the first languages of approximately 97% of Indians, covering roughly 1.2 billion people according to the 2011 Census. The release represents a major expansion in openly available synthetic training data for Indic languages, addressing a critical gap identified by researchers as the primary obstacle to advancing Indian language AI technology. The corpus includes native-script data for all 22 languages and romanized text for twelve languages, spanning 34 total language-script configurations. The training material combines curated English content—including mathematics, code, and question-answer material—systematically translated into Indian languages through a multi-stage workflow featuring translation, quality assessment, and targeted repair. Each output maintains complete provenance tracing back to its original English source document, ensuring accuracy and enabling auditable, reproducible development. By releasing the open data generation pipeline alongside the dataset, NVIDIA and AI4Bharat enable other developers to reproduce the work and generate additional training data from their own source materials. The infrastructure has already been deployed to continue-pretrain NVIDIA's Nemotron 3.5 Super VL model with 10 billion Indic tokens, demonstrating immediate value for building stronger multilingual AI systems that can serve Indian language speakers with technical capabilities.