Sarvam is partnering with NVIDIA to architect India’s sovereign AI infrastructure by utilizing NVIDIA’s full hardware and software stack. By training its foundational models from scratch on NVIDIA H100 GPU clusters using the NVIDIA NeMo framework and Megatron-LM, Sarvam is natively processing over 2 trillion authentic Indian language tokens. Leveraging open technologies like the NVIDIA Nemotron toolchains has enabled Sarvam to scale securely within national trust boundaries—effectively creating a high-performance, cost-efficient "UPI for AI" that brings population-scale intelligence to all 1.4 billion Indian citizens.