New 'Made in India' Foundational AI Models Launch

- Sarvam AI has raised a significant $41 million in a Series A round led by Lightspeed Venture Partners, with participation from Peak XV Partners and Khosla Ventures, marking one of the largest early-stage fundraises for an Indian AI startup. - The new models from Sarvam AI and BharatGen utilize a "Mixture-of-Experts" (MoE) architecture, which enhances efficiency by only activating a fraction of the model's total parameters for any given task. For instance, Sarvam-30B activates just 1 billion of its 30 billion parameters per token. - This push for homegrown AI is heavily backed by the Indian government's IndiaAI Mission, which has awarded BharatGen, a consortium led by IIT Bombay, ₹988.6 crore (around $112 million) to develop sovereign foundational models. The overall mission has an outlay of over $1 billion to build out India's AI compute capacity and support startups. - Gnani.ai's Vachana TTS model can perform "zero-shot" voice cloning, replicating a person's voice in 12 different Indic languages from less than 10 seconds of sample audio while preserving the speaker's pitch and style. - The development of these models is supported by a significant build-out of domestic infrastructure, with tens of thousands of NVIDIA GPUs being deployed through cloud providers like Yotta to power large-scale AI training. - Investment in India's AI sector has surged, with its share of total venture capital funding rising from under 5% in 2020 to approximately 12% in 2025, with AI-focused companies attracting $1.2 billion in 2025 alone. - These foundational models are already being tested in practical applications across various sectors, including governance (MahaGPT with the Government of Maharashtra), healthcare (the Medsum app), and education (an AI-driven spoken English assessment tool).