Sarvam AI: India's sovereign LLM breakthrough comes with Nokia & Bosch partnerships

Sarvam, an Indian AI startup, has joined the race for LLMs without fanfare. The company's accomplishment is remarkable for not being dependent on external inputs for data, algorithms or engineering. Equally remarkable is the frugality with which it has arrived at its product milestone, which is a pointer to what India brings to the table in terms of AI development. The country is key for dispersal of GenAI irrespective of where it originates, and indigenous efforts can be expected to be extra sensitive to the local market contours. Dispersal rates will climb as the pipeline of sovereign AI offerings broadens. Early successes should not fly under the radar on account of limited resources or ambition.

Sarvam has announced partnerships with Nokia and Bosch for AI applications. The startup's focus remains on enterprise initiatives, but it has launched a chat app for web and mobile users, entering territory carved out by global giants Google, OpenAI and Anthropic. The big boys take the Indian market seriously for its size and savvy.

Finding an Indian name in this crowd, however circumspect its vision, is an encouraging development. It also speaks to public-funded initiatives that are seeding the domestic AI ecosystem despite tall hurdles in energy and compute. Unlike the US or China, where big tech is stumping up the cash for AI investments, India is proceeding on the tested template of public infrastructure mated with private innovation. The model has succeeded beyond expectations and is key to expanding India's sovereign AI footprint.


Last week's global AI summit in New Delhi may have made the headlines for the wrong reasons, yet it does not take away from India's desire to assert its intent. Those efforts are in right earnest beyond the media glare, as Sarvam has demonstrated. An ecosystem for Indian AI innovation is likely to creep up on us through the entrepreneurial zeal of the country's formidable talent base. Local solutions for global problems are the way to go.