![JMA – Ninja & Deep Jandu [Official MV] Kaptaan](https://i.ytimg.com/vi/-dsfIh319s0/maxresdefault.jpg "JMA – Ninja & Deep Jandu [Official MV] Kaptaan")
2 min readFeb 19, 2026 01:04 AM IST
Senior White House Policy Advisor on artificial intelligence (AI) Sriram Krishnan Wednesday said that the US expects that its allies, including India, should build their AI solutions on top of America’s AI stack.
“We want to make sure the world uses the American AI stack. That means from the bottom up… it means our semiconductors and advanced AI chips, Nvidia, AMD, Google’s TPUs, and so on and so forth. Second, we want to make sure the world uses our AI models, all of the companies that I listed above. And third, it uses our applications and builds applications on top,” Krishnan said.
He was speaking at an event organised by the US-India Strategic Partnership Forum (USISPF) on the sidelines of the India-AI Impact summit.
Though India “should not give up strategic autonomy,” Krishnan said that the US ultimately wants “everyone” to build on top of the America AI stack, when asked about his take on India’s ambitions to build sovereign AI.
“India (as) a key ally, should be leveraging and building on top of this infrastructure. This does not mean giving up strategic autonomy. Indian companies will need to bring in local language support and local culture, and it needs to have local inference for low latency, built on infrastructure here. But at the end of the day, we want the American AI stack to be something that everyone builds on,”
Coincidentally, Krishnan’s comments came just as three Indian start-ups Sarvam, Bharatgen and Gnani.ai, launched sovereign large language and voice AI models on Wednesday, as domestic companies look to make a splash in the competitive AI ecosystem, searching for a ‘DeepSeek’ moment of their own. These models have been designed and trained entirely in India.
Sarvam launched two models: a 30-billion-parameter model and a 105-billion-parameter model, both trained from scratch in India. Gnani.ai unveiled a text-to-speech model, and Bharatgen launched a 17-billion-parameter multilingual foundational model.
© The Indian Express Pvt Ltd
