India risks becoming a digital colony if it keeps relying on foreign artificial intelligence, the co-founder of Sarvam AI has warned in an interview with the Financial Times. Vivek Raghavan said the country needs its own models, its own computing capacity and control over its own data. Sarvam AI, the Bengaluru company he founded with Pratyush Kumar, is building foundation models and voice technology for Indian languages. Raghavan told the newspaper that a trillion-parameter model is in development and should be ready in six to nine months.
Sarvam AI and the digital colony warning
Raghavan put the warning in blunt terms. “The big question that we have to ask, in an Indian context, is whether there is a risk of the country becoming a digital colony,” he said. The Sarvam AI co-founder’s argument turns on economics. The largest technology companies in the world now give away powerful AI assistants for nothing, and he said that makes it hard for domestic firms to compete on equal terms. “In any other industry, if someone invested $100bn and gave it away for free in another country, that would be called dumping. This is actually what’s happening in AI,” he said.
Data is the second part of his case. India is one of the biggest markets for the free tiers of global chatbots such as ChatGPT and Google’s Gemini. The Sarvam AI co-founder warned that free users pay in another way. “If you’re a free user, your data is the product,” he said. For a country of India’s size, he argued, strategic autonomy means the ability to build and run AI systems at home rather than depend entirely on providers based overseas. The point goes beyond technology. It is about who decides how the systems behave and where the data they learn from is held.
What Sarvam AI is building now
Sarvam AI has already built a 105-billion-parameter model using government-supported computing resources. The trillion-parameter model now in development would be far larger, and Raghavan said it should be ready within six to nine months. The company’s voice agents handle roughly 2.5 million to 3 million calls a day, helping businesses talk to customers in regional languages. Its end-to-end voice system supports 12 Indian languages, and some of its models cover up to 22. The focus on language is deliberate: India has 22 official languages, and most global models were not built with them in mind.
The company was chosen by the government in April 2025 to build India’s sovereign large language model from scratch under the IndiaAI Mission, with dedicated computing resources set aside for the work. In June, Sarvam AI announced a $234 million first close of its $300 million Series B funding round, with HCLTech as the lead strategic investor. Raghavan said the goal is first to win a large share of AI use in India, then to take its affordable technology to markets in Africa, the Middle East and other parts of Asia, where language, cost and access raise similar problems.
A debate beyond one company
The Sarvam AI co-founder has made the same case before. In October, speaking to students at the Scindia School’s Founder’s Day event in Gwalior, he said India should build its own AI models, computing infrastructure and data capabilities, and warned that the United States and China were racing ahead while India risked being left behind. He told the students that technological sovereignty would shape India’s ability to build AI systems suited to its own needs, and summed up the aim as AI built by Indians, for India, and then for the world.
The Sarvam AI co-founder’s remarks land as India moves on several fronts to build domestic AI capacity. The IndiaAI Mission subsidises computing power for Indian developers, and new frameworks are being drawn up for sharing data safely so models can be trained on it. The debate Raghavan has sharpened is now one that policymakers and company technology chiefs are having in public: how much of the country’s AI stack, and the data that feeds it, should sit under Indian control. For businesses choosing between free global tools and homegrown alternatives, the question covers capability and price, and also where their data goes.
Raghavan ended the interview on an optimistic note. “My hope is that we are at the vanguard of change and there will be a hundred Sarvams in the future. I’m optimistic,” he said. Whether that future arrives will depend on funding, computing power and the talent to build models at a scale India has not attempted before.

