Sarvam AI has raised $234 million, crossed a $1.5 billion valuation, and become India's newest AI unicorn. The lead check, $150 million, came from HCLTech, a large Indian IT company that typically sells services to global clients rather than building its own technology. That is the interesting part.
HCLTech buying a 10.46% stake in Sarvam is an unusual move for an IT services company. The old model for firms like HCLTech is to supply engineers and support to foreign companies. That model is under serious pressure as AI automates large portions of exactly that kind of work. Investing in Sarvam is a hedge: rather than only selling labor, HCLTech is buying into a technology it can bring to its own enterprise clients.
The combination makes sense on paper. Sarvam has built AI models that work in Indian languages and understand Indian-specific documents, voice patterns, and government workflows. HCLTech has relationships with large enterprises and governments globally. Put those two together and you have a product company, not just a research lab.
The practical proof is already there. Sarvam's conversational AI platform handles over 2 million interactions a day. Its voice agents collected agricultural data from 17 million Indian farmers for the Ministry of Agriculture. An insurance campaign running on its platform supported policy renewals for 45 million policyholders. A large Indian fintech company is using Sarvam's platform to support a sales team of 350,000 people. These are not prototype numbers.
The timing of the funding also happened to coincide with a sharp reminder of why countries want their own AI. Last Friday, the US government ordered Anthropic to shut off access to its two most advanced AI models for every foreign national, anywhere in the world, without warning. Companies and teams outside the US that relied on those tools lost access immediately. The order applied even to Anthropic's own non-US employees.
This is the first time the US government has used export controls not just on AI hardware, but on the AI models themselves. It will not be the last. Governments increasingly treat advanced AI the same way they treat fighter jets: strategic assets that can be switched off for foreign users at any moment.
Sarvam has been positioning itself around this risk since the beginning. The Indian government selected Sarvam from 67 proposals to build India's official sovereign AI model, and backed it with subsidized access to 4,096 high-end Nvidia GPUs through the government's IndiaAI Mission program. That infrastructure subsidy was worth roughly $30 million at market rates, giving Sarvam a cost advantage that a private startup building from scratch simply cannot replicate.
The question for any business operator using AI in markets outside the US is straightforward: what happens to your operations if the tools you depend on get switched off? Sarvam's pitch is that Indian-built AI, running on Indian infrastructure, cannot be turned off by a foreign government. That argument now has a very concrete recent example behind it.
Sarvam still has real challenges. Its revenue for the last reported financial year was modest relative to its $1.5 billion valuation, and the HCLTech deal values it at roughly seven times its previous funding round's implied price. Whether the commercial scale matches the infrastructure ambition remains to be seen. But the underlying problem it is solving, AI that works for non-English speakers on infrastructure that cannot be remotely revoked, is real and getting more urgent.