facebook icontwitter iconlinkedin iconvideo icon

Sekar Vembu Questions Government’s AI Champion Strategy After Sarvam Token Factory Launch

Sekar Vembu Questions Government’s AI Champion Strategy After Sarvam Token Factory Launch
Rosalin Biswal
SUMMARY
  • Sekar Vembu questioned the government’s approach of selecting individual companies as India’s AI champions.
  • His comments followed Sarvam AI’s launch of Token Factory, an India-hosted inference platform offering access to domestic and international AI models.
  • Vembu argued that GPUs and computing infrastructure should be accessible to startups, universities and independent researchers.

Tech entrepreneur Sekar Vembu has questioned the Indian government’s approach to supporting domestic artificial intelligence (AI) companies, arguing that public resources should be made accessible to startups, academic institutions and independent researchers rather than being concentrated among companies selected as national AI champions.

His comments follow the launch of Sarvam AI’s Token Factory, an inference platform that allows developers and enterprises to access the company’s own AI models alongside open-weight models developed by international companies through computing infrastructure hosted in India.

In a post on X, Vembu said he did not blame Sarvam for offering Chinese-developed models alongside its own products, arguing that startups must respond to customer requirements to remain commercially viable. However, he questioned the government’s decision to select individual companies to lead India’s sovereign AI development efforts and provide them with subsidised computing resources.

“A startup has to survive on its customers, so it has to give them what they want,” Vembu wrote, adding that customers should be able to choose models based on their requirements, regardless of where they originate.

He argued that government-backed selection of AI champions could discourage other researchers and startups that might otherwise compete to develop indigenous AI models.

Sekar Vembu Calls for Market-Led AI Development

According to Vembu, the government should focus on making computing infrastructure more accessible rather than deciding which companies should lead India’s AI development efforts.

He suggested reducing import duties and other barriers associated with acquiring computing resources, particularly graphics processing units (GPUs), which are essential for training and running large AI models.

Vembu also called for wider access to computing infrastructure for startups, universities, research laboratories and individual researchers.

“Great AI models from India will come from hundreds of motivated people competing in the marketplace, not from one company chosen by the government,” he said.

His argument centres on the role of competition in developing India’s AI capabilities. Instead of concentrating resources among a limited number of selected companies, he believes a broader ecosystem of researchers and technology startups should have the opportunity to develop competing models.

The comments also highlight a distinction between commercial AI infrastructure and indigenous foundation model development. Hosting international AI models on Indian servers can support data residency and local inference requirements, but it does not mean those models were researched or developed in India.

Sarvam’s Token Factory Reopens India’s Sovereign AI Debate

Sarvam AI launched Token Factory on November 16, 2026, offering access to eight models through a single application programming interface (API). The catalogue includes its own Sarvam 105B model, alongside models from Chinese AI developers DeepSeek, Moonshot AI and Z.ai, as well as Google’s Gemma.

The platform runs these models on GPU infrastructure located in India. Sarvam said the service is designed to give enterprises access to multiple models while offering control over their AI infrastructure without dependence on a single model provider.

The launch triggered discussion over whether India’s sovereign AI ambitions should primarily focus on developing indigenous foundation models or building domestic infrastructure capable of running the world’s leading AI systems.

Sarvam was selected on April 26, 2025, under the IndiaAI Mission’s Foundation Model initiative to develop an indigenous large language model ecosystem. The government initially selected four startups, including Sarvam AI, Soket AI, Gnani AI and Gan AI, to develop Indian foundation models.

The IndiaAI Mission has an overall approved outlay of ₹10,372 crore. A February 2026 government disclosure listed ₹246.72 crore in compute support for Sarvam under the Foundation Models pillar. The mission also provides subsidised access to shared computing infrastructure for startups and academic institutions.

The government’s programme has since expanded to include 12 selected organisations and consortia working on indigenous foundation models.

Find this useful? Pass it on.
CEO VINE covers India's startups, founders and funding news every day.
Add CEO VINE as a preferred source on GoogleGet more of our startup and business stories in your Google Top Stories.

You might also like