Sarvam’s $10-billion AI gamble faces the frontier test

The $10 Billion Question

Sarvam AI, India’s newest AI unicorn, has set its sights on building a foundation model that can rival the latest offerings from OpenAI and Anthropic. The company has raised around $350 million cumulatively, including an extended $75 million Series B round from Nvidia in August. Co-founder and CEO Pratyush Kumar believes that a $10 billion investment today could take Sarvam to the current AI frontier by Independence Day next year.

But experts say the harder question is whether Sarvam can remain at the frontier as the benchmark keeps moving.

“The gap is actually not as big as it looks. If we were to invest $10 billion today, we could reach the frontier by Independence Day next year,” Kumar said earlier this month. Such an investment would give Sarvam access to large-scale compute, advanced chips, infrastructure, and the ability to hire talent. However, catching up with models from OpenAI, Anthropic, and Google involves more than assembling these resources.

Talent and Research Depth Are Key

“The challenge is therefore not merely capital, but research depth, compute availability, training efficiency and ability to attract a very small pool of world-class researchers,” said Jaspreet Bindra, co-founder and CEO of AI&Beyond. Sarvam could plausibly get close to today’s frontier within a year, he said, but matching the frontier of that time would be considerably harder because global AI labs would have moved ahead.

Pramod Gosavi, senior principal at Bay Area-based VC firm Blumberg Capital, said Sarvam would struggle to build a credible frontier model without experienced AI researchers leading the training effort. He pointed to the research depth at OpenAI, Anthropic, and Google’s DeepMind.

Pratyush Choudhury, co-founder of Activate, said compute and data are bigger limitations than talent. Data, he noted, includes usage data for fine-tuning models.

Devendra Chaplot, a founding member of Mistral AI and Thinking Machines Lab, has joined as an adviser for Sarvam’s planned trillion-parameter frontier model. Rishi Bal, CEO of BharatGen, said the narrative has shifted to whether India can compete at the global frontier.

India’s AI Momentum

Beyond Sarvam, India is making strides in AI. Rajan Anandan, a prominent investor, highlighted that this is the first year several Indian ground-up models have been launched. Sarvam has launched models trained ground up in India, and BharatGen from IIT Bombay is another example.

Anandan pointed to India’s digital payments transformation: In 2016, India wasn’t in the top 100 countries for digital payments; today, 55 to 60 per cent of real-time payments are in India. He suggested a similar trajectory could play out in AI, with the verdict on India’s AI game coming eight to 10 years from now.

Sarvam models are already being deployed. Kyndryl India CTO Sreekrishnan Venkateswaran said Sarvam and BharatGen models are used by clients in banking, financial services, insurance, and telecom. Sarvam works across 22 Indian languages and is expanding enterprise deployments.

Cost Advantages and Challenges

Sarvam models are five times cheaper than GPT mini and nine times cheaper than Gemini Flash, according to Anandan. He also noted that Chinese models are significantly cheaper than frontier labs and were built with much fewer GPUs than the US. Anandan said models smaller than 100 billion parameters can handle 90-95% of what companies need with AI.

India has a renewable energy advantage for data centres, but its chip fabs are 28-nanometre, not advanced AI chips. The India AI Mission’s Rs 10,000-crore fund is a step in the right direction, Anandan said.

The Investment View

Anandan said India has had over 100 IPOs this year, but cautioned that “tough to justify valuations” is the biggest risk in AI investing. He noted that “we are certainly not in a cheap AI market.”

He illustrated the valuation challenge: “See, $1.5 billion to $100 billion… $1.5 billion was the last round. That's about 60x. Anthropic would have to be a $60 trillion company.”

Anandan recommended investing in Sarvam and companies at the application or model layer. He regretted not investing in Anthropic. He also mentioned the Equal AI app for answering phone calls.

The Bigger Picture

Kumar said building AI no longer requires a “magic source” exclusive to a few companies. He dismissed the AGI singularity claim as “completely bogus.” He praised Anthropic’s bet on coding as amazing, called DeepSeek a “great systems company,” and said open-source models match proprietary options. He also said Google is full-stack and he likes Google.

Asked about prominent figures, Kumar said Elon Musk is “biomodal,” while of Sam Altman he said, “Don't know enough,” and of Donald Trump, “Definitely don't know at all.”

On India’s future, Anandan said, “India is a wide-open playing field. That's why it's an exciting market, because we believe that if you win in India, you win.” He added, “The risk really is that we don’t move fast enough as a nation… that we don’t aim to be in the winner’s circle. This is the industrial revolution, repeat.”

Kumar echoed that optimism: “We should all aim towards a world of abundance and positive outcomes.”

A Decade of Uncertainty

Sarvam’s biggest model is still months away, leaving the outcome uncertain. The challenge is not just reaching the frontier but staying there. As Bindra noted, global AI labs will not stand still. The next few years will test whether Indian AI companies can turn capital and ambition into lasting leadership.