Inside India newsletter: Anthropic curbs ignite AI debate in India

Hello, this is Priyanka Salve, writing to you from Singapore.

Welcome to the latest edition of  “Inside India“ — your one-stop destination for stories and developments from the world’s fastest-growing large economy.

While the U.S. and China have been racing to develop a sovereign artificial intelligence stack, India was confident of making its mark by building the AI application layer on top of foreign foundational models. But now New Delhi is being pushed to rethink its strategy as Washington moves to restrict certain AI offerings.

Any thoughts on today’s newsletter? Share them with the team.

India’s artificial intelligence strategy was simple: become an AI innovation hub by leveraging its vast information technology talent to develop applications using foreign foundational models.

But the fragility of that ambition was exposed last week when Anthropic disabled access to its new models— Fable 5 and Mythos 5 — for foreign nationals, complying with an export-control directive from the U.S. government.

“The fact that frontier access can vanish overnight on a foreign government’s order is the whole problem,” Saket Dandotia, co-founder and chief executive at Onetab.ai, told CNBC.

Anthropic’s suspension of access could have broken Dandotia’s business of making AI applications for enterprises had he not diversified across multiple models.

But “diversification buys time; it doesn’t buy independence,” Dandotia said, adding that India needs sovereign AI, so that startups like his do not “lose their edge” with one directive from a foreign government.

An ADP Research report released Thursday found that 41% of Indian workers use AI nearly every day, higher than 26% in China and 19% in the U.S. But without a sovereign AI stack, that adoption reflects the extent of India’s reliance on foreign tech.

India does not yet produce cutting-edge chips domestically, nor does it yet have a frontier-scale foundation model on a par with leading U.S. or Chinese models. While growing fast, its data center capacity also lags considerably behind the U.S. and China.

Government efforts are underway on all three fronts, through an India semiconductor mission, an AI mission, and tax breaks for global hyperscalers setting up data centers in the country.

The private sector is also beginning to realize the need to invest in the domestic AI stack. On Monday, India’s Sarvam AI, which is building sovereign AI models, raised $300 million at a $1.5 billion valuation from a clutch of investors that included India’s third-largest software services company by market cap, HCL Technologies.

But the efforts might be too little, and likely too late, industry experts said, adding that the biggest challenge for India is access to computing power and a lack of deep-tech investment capital.

India’s strength is its strong domestic market, both at the consumer and enterprise front, and its deep tech talent pool, but it lacks the capital that is available to sovereign AI companies in the U.S. and China, said Manish Agarwal, co-founder of Humyn Labs, a physical AI data company.

Last year, Indian startups raised $10.5 billion in funding, the third highest in the world after the U.S. and U.K., according to a report from private market intelligence firm Tracxn in December. However, most of these funds went to startups in enterprise applications, retail, and fintech sectors, and not deep-tech companies — firms working on cutting-edge disruptive technologies.

Investments by venture capitalists in India in deep-tech startups are quite small compared with the billions being invested overseas, experts told CNBC, adding that private investors are more conservative when betting on deep-tech ideas.

For example, HCL Tech’s investment of 14.27 billion rupees ($151 million) in Sarvam was less than 10% of the what it paid out to shareholders as dividends in the financial year ending March 2026.

So, calls are growing within the country, asking the government to invest more heavily in developing sovereign AI.

Mohandas Pai, a prominent venture capitalist and angel investor, has urged Prime Minister Narendra Modi to start an AI mission, calling existing government programs “too slow, way too small to make any large impact.”

For a country the size of India, building a foundational AI model that does not hallucinate will need to have a few trillion parameters, experts said, adding that this requires a lot of capital and computing power. The flagship model of India’s Sarvam, for example, has a little over 100 billion parameters.

Currently, sovereign AI models being developed in India are using Nvidia architecture, but if the U.S. restricts access to the Blackwell chips as it did with China, it will leave India helpless, Neil Shah, vice president of research at Counterpoint Research, told CNBC.

This is a risk that other influential voices in the Indian tech industry are also expressing.

Sridhar Vembu, co-founder of Indian tech multinational Zoho and who founded Arattai to rival WhatsApp in India, said in a post on X that “Technology is the ultimate weapon,” and that India must find its “own way ahead.”

But to do that, the country needs capital and sovereign computing power, both of which it severely lacks.

Without a strong government drive to address those issues, the conversation around building sovereign AI is at risk of “ephemerality,” Agarwal of Humyn Labs said.