MUMBAI: Tech giants such as SoftBank’s shrinking appetite for India investments, more of smaller startup cheques being written by venture capital (VC) funds or the lack of euphoria in domestic stock markets which has now been overtaken by Taiwan in terms of value for that matter—all are somehow reflective of the fact that India is far behind in the AI race.At a time when two US AI startups Anthropic and OpenAI are heading towards public listings estimated to be worth several billion dollars, India is barely scratching the surface. For the first time since at least 2000, local companies are understood to be out of the top 10 constituents of the MSCI Emerging Markets (EM) Index. Consider this: TSMC alone accounts for 42% of Taiwan’s benchmark index–thanks to its advanced chips which power the world’s top AI models.Not to say that Indian founders are not building high quality AI startups but lagging compute capacity, underdeveloped research ecosystem and slow domestic consumption are the reasons why India is trailing in AI, said investors. The recent move by the US govt to restrict foreign access to Anthropic’s Fable 5 and Mythos 5 models only emphasises the need for India to step up its own AI capabilities, they said.
Restricted Access To Advanced Models Shows The Need For India To Step Up
“While initiatives such as Sarvam AI’s Sarvam-30B and Sarvam-105B represent important progress toward sovereign AI capabilities, India continues to rely largely on models developed and controlled by foreign companies,” said Anup Jain, founding partner at BlueGreen Ventures.
$40-50mn revenue elusive
Indian AI startups raised nearly $1.5 billion in the March quarter, about 38% of all startup funding, said Ashish Bhatia, founder & CEO at India Accelerator, arguing that when one in three rupees a VC writes is going to AI, the ecosystem isn’t slow.This, however, pales in comparison to leading markets like the US and China where the flow of investments is way higher. Anthropic alone raised $65 billion in recent funding at a valuation of $965 billion.“The depth of the market in India today is not yet enough to support massive billion dollar funding rounds and trillion dollar valuations like in the US,” said Pankaj Mitra, partner at Bessemer Venture Partners India, adding that the country needs deeper AI talent in fundamental research.India lacks a competitive position at the foundation model layer. Besides, the country does not yet have an AI-first company generating $40-50 million in annual revenue, a milestone that serves as the trigger for large late-stage capital.
Sovereign compute key
Big tech giants Meta and Google are sure building AI data centres in India but it is not enough and more domestic players need to contribute. “We cannot build serious AI without GPU access at scale,” said Poorvi Vijay, principal at Elevation Capital. India needs to have sovereign compute and a real say over intelligence built on Indian data, not just where the data sits, said an investor, adding that the policy focus so far has only been data localisation. “Money worldwide is flowing into global AI giants. The market perception is that India is not doing enough in terms of building LLMs or in areas like semi-conductor manufacturing,,” said Nikunj Doshi, managing partner at Bay Capital.
Innovation across AI stack
So what are Indian AI founders building? Innovation is happening across the stack, say investors, although more founders are building a lot on the middle-ware layer—the plumbing between the foundation models and the enterprise deployment. “India doesn’t need to win the foundation model race….that ship has sailed.“What India needs to win is the deployment race: taking AI into agriculture, healthcare, climate and financial services at scale that no country can match,” said Jain.
