Sarvam and Krutrim's fresh rounds show investors still believe in an India-first AI stack

The latest funding rounds closed by Sarvam AI and Krutrim - the two most closely watched Indian foundation-model startups, each pursuing a somewhat different strategy for building AI capability tailored to Indian languages and use cases - suggest that domestic and international investors remain willing to fund India-specific AI infrastructure even as the astronomical valuations and funding scale at the American and Chinese frontier labs make any India-based effort look small by comparison. Both companies have leaned into partnerships with the government's IndiaAI Mission compute allocation while also raising private capital to fund the additional compute and talent that government subsidy alone does not cover. The strategic bet underlying both companies differs in emphasis. Krutrim, founded by Ola's Bhavish Aggarwal, has pursued a broader full-stack ambition spanning foundation models, an AI-focused cloud and even hardware ambitions, betting that vertical integration across the AI stack gives it defensibility that a pure model-layer play would lack. Sarvam has focused more narrowly on enterprise and government-deployment-ready Indic-language models, prioritising the kind of data-residency, security and multilingual accuracy features that Indian government and BFSI-sector customers specifically require and that foreign frontier labs have been slower to prioritise for the Indian market. Both companies face the same structural challenge: building genuinely frontier-competitive foundation models requires compute and top research talent at a scale that is difficult to fund purely from India-sized venture rounds, and both have had to make hard choices about which capability frontiers to pursue aggressively - agentic reasoning, multimodal understanding, raw parameter scale - versus which to deprioritise in favour of the Indic-language and deployment-context advantages that constitute their actual differentiation from simply using GPT-5 or Gemini with a translation layer bolted on. Enterprise and government customer traction has been the most concrete evidence of progress for both companies, with several public-sector deployments and BFSI pilot programmes now running in production rather than proof-of-concept, giving both companies revenue and reference customers that support the case for continued funding even without frontier-benchmark-topping model releases. What to watch: whether either company raises a round large enough to meaningfully close the compute gap with well-funded international competitors, how government procurement preferences for India-domiciled AI providers evolve as data-sovereignty policy discussions continue, and whether either company's models achieve adoption meaningfully beyond government and BFSI customers into the broader Indian startup ecosystem.

Original source: Inc42