Sarvam AI Hits $1.5B Valuation as HCLTech Leads $234M Unicorn Round

India's newest AI unicorn just rewrote the rules of who funds frontier models

On June 15, 2026, Bengaluru-based Sarvam AI announced a $234 million first close of a planned $300 million Series B at a $1.5 billion post-money valuation, crossing the billion-dollar threshold and becoming India's newest AI unicorn112021. The story would be a routine venture milestone except for one detail: the largest check in the round — $150 million, buying roughly a 10.46% stake — came not from a Silicon Valley growth fund but from HCLTech, the IT services arm of Indian conglomerate HCL Group and a company with more than 227,000 employees and roughly $14.7 billion in annual revenue1421. Bessemer Venture Partners joined the round as a new institutional investor, with existing backers Khosla Ventures and Peak XV Partners continuing their support1113.

Read across the coverage, two threads dominate: the sovereign AI thesis now driving serious capital, and the unusual structure of a round in which a domestic IT incumbent took an equity position in a three-year-old startup rather than merely partnering with it.

From $41M to $1.5B in under three years

Sarvam was founded in August 2023 by Vivek Raghavan and Pratyush Kumar, both previously of AI4Bharat, the Indian-language AI research initiative at IIT Madras backed by Nandan Nilekani1119. The startup emerged from stealth in December 2023 with $41 million across seed and Series A rounds led by Lightspeed Venture Partners, with Peak XV and Khosla participating — at the time, the largest early-stage raise for an Indian AI startup23.

The valuation trajectory since then is steep by any standard. Reporting puts Sarvam's pre-Series B valuation at roughly $196 million in 2025, meaning the $1.5 billion figure represents a jump of about 7.6 times in under a year141921. Notably, Lightspeed — which led the earlier rounds — sat out the Series B first close, a divergence from its prior support that the coverage flags without explanation1925. The company reported unaudited revenue of Rs 45.1 crore (roughly $5–6 million) for FY26, underscoring that investors are pricing strategic positioning and growth rather than current earnings1924.

What Sarvam has actually shipped

The technical credibility behind that valuation rests on Sarvam's February 2026 release of two open-weight models trained entirely in India — a 30-billion-parameter model and a 105-billion-parameter flagship — using compute provisioned through the government's IndiaAI Mission, with no foreign model weights or transferred architectures141719. Both use mixture-of-experts designs to keep per-token inference costs down, a deliberately Indian trade-off: cheap inference is not academic when serving 1.4 billion people14.

Those models are already deployed at population scale, according to figures the company supplied and the coverage largely repeats without independent verification. Sarvam says its conversational platform handles more than 2 million interactions daily (doubling in two months), its inference layer processes about 10 million API calls a day (tripling in three months), and its speech models transcribe over 500,000 hours of audio monthly1119. Its multilingual voice agents have collected data from 17 million farmers for India's Ministry of Agriculture and Farmers Welfare, and a nationwide voice campaign supported policy renewals for 45 million policyholders at a leading insurer1112. A large fintech reportedly uses its agentic platform to support a sales force of more than 350,000 people11.

Why HCLTech wrote the biggest check

The most consequential reading of this round is what it says about India's IT services establishment. TechTimes frames HCLTech's move as industry self-preservation: Indian IT firms built a $250 billion industry implementing technology others built, and equity ownership of a domestic frontier-model lab is a hedge against that model eroding in the AI era14. CEO C. Vijayakumar called the investment "a significant step toward building India's trusted and globally competitive AI ecosystem"2021.

The plan is explicit division of labor — Sarvam builds the models; HCLTech builds industry-specific solutions on top and sells them through enterprise relationships across North America, Europe, and Asia1420. HCLTech retains its existing AI partnerships with OpenAI and Google Cloud, positioning the Sarvam stake as an India-specific layer rather than a wholesale pivot14. For Sarvam, the deal provides what few sovereign AI startups anywhere possess: an equity-linked distribution pipeline into global enterprise sales.

Sovereignty as an investment thesis — sharpened by Washington

Coverage converges on the timing being no accident. The round closed days after the US government blocked foreign nationals from accessing Anthropic's latest frontier models, a move Indian co-founder Pratyush Kumar pointedly connected to the fundraise, writing that observers "should not confuse access with ownership"1921. Both OpenAI and Anthropic describe India as their second-largest market after the US, yet India has produced few serious contenders in frontier model development, squeezed by high compute costs and limited capital1121.

Forbes India adds a skeptical counterweight: experts there say the real challenge of indigenously building every layer of a full-stack sovereign AI platform remains mostly unresolved, and the valuation is a bet on future scale rather than demonstrated economics1524. The Deep Tech's reporting is blunter still, noting Sarvam's compute moat is policy-dependent — the economics work because the state subsidized the compute bill under the IndiaAI Mission, which does not yet prove Sarvam can train its next generation at full market cost and still clear a venture return23.

The unicorn context: what this says about India's AI funding cycle

Sarvam's unicorn moment lands in a suddenly crowded field. Krutrim — Bhavish Aggarwal's Ola-backed venture — became India's first GenAI unicorn in 2024 at a $1 billion valuation; Neysa, an AI cloud infrastructure startup, crossed the threshold in February 2026 at roughly $1.4 billion after a $600 million round led by Blackstone; and Emergent followed in July 2026 with a $130 million Series C that also pushed it past $1 billion162223. Two Indian AI unicorns within weeks of each other is a pace the country had not previously seen, and Financial Express counts Sarvam as India's fifth unicorn of 2026 overall, alongside Juspay, Neysa, KreditBee and Skyroot Aerospace — three of the five being deep-tech plays1624.

The contrast that proves the thesis, though, is Krutrim's retreat. In May 2026, Krutrim paused its in-house chip-design and frontier-model programs, pivoted to enterprise AI cloud, cut more than 200 roles and withdrew its consumer app23. Krutrim attempted everything on its own balance sheet and found foundation-model economics unsustainable; Sarvam concentrated on open weights trained on subsidized compute, then attached itself to a distribution-rich strategic backer. One backed off the frontier; the other reached unicorn status on it in the same quarter. That divergence is the strongest available evidence for the specific path India's AI funding environment is now rewarding.

The open questions

The execution risk is real and the coverage names it plainly. Sarvam is simultaneously targeting three frontier research fronts — agentic AI, coding and cybersecurity — on a budget that is modest by the standards of American frontier labs, while scaling deployments across a dozen verticals including banking, insurance, agriculture, government services and defense, and while still needing to close the remaining $66 million of the Series B on an undisclosed timeline111419. Independent benchmark validation was only just beginning to appear as of mid-2026; most model scores remain self-reported14.

The bet being made here — by India's own IT establishment, with a sovereign-AI tailwind — carries a message beyond the valuation number. India's technology industry has decided it can no longer afford to rent its intelligence from abroad14. Whether Sarvam converts that conviction into a durable platform over the next twelve months, or becomes another cautionary data point in the graveyard of well-funded frontier ambitions, is the question this round leaves open.