Sarvam AI

Sarvam AI is an Indian AI company focused on building foundation models optimized for Indian languages and the diverse linguistic landscape of the subcontinent. Backed by Lightspeed and Khosla Ventures, the company is addressing the challenge of making AI accessible to India's 1.4 billion people, most of whom communicate in languages other than English.

Sarvam's flagship is Sarvam-105B, a 105 billion-parameter model launched in February 2026 and trained entirely from scratch on government-subsidized GPUs — positioned as India's first sovereign foundation model. It supports all 22 official Indian languages and is competitive with global models on standard benchmarks. To keep training and serving within India, Sarvam operates its own compute infrastructure, including an H100 cluster of roughly 3,400 GPUs alongside a newer Blackwell deployment. The company has also signed a memorandum of understanding with Tamil Nadu to build India's first Sovereign AI Park, a roughly 1.2 billion dollar initiative.

Sarvam's models are trained specifically on Indian language data across Hindi, Tamil, Telugu, Bengali, Marathi, and other major Indian languages, providing significantly better performance on Indian-language tasks than English-centric models adapted for India. The company offers APIs for text generation, translation, speech recognition, and text-to-speech in Indian languages.

Sarvam AI represents the growing importance of regional AI models that serve populations underserved by Silicon Valley's primarily English-focused AI systems. As a sovereign-AI champion, it has become a focal point in India's debate over how much of the country's AI future should depend on a small number of foreign frontier-model providers. Sarvam competes with Krutrim (Ola's AI company) and international players adapting models for the Indian market.

In 2026, Sarvam raised $234 million at a $1.5 billion valuation — its first round as a unicorn — in a financing led by IT-services giant HCLTech. The round underscored growing domestic conviction that India needs home-grown frontier-model capacity rather than depending on a handful of foreign providers.