Tamil Nadu Digital Sangam — Sovereign AI Research Park

Tamil Nadu announces $1.2B sovereign AI research park at Digital Sangam, partnering with Sarvam AI. 20MW AI data center. First sovereign AI research park in India. State-level commitment to indigenous AI compute and multilingual model development.

Sources

Odisha Sovereign AI Capacity Hub — $2.3B MoU

Odisha signs $2.3B MoU with Sarvam AI for a Sovereign AI Capacity Hub. 25,000 GPUs, 50MW facility. Largest single sovereign AI commitment by an Indian state. Part of Sarvam's full-stack sovereign AI platform selected under MeitY's IndiaAI Mission with 4,096 government-subsidized H100 GPUs at Yotta data center.

Sources

Sarvam-105B Sovereign Model Release

Sarvam AI releases Sarvam-30B and Sarvam-105B foundation models. 105B parameters, 128K context window, 9B active parameters per token (mixture of experts). Trained on 20+ Indian languages. $41M funding (Lightspeed, Peak XV, Khosla Ventures). First Indian sovereign model at competitive parameter scale.

Sources

Sarvam AI Closes $350M Round at $1.55B — Largest Indian AI Raise

Bengaluru-based Sarvam AI nears $300-350M Series funding at $1.5-1.55B valuation. Bessemer Venture Partners leads, with Nvidia, Amazon, and Prosperity7 Ventures participating. Largest pure-play Indian AI capital infusion to date. Funds the deployment scale-up of Vikram-30B (real-time conversational, 32K context) and the 105B reasoning model unveiled at AI Impact Summit February 2026.

Sources

Sarvam AI Open-Sources 30B and 105B Indian Language Models

Sarvam AI open-sourced Sarvam-30B and Sarvam-105B (MoE, 10.3B active params) under Apache 2.0 on Hugging Face and AIKosh. Trained from scratch on Indian language datasets. 25K+ developers on Sarvam Cloud by March 2026. Indian language processing 2-4x cheaper than GPT-4. Sarvam Startup Program launched for early-stage companies.

Sources

Gorilla-Yotta $2.8B AI Infrastructure Deal (20K+ B300 GPUs)

Gorilla Technology and Yotta expanded AI infrastructure collaboration to $2.8B with 20,000+ NVIDIA B300 GPUs targeted for deployment by September 2026. Establishes major position in India sovereign AI buildout.

Sources

Reliance/Jio commits Rs 10 lakh crore (~$110B) over 7 years to AI infrastructure

At India AI Impact Summit, Mukesh Ambani announced Reliance/Jio commitment of Rs 10 lakh crore (~$110B) over 7 years for AI-ready infrastructure: multi-GW data centers (Jamnagar 120MW+ online H2 2026), nationwide edge network, Jio platform AI integration, green power leverage. Largest single private-sector sovereign AI infrastructure commitment globally.

Sources

Adani Group targets $100B by 2035 for renewable-powered hyperscale AI data centers

Adani Group announced parallel commitment of $100B by 2035 for renewable-powered hyperscale data centers optimized for AI workloads. Complements Reliance/Jio push, creating competitive private-sector compute buildout within India's sovereign AI ecosystem.

Sources

Gorilla-Yotta Closes $2B Supermicro AI Infrastructure Deal — 20,736 B300 + 5,120 B200

Gorilla Technology Group (NASDAQ: GRRR) and Super Micro Computer closed an approximately $2B AI infrastructure supply deal in India, delivering 20,736 B300 cards, 5,120 B200 cards, networking, and related infrastructure to support Gorilla's Yotta AI infrastructure project in India. Companies entered a strategic framework to pursue further multi-billion-dollar opportunities across India and the broader Asia-Pacific (Singapore, Malaysia, Thailand, Indonesia, Philippines, Japan, Taiwan).

Sources

Sarvam Crosses Unicorn — $234M Series B First Close, $1.5B Post-Money, HCLTech $150M Anchor

Sarvam — India's full-stack sovereign AI company and lead foundation-model designate under IndiaAI Mission — announced $234M first close of $300M Series B at $1.5B post-money valuation. HCLTech committed $150M as strategic investor (round anchor). First sovereign-AI-mission-anchored Indian foundation-model lab to cross $1B threshold. Sarvam 30B + Sarvam 105B both pre-trained, fine-tuned, RL-aligned end-to-end in India on IndiaAI-Mission-provisioned GPU compute.

Sources