Maharashtra Signs LoI With Sarvam to Build 20 MW Sovereign AI Infrastructure

The Maharashtra government has signed a letter of intent with Sarvam AI to build sovereign AI infrastructure across the state, with plans to create 20 MW of AI-optimised compute capacity, scalable up to 50 MW, officials said at the India AI Impact Summit 2026 recently held in New Delhi.

Chief Minister Devendra Fadnavis and Sarvam AI Co-Founder Pratyush Kumar presided over the exchange of the LoI, according to the Chief Minister’s Office. The collaboration aims to deploy AI systems at scale across healthcare, education, agriculture, welfare delivery and public administration.

In a statement, Sarvam AI said the partnership seeks to strengthen Maharashtra’s sovereign compute capacity and deliver measurable impact in public services. The company described the initiative as part of a broader push to build state-controlled AI infrastructure for real-world deployment.

The agreement outlines plans to establish large-scale AI infrastructure within the state, advance research and innovation, and promote skill development in emerging AI domains. The proposed compute capacity of 20 MW, expandable to 50 MW, is positioned as a foundation for training and deploying advanced AI models locally.

The move comes as Sarvam AI deepens its alignment with national AI policy. The company is a primary beneficiary of the IndiaAI Mission and has received government-backed GPU subsidies and infrastructure support. 

At the India AI Impact Summit, Sarvam launched India’s sovereign foundational model, including Sarvam 30B and Sarvam 105B.

Sarvam has also partnered with the Unique Identification Authority of India to integrate multilingual, voice-based interfaces into Aadhaar services. The company argues that voice-enabled public platforms could expand access in a country marked by linguistic diversity and uneven literacy levels.

Beyond Maharashtra, Sarvam has entered into similar state-level partnerships with Odisha and Tamil Nadu. These initiatives include plans for sovereign AI parks, data-residency frameworks and sector-specific AI tools for agriculture, mining and workforce training.