In a landmark collaboration, IIT Madras and the Indian Space Research Organisation (ISRO) have announced a joint initiative to develop India’s first domestically designed AI inference chip. The project, codenamed ‘Shakti-AI’, aims to reduce the country’s dependence on imported semiconductor technology for machine learning workloads.

The chip will be based on the open-source RISC-V architecture and is designed to handle edge AI tasks in satellite imagery analysis, autonomous navigation, and natural language processing for Indian languages.

Professor V. Kamakoti, Director of IIT Madras, said the chip could be ready for fabrication by late 2026. “We’re not trying to compete with NVIDIA on training workloads. Our focus is on efficient, low-power inference that can run on Indian-made hardware,” he explained during the announcement at the India Semiconductor Summit in Bengaluru.

The initiative is part of the broader India Semiconductor Mission (ISM), which has allocated ₹76,000 crore ($9.1 billion) for building a domestic chip ecosystem. The government has already approved three semiconductor fabs, including the Tata-PSMC joint venture in Gujarat.

Industry observers see the Shakti-AI project as a critical step towards India’s semiconductor self-sufficiency. With global AI chip demand expected to triple by 2028, domestic production capability could give Indian startups a significant competitive edge.