India to compete with ChatGPT and DeepSeek R1 with new language model

In a significant push towards AI self-reliance, the Indian government has announced plans to develop its own large language model (LLM) as part of the Rs 10,370 crore IndiaAI Mission, IT Minister Ashwini Vaishnaw revealed on Thursday.

The initiative aims to create an AI model tailored to India’s diverse linguistic, cultural, and contextual needs, while eliminating biases.

The government has approved the procurement of 18,693 graphics processing units (GPUs) - essential for training advanced AI models. Ten companies have been selected to supply these high-end chips, including Jio Platforms, Tata Communications, E2E Networks, CMS Computers, Ctrls Datacenters, Locuz Enterprise Solutions, NxtGen Datacenter, Orient Technologies, and Vensysco Technologies. However, Yotta (Hiranandani Group-backed) is providing 9,216 GPUs (nearly half of the total).

The government is collaborating with at least six developers for the project, with an expected completion timeline of four to eight months.

To facilitate AI development, the government will soon launch a common compute facility, allowing startups and researchers to access GPU resources at subsidized rates.

- High-end GPUs: Rs 150 ($1.80) per hour

- Lower-end GPUs: Rs 115.85 ($1.40) per hour

- 40% subsidy for users to further lower costs

Minister Vaishnaw noted that global GPU access costs around $2.5–$3 per hour, whereas India’s initiative will provide the same for around $1 per hour after subsidies.

As part of the IndiaAI Mission, the government has also selected 18 AI-driven solutions for funding in agriculture, climate change, and learning disabilities.

This move comes just days after China’s DeepSeek AI lab launched its low-cost foundational model, DeepSeek R1, underscoring the growing competition in the global AI landscape.