Global AI race: India to build own foundational models
New Delhi: India on Thursday outlined its global artificial intelligence (AI) ambitions with plans to develop its own foundational AI model, aiming to compete with existing models such as ChatGPT and DeepSeek R1. As part of its AI strategy, India announced the creation of a common compute facility powered by 18,693 GPUs, which will be made available at affordable rates for startups and researchers.
The announcement comes as Chinese company DeepSeek has gained prominence, with its AI model surpassing ChatGPT as the top-ranked free app on Apple’s App Store. The move also signals India’s intention to challenge the dominance of US-based AI firms, particularly OpenAI.
IT Minister Ashwini Vaishnaw stated that India is committed to building a world-class foundational model capable of competing with the best AI models globally. The government has invited proposals to develop AI models that align with India’s linguistic and cultural context, ensuring the datasets are locally relevant and free from biases.
A foundational AI model serves as the base for various AI applications and requires significant computational resources. The government’s common compute facility will be made available at a cost significantly lower than global benchmarks.
The minister assured that the cost per GPU hour will be under one dollar, with the government covering 40 per cent of the total cost. Attractive six-month and annual plans will make it even easier to access the services offered to developers and researchers.
The government has empaneled multiple companies, including Jio Platforms, CMS Computers, Tata Communications, E2E Networks, and Yotta Data Services, to contribute to the AI infrastructure. These companies will offer a variety of AI computing units like NVIDIA H100, H200, A100, AMD MI300X, and Intel Gaudi 2 at competitive prices to further support AI research and development. The initiative also includes the establishment of an AI safety institution, with eight projects approved to focus on AI risk mitigation.
Vaishnaw stated that at least six major developers and startups are expected to build foundational AI models within the next 4-10 months. He emphasised that advancements in algorithmic efficiency could significantly reduce development time and costs. He also announced that DeepSeek’s AI model will be hosted on Indian servers to address data privacy concerns. Experts have raised issues regarding DeepSeek’s data storage policies, as its documentation indicates that user data is stored on servers in China.
As part of the IndiaAI Mission, the government has approved a funding outlay of Rs 10,372 crore to strengthen AI infrastructure. The planned 18,693 GPUs are expected to provide significant computing power, considering that DeepSeek was trained on 2,000 GPUs and ChatGPT-4 on 25,000 GPUs.
The IndiaAI Mission aims to harness AI for large-scale applications in healthcare, education, agriculture, logistics, and weather forecasting. The government has approved funding for 18 projects focusing on agriculture, learning disabilities, and climate change.
AI safety remains a key focus, with India adopting a collaborative approach through a hub-and-spoke model for AI safety institutions. Approved projects include initiatives on machine unlearning (IIT Jodhpur), synthetic data generation (IIT Roorkee), AI bias mitigation strategies, explainable AI frameworks (Defence Institute of Advanced Technology Pune), privacy-enhancing strategies (IIT Delhi, IIIT Delhi, TEC), AI ethical certification frameworks, AI algorithm auditing tools, and AI governance testing frameworks.
India’s AI ecosystem promises a bright future, with 240 universities offering AI courses and 100 institutes building 5G labs. On this occasion, Minister of State for Electronics & IT Jitin Prasada, senior MeitY officials, and IndiaAI representatives were present.
Regarding potential export restrictions on GPUs and AI chips from the US, Vaishnaw stated that India is viewed as a trusted technology partner worldwide. He expressed confidence that India’s commitment to respecting intellectual property rights and fostering technological collaboration would ensure the continued growth of the IndiaAI Mission.
With over $30 billion invested in the semiconductor sector, India is well-positioned to advance AI-driven industrial applications. The initiative has already garnered global recognition, including acknowledgement at the World Economic Forum in Davos.
