The Indian government is set to develop indigenous generative artificial intelligence (AI) models within a timeframe of six to eight months, Union Minister of Electronics and Information Technology Ashwini Vaishnaw announced on Thursday. This initiative was unveiled during the Utkarsh Odisha Conclave, where the minister shared that the DeepSeek-R1 AI model has been successfully implemented on domestic servers, and a technical report on the model is currently under expert evaluation. This announcement comes shortly after India’s participation in the 55th annual meeting of the World Economic Forum held in Davos-Klosters, Switzerland.
India Poised to Develop Indigenous AI Models
In a recent report by Moneycontrol, Vaishnaw elaborated on various AI-related initiatives during his address at the Conclave. He indicated that India has acquired more than 18,600 high-performance GPUs necessary for establishing the AI infrastructure and enhancing computing capabilities. These resources are slated to serve as the foundation for the generative AI models set to be developed this year.
“We believe there are at least six major developers capable of producing AI models within a range of six to eight months, potentially even four to six months under more optimistic conditions,” said Vaishnaw.
The Minister also emphasized the urgent necessity of creating a shared computing facility, deeming it a critical pillar for fostering a robust AI ecosystem. This initiative is part of the India AI Mission and aims to support the computational needs of researchers, startups, and academic institutions across the country.
Elaborating on the planned computing facility, which will comprise cloud-based servers designed for AI inference, Vaishnaw reported that expectations have been surpassed with the country’s acquisition of over 18,600 GPUs. This includes 12,896 Nvidia H100 GPUs, 1,480 Nvidia H200 GPUs, and 742 AMD MI325 and MI325X GPUs. The initial target was to procure 10,000 high-end AI chipsets, he mentioned.
“DeepSeek AI was trained using 2,000 GPUs, while ChatGPT utilized 25,000 GPUs, and now India boasts 18,000 high-end GPUs at its disposal. We have established a strong compute facility to support our AI goals,” he added. Once operational, the Indian AI models are expected to be accessible to all. However, the minister did not clarify whether these models would be open-source or available through specific platforms and application programming interfaces (APIs).
On the topic of DeepSeek, Vaishnaw noted that the DeepSeek-R1 AI model has already been implemented, and experts are in the process of evaluating the technical report to enhance their understanding of the model, as referenced in a CNN-News18 video.
Shifting focus to AI safety and security, Vaishnaw announced plans to establish an AI safety institution based on a hub-and-spoke model. This approach differs from those of other nations, where a singular body governs AI regulation; instead, India will facilitate collaboration among multiple institutions to create safety tools and regulatory frameworks.
