New Delhi: India has introduced AI Kosh, a repository of non-personal datasets designed to support the development of large language models (LLMs). This initiative, led by the Ministry of Electronics and Information Technology (MeitY), is part of the country’s broader strategy to strengthen its artificial intelligence (AI) ecosystem.
AI Kosh will also facilitate startups’ access to the India AI Mission’s 18,693 GPUs (Graphics Processing Units), which will be available at subsidized rates, according to MeitY. The initiative is expected to accelerate AI research and model training within the country.
India’s Ambition to Build Its Own GPUs
At the same time, Union IT Minister Ashwini Vaishnaw announced that the government has initiated discussions with industry stakeholders to develop India’s own GPUs. While acknowledging that the process could take three to four years, Vaishnaw emphasized the country’s commitment to achieving self-reliance in semiconductor and AI hardware manufacturing.
“Developing GPUs is an extremely complex process, but we are determined to make it happen. As we progress, we will share more details,” Vaishnaw stated.
The Importance of GPUs in AI Development
GPUs play a crucial role in AI model training, as they power the algorithms that generate human-like text, speech, and videos. Since the introduction of OpenAI’s ChatGPT in 2022, the demand for GPUs has surged, with companies like Nvidia seeing their market valuation rise from $300 billion to $3.6 trillion.
Currently, India does not hold patents for manufacturing GPUs, making the country reliant on international chipmakers such as Nvidia, AMD, Amazon, Qualcomm, and Intel. However, the long-term goal is to develop India’s own advanced semiconductor technology and establish local fabrication units.
A senior government official noted, “Our objective is to create a sustainable AI ecosystem in India by developing domestic chip manufacturing capabilities. While this will take time, our immediate priority is to provide sufficient GPU access to build foundational AI models.”
According to previous reports, India aims to develop its first indigenous AI chip within five years.
Startups to Get Subsidized GPU Access
As part of the AI Kosh initiative, MeitY has received 67 applications from startups to develop AI models. Initially, three to five startups will be selected to receive immediate access to the GPU infrastructure. The cost of accessing a GPU in India has been set at ₹67 ($0.77) per hour, significantly lower than the global average of $2-$2.5 per hour.
“The compute platform will also provide access to public datasets from various ministries, covering key sectors such as agriculture and logistics. Additionally, MeitY’s Bhashini project, which focuses on Indian languages, will contribute its linguistic datasets to help startups develop AI models tailored for India’s unique needs,” Vaishnaw explained.
Industry Leaders Welcome the Initiative
Industry stakeholders have praised AI Kosh and India’s efforts to become self-reliant in AI and semiconductor technology. Rajkumar Sharma, President of the AICRA, emphasized the importance of AI Kosh in fostering innovation and reducing dependency on foreign data resources.
“This is a historic move for India’s AI ecosystem. Access to high-quality datasets and affordable compute infrastructure will empower Indian startups and researchers to develop AI solutions tailored to our country’s needs. Moreover, by setting the foundation for indigenous GPU development, India is taking a crucial step toward self-sufficiency in AI technology,” Sharma stated.
Future Roadmap for AI Development in India
One of the startups participating in the AI Mission is Bengaluru-based CoRover.ai, which specializes in Indic-language AI applications. CEO Ankush Sabharwal stated that their AI model, BharatGPT, will be designed as a general-purpose foundation model with sector-specific expertise added over time.
The AI compute infrastructure, launched under AI Kosh, includes GPUs from Nvidia, AMD, Intel, and Amazon Web Services, supplied by 10 shortlisted technology vendors. According to Sunil Gupta, CEO of Yotta Data Services, the GPU mandate under MeitY is expected to be a significant revenue stream for these vendors in the long run.
To support AI development, the Cabinet has allocated over ₹10,000 crore, with 40% of the funding dedicated to subsidizing GPU access for startups. However, Vaishnaw did not confirm whether the government would hold ownership stakes in the AI models developed under this initiative.
With these efforts, India is laying the groundwork to become a global AI hub, reducing reliance on foreign technologies while fostering homegrown innovation in artificial intelligence.
