In a major push towards self-reliance in artificial intelligence, the IndiaAI Mission has signed a Memorandum of Understanding (MoU) with the Indian Parliament to leverage its vast multilingual data for training indigenous AI models. IndiaAI mission will use this data for the development of indigenous AI technology, Union IT Minister Ashwini Vaishnaw announced during a panel discussion at the Raisina Dialogue 2025, organised by the Observer Research Foundation (ORF), according to a PTI report.
Vaishnaw reportedly said it is important to develop India’s own LLM (large language model like ChatGPT), as open source technology available at present may not remain open in future like OpenAI.
Highlighting the importance of developing India’s own large language model (LLM), Vaishnaw noted that publicly available AI technologies, such as OpenAI’s models, may not remain open-source in the future. To address this, the government has established AI Kosh, a common compute infrastructure for AI research.
“The most important part of the entire AI journey is getting the data sets. We have set up a common compute kind of structure called AI Kosh. Today morning, the India AI Mission and the Indian Parliament, we signed an MoU because Parliament has built a very large data sets in multiple languages over a period of time. That will be a very good training resource for training our models,” Vaishnaw was quoted as saying in the report.
He also said there are many similar data sources available through organisations like Doordarshan, All India Radio, etc.
OpenAI and Elon Musk are in a legal tussle over the transition of OpenAI from a non-profit to a for-profit organisation. According to the report, Vaishnaw, on a lighter note said that maybe OpenAI should change its name as well if it becomes a proprietary technology.
Vaishnaw also emphasised the need for India to develop its own graphics processing unit (GPU) chips, crucial for AI computing. The government has already initiated work on this project, aiming for a functional GPU within three to five years. Two different instruction set architectures are under evaluation, with experts expected to finalise the approach soon.
When asked about the timeline for achieving indigenous GPU capability, Vaishnaw reportedly said, “In three to five years to get our good, reasonably good capability, GPU up and running. Basically, there are two approaches which we are evaluating seriously – two different types of instruction sets. Which one will we finalize, or will we take both the instruction sets and move forward? That is something which the experts need to decide,” the minister said, according to the report.
Massive AI Investment Proposed
Vinod Dham, known as the Father of the Pentium Processors and founder of Indo-US Venture Partners, suggested that India should allocate USD 50-100 billion for AI development, similar to the India Semiconductor Mission (ISM). He stressed that AI requires significant computing power, which in turn demands large-scale GPU production and energy resources. Vaishnaw responded positively, stating that such funding models could be considered.
Dham reportedly said that the timeline given by the minister for the development of GPU is fairly reasonable, adding that India should leverage the OpenAI model to build its own AI model but should not use Western AI models to do confidential things.
According to the report, OpenAI Chief Economist Ronnie Chatterji said during the discussion that “data sets are a huge ingredient to build large language models and India, and the government in particular, has done a fantastic job in terms of making those data sets public.”
“If you have a million people coming out every year with a new set of skills and a population of 30 years old or younger for the next 30 years, it’s a huge opportunity for India to stay on the cutting edge,” Chatterji was quoted as saying in the report.
Some people read for free. A few choose to support. If you found TelecomTalk useful, you can help keep us running.
Support Telecomtalk
