'We will start reaching out to researchers, academics, and experts on AI from top universities and companies from across the world.'
Kindly note the image has been posted only for representational purposes. Photograph: Kind courtesy DIY Team/Pixabay.com
The government will soon start reaching out to a global network of experts on artificial intelligence (AI) and machine learning (ML) to help mentor companies and startups selected to build India's domestic AI foundation model, according to a senior government official.
The ministry of electronics and information technology (Meity) aims to engage these mentors -- including experts from top AI companies -- to provide guidance in the initial phase of setting up the framework for foundation models, the official said.
"We will start reaching out to researchers, academics, and experts on AI from top universities and companies from across the world. Once we have selected the companies that will work on the foundation models, our plan is to pair them with the mentors that come on board," the official said.
These mentors, another official said, would also help the selected companies design and develop the foundation model in accordance with the demand and further help in distilling the model to fit the right tools.
"A lot of initial work is likely to be academic in nature and will involve training, research, and development.Once we have enough manpower that is trained on the basics, we will begin work on collation and training of models on datasets," the official said.
Work on the AI foundation model will take place under the Rs 10,372 crore (Rs 103.72 billion) India AI mission.
On January 30, Union IT Minister Ashwini Vaishnaw said the government was talking to at least six startups to develop multiple foundation models in various areas.
Vaishnaw said the government had been working with experts in the field of large language model (LLM) and small language model (SLM) over the past 18 months to develop the framework for India's foundational AI model.
This is not the first time the government is tapping into a global network of high-skill talent.
In the Budget 2025, the finance minister proposed a presumptive taxation regime for non-Indian residents providing services to electronics-manufacturing companies operating in India.
Senior government officials told Business Standard this would help companies attract top global talent easily.
Under the new presumptive regime, which will take effect on April 1, the government has proposed that only 25 per cent of the remuneration received by non-residents will be treated as taxable profits.
This, in turn, will mean that a majority of non-residents who stay in India for extended periods will pay an effective tax of less than 10 per cent on their gross income, experts said.
Nvidia, AMD, HPE in Talks with Govt for Local GPUs
Surajeet Das Gupta
Three global technology players -- Hewlett Packard Enterprise (HPE), Advanced Micro Devices (AMD), and Nvidia -- are in discussions with the government for a collaborative effort to help design and develop 'indigenous' graphic processing units (GPUs), as well as in building the artificial intelligence (AI) ecosystem in the country.
Talks are ongoing, according to a senior government official, with an announcement on the final terms of collaboration expected soon.
In a recent interview with Business Standard, Minister for Electronics and Information Technology Ashwini Vaishnaw stated that the government has begun consultations with industry on India-designed GPUs and will support the initiative.
Discussions have already taken place with the Centre for Development of Advanced Computing, as well as with global firms that could assist in the effort, the minister added.
These discussions come as Meity finalises a list of companies empaneled to receive government support for procuring GPUs and developing computing infrastructure critical for AI.
The government has received offers for more than 18,000 GPUs -- nearly double its initial target of 10,000.
For context, India's largest supercomputer currently operates on 700 GPUs.
Industry applications, according to sources, include a range of GPUs, such as Nvidia's H100, H200, A100, and L4; AMD's MI 325X; and Intel's Gaudi-2, among others.
Prices for on-demand computing start as low as Rs 64 per hour, according to industry sources.
Vaishnaw has said the government aims to provide computing power to startups, research centers, and universities at roughly $1 per hour -- one of the lowest rates globally.
While the proposed India-designed GPUs may not match the world's most advanced processing power, sources familiar with the matter say they will offer a cost-effective solution with potential for innovation in AI applications.
A similar approach has been seen in China, where the DeepSeek generative AI tool has developed a lower-cost model compared to ChatGPT, relying on less advanced GPUs with reduced processing power.
The Indian government is also considering asking proposals from startups and developers, offering incentives to create foundational AI models under its Rs 10,000 crore AI mission.
Vaishnaw, in a recent press conference, stated that six developers are currently working on foundational AI models in the country.
Nvidia, for its part, has expressed interest in co-developing a chip in India by leveraging the country's chip design capabilities.
HPE plays a critical role in providing server infrastructure optimised for GPUs, enhancing AI-driven computational workloads.
AMD, an old player in the GPU market since its 2006 acquisition of ATI Technologies, has advocated for a collaborative AI ecosystem, warning that monopolisation could stifle innovation and harm everyone.
Feature Presentation: Ashish Narsale/Rediff.com
