India's AI Push: 3 New Startups & 34K GPUs

New Delhi, May 30 (PTI) After Sarvam AI, India on Friday selected three more startups -- Soket AI, Gan AI, Gnani AI -- for building indigenous artificial intelligence foundation models.

In line with its global AI ambitions backed by a comprehensive plan that entails enhanced AI infrastructure and local language model development, India has also announced availability of 16,000 more GPUs that would take the compute facility available to startups and researchers here to 34,000, with the support of industry partners.

The expanded compute capacity on cloud will provide a common computational AI platform for training and inference, crucial to develop indigenous foundational models and AI solutions tailored to the Indian context.

IT Minister Ashwini Vaishnaw said significant progress has been made on India AI Mission, with focus on democratisation of technology.

The compute facility supercharged with 34,000 GPUs will enable India to develop AI ecosystem in a big way, he said.

Seven bidders have offered their commercials for various categories of AI compute units (GPUs). These include Cyfuture India, Ishan Infotech, Locuz Enterprise Solutions, Netmagic IT Services, Sify Digital Services, Vensysco Technologies, and Yotta Data Services.

At the same time, three more teams -- Soket AI, Gan AI, Gnani AI -- have been selected for building indigenous artificial intelligence foundation models.

"Like Sarvam, these three teams also have a very big target ahead of them. Whichever sector they focus on, they must be among the top five in the world," Vaishnaw said.

Put simply, foundation models in generative AI are large, pre-trained models that form the base for a variety of AI applications.

The Minister further said that 367 datasets have already been uploaded to AI Kosh.

He also highlighted IndiaAI Mission's role in driving reverse brain drain, and creating a comprehensive ecosystem entailing foundational models, compute capacity, safety standards, and talent development initiatives.

Vaishnaw emphasised that these efforts are aimed at building a complete and inclusive AI ecosystem in India.

In April this year, Sarvam AI was selected to build India's first indigenous AI foundational model, marking a key milestone in the country's AI innovation ecosystem.

Soket AI will develop open source 120 billion parametres foundation model optimised for the country's linguistic diversity targeting sectors such as defence, healthcare, and education.

Gan AI will create 70 billion parameters of multilingual foundation model targeting capabilities to surpass the current global leader.

Gnani AI will build a 14 billion parameter Voice AI foundation model delivering multilingal real-time speech processing with advances reasoning capabilities.

Ganesh Gopalan, Co-Founder and CEO of Gnani.ai, said in a statement, "We are honoured to be selected under the IndiaAI Mission to develop large language models that truly represent India's linguistic diversity. At Gnani.ai, our mission has always been to make technology more inclusive and accessible".

Gopalan further said Gnani.ai is keen to "lead the way in developing voice-to-voice large language models for India and the world, because we believe transformative AI must speak the language of the people it serves".

Meanwhile, under the IndiaAI Applications Development Initiative, Vaishnaw also announced the winners of the IndiaAI I4C CyberGuard AI Hackathon, jointly organised with Indian Cyber Crime Coordination Centre (I4C), Ministry of Home Affairs.

"The Hackathon resulted in the development of AI-based solutions to enhance the classification of cybercrime complaints and support the identification of emerging crime patterns, trends, and modus operandi on the National Cyber Crime Reporting Portal (NCRP). These models can interpret complex inputs such as handwritten FIRs, screenshots, and audio calls with improved speed and accuracy," an official release said.