SOURCE : FINANCIAL EXPRESS | PUBLISHED : 8 MARCH 2025

The country will develop its own graphics processing units (GPUs) within the next three to four years, marking a major milestone in technological advancements, electronics and IT minister Ashwini Vaishnaw said on Thursday.

The initiative to build indigenous GPUs is significanct as the global AI chip industry is currently dominated by US-based companies. Additionally, with potential US restrictions on AI chip exports, India’s capability to develop indigenous GPUs will be crucial for technological self-reliance.

The government is also focusing on developing its own foundational AI models within the next eight to ten months, a move that will further strengthen the country’s AI infrastructure.

The datasets platform, AI Kosha, which was launched on Thursday, will serve as a repository of anonymous and non-personal data, providing crucial resources for startups to train AI models.

The government has invited private companies to contribute to the datasets platform. This is only the first version, with more data expected to be added by various government departments in the future.

The other platform which was also launched – the AI compute platform — will enable startups, researchers, and academic institutions to access GPUs at an affordable hourly rate. “The IndiaAI mission was started a year ago, and today, we have a common compute facility with 14,000 GPUs up and running,” Vaishnaw said. He added that more GPUs would be added every quarter by empanelled companies.

Currently, ten companies, including Jio Platforms, E2E, NxtGen, Yotta, and Tata Communications, have been empanelled to provide AI compute infrastructure.

Vaishnaw said that India’s approach is unique as it involves public funds to create a shared compute facility, contrasting with other countries where AI computing power is concentrated within a few large corporations.

To make AI compute even more accessible, the government is offering a 40% subsidy, bringing down the cost of accessing GPUs for researchers and startups to under Rs 100 per hour.

For building its indigenous foundational AI models, the government has received 67 proposals from startups. “Our team is inundated with applications, and we are currently evaluating them. Initially, we will select three to five mature proposals to move forward,” Vaishnaw said.

On the regulatory front, Vaishnaw highlighted the country’s distinct approach. “Unlike other regions, we believe techno-legal solutions are the right way to address challenges in AI,” he said.

The AI Safety Institute, established by the government, is currently working on 13 projects in collaboration with industry stakeholders to mitigate potential AI-related risks. One such project, being undertaken by IIT Jodhpur, focuses on detecting deepfakes.

The industry is also developing a code of ethics to ensure responsible AI deployment and mitigate potential harms to users.

SOURCE : FINANCIAL EXPRESS | PUBLISHED : 8 MARCH 2025