India races to build own AI models as DeepSeek leaps ahead

    • India's initiative comes the same week as China’s DeepSeek AI startup wowed the world with its advances in competing with US industry leaders like OpenAI.
    • India's initiative comes the same week as China’s DeepSeek AI startup wowed the world with its advances in competing with US industry leaders like OpenAI. PHOTO: BLOOMBERG
    Published Fri, Jan 31, 2025 · 08:17 AM

    INDIA is racing to catch up in artificial intelligence as the government engages researchers, startups and companies to create foundational AI models within the next 10 months.

    Prime Minister Narendra Modi’s administration has selected 18 proposals that will be supported with computing infrastructure, data and capital to build AI-related applications in sectors such as agriculture and climate change, Technology Minister Ashwini Vaishnaw said on Thursday. 

    “The foundational models made in India will be able to compete with the best of the best in the world,” he said, adding six major developers will be able to build foundational AI models by year end.

    India is seeking to build computing capacity of just over 18,000 graphics processing units, Vaishnaw said. Homegrown E2E Networks and tycoon billionaire Mukesh Ambani’s Jio Platforms are among the companies vying for building this capacity using processing units such as Nvidia’s H100. 

    “The average per AI compute unit is 111.85 rupees (S$1.74) per hour,” Vaishnaw told reporters in New Delhi.

    India, which expects companies to invest US$30 billion to build data centres over the next few years, will fund 40 per cent of the computing price of the proposals, the minister said.

    BT in your inbox

    Start and end each day with the latest news stories and analyses delivered straight to your inbox.

    The initiative comes the same week as China’s DeepSeek AI startup wowed the world with its advances in competing with US industry leaders like OpenAI.

    Foundation models are a type of AI paradigm trained on vast datasets to perform a wide range of tasks. The models are designed to adapt and can be further fine-tuned for specific applications making them versatile.

    The federal government is aiming to get developers to build large as well as small language models as part of its ambitious US$1.2 billion IndiaAI mission. 

    The market for open-source AI models is dominated by US tech giants Alphabet and Meta Platforms, with DeepSeek emerging as a new challenger. India will soon host DeepSeek on local servers, Vaishnaw said.

    OpenAI, which operates ChatGPT, will support the development of applications at the IndiaAI Mission, the company has previously said. BLOOMBERG

    Share with us your feedback on BT's products and services