In the move toward AI self-reliance, the Indian government has accelerated its IndiaAI Mission, planning to roll out a National Large Language Model (LLM) by the end of 2025, expected to be launched at the India–AI Impact Summit scheduled in February 2026. This sovereign initiative under the IndiaAI Mission, backed by a massive deployment of 38,000 GPUs available at an ultra-affordable rate of ₹65 per hour, is empowering 12 domestic companies to pioneer foundational AI models tailored to India’s diverse linguistic and cultural landscape.
This was announced at a pre-summit event for the upcoming India–AI Impact Summit 2026 at the India Mobile Congress (IMC) 2025 in New Delhi today. The India–AI Impact Summit 2026 is scheduled to be held on 19–20 February 2026 at Bharat Mandapam, New Delhi.
In the inaugural address, Secretary MeitY, S. Krishnan said that the Indian model will be ready by end of this year, ready to be launched at the the India–AI Impact Summit 2026. He said, “End of this year, we should have entirely an Indian foundational model. By the time of the India AI Summit, we should be able to launch India’s sovereign model as well.”
He further added, “We have adopted innovative means by learning from the experiences of others to build viable projects and products that will truly make a difference for us. In partnership with the private sector, accessible to the public sector – ensuring a level-playing field for all providers, is both innovative and frugal. It ensures that with the least amount of resources we are able to ensure availability for all. A number of international agencies have also found our approach very appealing, to build a model which can be used for the rest of the global South.”
Additional Secretary -MeitY, CEO – IndiaAI and DG- NIC, Abhishek Singh, said, “The Government’s IndiaAI Mission, approved last year, is addressing key gaps by enabling affordable compute, creating a national data platform, supporting foundation models, advancing AI skilling, and ensuring safe and trustworthy AI. Today, with 38,000 GPUs available at just ₹65 per hour and 12 companies developing foundation models, we envision to launch an Indian Large Language Model by year-end, reducing dependence on foreign systems.”
Singh added that India’s sovereign AI model will be completely trained on Indian data sets and will be hosted on India servers.
In February this year, PM Narendra Modi announced that India is building its own Large Language Model, underscoring the nation’s growing focus on advancing artificial intelligence technologies. The aim of the model, called Bharat Gen, is to create AI that is ethical, inclusive, multilingual, and deeply rooted in Indian values and ethos. The platform integrates text, speech, and image modalities, offering seamless AI solutions in 22 Indian languages.
By providing subsidised GPU clusters, expanded recently with an additional 3,850 units, the Modi government aims to lower barriers for startups and researchers, fostering breakthroughs in sectors like healthcare, agriculture, and governance. The forthcoming National LLM is designed to process and generate content in India’s 22 official languages and beyond, addressing gaps in global models that often overlook regional nuances.
Fuelling this momentum are 12 selected companies, including Tech Mahindra, Sarvam AI, Soket AI, Avataar.ai, Fractal Analytics, and others like Genloop and Zenteiq Aitech. These firms, chosen through competitive calls for proposals under the ₹10,300 crore IndiaAI Mission, are leveraging the GPU infrastructure to train indigenous foundational models.
The mission’s latest tranche brings the total GPU count to over 38,000, distributed across national clusters to ensure equitable access. “This is not just about hardware; it’s about building AI that understands India,” said a MeitY spokesperson, emphasizing the focus on ethical, bias-free models trained on localized datasets.
Complementing the compute power, the initiative includes the establishment of 600 AI-focused data labs nationwide, aimed at grassroots innovation and skill-building. These labs will serve as hubs for experimentation, data annotation, and application development, targeting underserved regions to bridge the digital divide. Early adopters report cost savings of up to 90% compared to international cloud providers, enabling rapid prototyping. Some models are already in beta for applications like multilingual chatbots and predictive analytics in public services.
