India cannot afford to remain a passive consumer in the AI era and must urgently build its own frontier-scale artificial intelligence models to shape global technology rules, according to Sarvam AI co-founder Pratyush Kumar. Speaking at the CII Business Summit during the session ''''AI & India's Future: Rule Maker or Rule Taker'''', Kumar emphasized that India must move beyond debates around whether it should build its own AI models and instead focus on creating long-term strategic capability. As reported by The Times of India, Kumar argued that ownership of foundational models is critical for economic value creation across industries, governance, science and manufacturing, warning that 'We can rent it for now until we don't have it, but you have to build it. You have to own the destiny around that' - emphasizing the importance of building domestic AI capabilities rather than relying on external solutions.
According to Kumar's statements at the CII Business Summit, Sarvam AI is now preparing to train its first trillion-parameter AI model within the next nine months, marking what could become a major milestone for the country's indigenous AI ambitions. Kumar revealed that the company has successfully built what he described as a proof of concept for India's indigenous model-building capabilities, stating that 'India can train a model where the data work, the algorithm work, the research work and the infrastructure work could lead to training a model end to end'. This timeline represents a significant commitment to domestic AI development and positions India's private sector in the global race for large-scale AI infrastructure.
Gautam Adani outlined a comprehensive three-layer AI framework consisting of power generation, compute infrastructure and AI applications, emphasizing that sovereign ownership of data and compute infrastructure would become critical in the next phase of technological competition. Speaking at the CII Annual Business Summit 2026, Adani stressed that 'If our data is processed on distant shores, it means our future is being written on foreign shores' and urged India to treat AI not merely as software but as strategic infrastructure spanning energy, data centres, chips, networks, compute and talent. He noted that India's scale of domestic demand across manufacturing, mobility, logistics and digital services positioned it uniquely to build large-scale AI and energy infrastructure, with the country having crossed 500 gigawatts of installed power capacity and being on track to quadruple capacity to 2,000 GW by 2047.
The access question is equally important for India's AI ambitions, with executives pointing to the success of UPI, Aadhaar, and Open Network for Digital Commerce (ONDC) as models for affordable, scalable infrastructure. Several industry leaders believe subsidized compute access for researchers, startups, vernacular AI teams, and founders outside major technology hubs could help create a broader and more inclusive AI ecosystem instead of concentrating access among only a few large players. Vishal Sirohi, CEO of Island Computing, noted that India still lacks enough engineers who have operated AI systems at hyperscale, emphasizing that 'GPUs, datacenters, and capital can be procured, financed, or partnered. Operational depth has to be built at home'. The real challenge may come down to megawatts rather than just GPUs, as Sanjay Phadke from Vayana argued that 'The constraint that will quietly determine whether India trains its own frontier models — or remains a fine-tuner of others' — is megawatts, not just GPUs'.
The IndiaAI Mission has committed more than ₹10,000 crore toward compute capacity, foundational models, skilling, and ecosystem development, with more than 38,000 GPUs now being made accessible through empanelled providers. Major infrastructure developments include Google's work on a large AI-focused datacenter hub in Visakhapatnam and Reliance Industries' plans for a massive AI infrastructure campus in Jamnagar. The government has also expanded semiconductor initiatives under the India Semiconductor Mission as India tries to reduce long-term dependence on imported chips and foreign cloud ecosystems. Industry estimates suggest India's datacenter market is expanding rapidly, with operational capacity expected to grow several-fold by the end of the decade as AI adoption accelerates across enterprises, financial services, healthcare, manufacturing, and public infrastructure.
The Indian AI market is experiencing significant growth momentum, with projections showing the market could reach an estimated $27.7 billion by 2032, with a compound annual growth rate (CAGR) of 19.2%. However, India faces intense global competition, particularly from China's rapid advancement in AI applied to physical tasks and robotics, where the country controls key components like lidar sensors and leads in industrial robot installations. The urgency is heightened by major investments flowing into data center expansion, with projections showing capacity could increase fivefold by 2030, reaching over 8 GW. Companies like Yotta Data Services are investing over $2 billion to build AI superclusters, planning to deploy thousands of Nvidia Blackwell GPUs by August 2026, positioning India among locations capable of large-scale AI infrastructure. Adani's commitment includes a further USD 100 billion towards the group's data centre business, including partnerships with Google to build what he described as India's largest gigawatt-scale data centre campus in Visakhapatnam, with Microsoft and companies including Flipkart and Uber partnering on data infrastructure initiatives.
