The sovereign AI race: Why India can't afford to depend on foreign tech

India is gearing up for one of its most ambitious technology missions yet, building a sovereign AI tech stack that spans compute, data, models, and applications. The goal is to reduce reliance on foreign platforms, safeguard national security, and capture the economic gains of the AI era. Recently, Union Minister Ashwini Vaishnav also announced that India will soon have its chip. But experts warn that while the intent is strong, the journey has only just begun, and the terrain ahead is steep.

The Ministry of Electronics and Information Technology (MeitY) has laid out a roadmap to make this vision a reality with IndiaAI Mission, where industry experts say that in an AI-first world, control over data, algorithms, and compute infrastructure defines who sets the rules.

Under the IndiaAI Mission, the government has procured about 34,000 GPUs from industry giants like Nvidia, AMD, AWS, and Intel; a significant step, but still just a fraction of what’s needed to compete globally. The aim is to pool national demand, much like MeitY’s bulk GPU procurement, to ensure startups, researchers, and enterprises can access cutting-edge compute without prohibitive costs. Startups like Sarvam, Gnani.ai, and Soket AI have been selected to build a homegrown LLM.

Yet, the hardware is just one piece of the puzzle. “Without building domestic capacity, India faces structural ceilings,” warns Nitin Mittal, Deloitte’s Global AI Leader, speaking to Business Today. In his view, the 2025–2030 window offers a chance to leverage India’s design talent and pool resources. But beyond that, the absence of indigenous advanced-node semiconductor fabs, High Bandwidth Memory (HBM) capabilities, and advanced chip packaging could undermine long-term competitiveness. “Failing to secure the silicon stack means forfeiting AI-led GDP gains, losing standards-setting influence, and watching our best talent migrate abroad,” he says.

That risk extends beyond economics. Dr Vivek Raghavan, co-founder of Sarvam, cautions that relying solely on overseas AI models is strategically short-sighted. “Foundational models shape language, decision-making, and whose worldview the AI reflects,” he says. “India has all the ingredients to lead—linguistic diversity, a rapidly digitising economy, ambitious public infrastructure, and deep technical talent. What’s missing is the will to align these into a national capability.”

Building a national tech stack is crucial, but experts say that colonial history may have made us risk-averse. However, that must change.

Punit Pandey, founder and CIO of Ojas Softech and Astrosage AI, believes cloud sovereignty is equally urgent. “If India had its cloud, we’d control our data. Without it, we’re dependent,” he notes. Some argue we’re doing fine, but looking around, we have no global tech products despite the nation’s dominance in Silicon Valley’s workforce.

For policymakers, the way forward is clear but capital-intensive and that is, invest in sovereign IP cores, fund open-source EDA toolchains, and prepare for local fabrication and memory ecosystems. Only by securing the full silicon value chain, they say, can India transform from a global back office into an end-to-end AI powerhouse.

Practical applications will be crucial to winning public support. Abhishek Singh, Additional Secretary at MeitY and CEO of IndiaAI Mission points to Bhashini, India’s natural language processing platform, which enables speech-to-speech, text-to-text, and speech-to-text translation across all Indian languages, speaking at the BT India@100 event. The vision is bold, for instance, a woman in a remote tribal village could call a toll-free number, explain a health problem in her native tongue, and receive accurate medical guidance instantly. “These are the kinds of solutions that make sovereignty tangible,” Singh says.

The dream of a sovereign AI stack is bold, necessary, and, by most accounts, still far away. But the stakes like economic, strategic, and social are too high to ignore. In the race to define the AI future, India can’t afford to be a passive consumer. It must build, own, and lead.