India risks becoming permanently dependent on foreign artificial intelligence (AI) models unless it develops its own sovereign large language models (LLMs), according to a report by global brokerage Bernstein. The report argues that while building AI applications and data centres has value, India's long-term competitiveness depends on owning the foundational AI models that power them.

The report comes amid growing global restrictions on access to advanced AI technologies. Bernstein says recent US curbs on frontier AI models demonstrate that cutting-edge AI is increasingly being treated as a strategic national asset rather than a freely available technology, raising concerns for countries relying on foreign platforms.

AI needs its own 'DeepSeek moment'

Bernstein argues that India cannot build its AI future on borrowed technology. It likens today's AI race to the evolution of defence technology, where access to advanced capabilities is often tightly controlled.

The report warns that India could eventually find itself locked out of the latest AI models during geopolitical tensions, leaving domestic companies and institutions dependent on older technologies while competitors in the US and China continue to advance. It describes AI as the next "fighter jet"—a strategic capability that nations are increasingly unwilling to share without restrictions.

MUST READ: OpenAI's next flagship AI model faces launch delay; White House demands safety checks

India needs AI sovereignty

According to Bernstein, India's focus should extend beyond developing AI-powered applications or chatbots. Instead, it argues the country must own the "intelligence layer"—the foundational LLMs that underpin enterprise software, defence systems, healthcare, finance and scientific research.

The report cautions that relying entirely on foreign LLMs exposes India to long-term strategic risks. If access to these models is restricted or delayed, Indian businesses may be forced to build products on older-generation AI systems, making them less competitive globally.

Services model may not work in the AI era

Bernstein also questions whether India's successful IT services model can be replicated in artificial intelligence.

MUST READ: India attracts AI billions, but HSBC's top investment picks are China, Japan and Korea; know why

Historically, Indian technology companies excelled by building software and services on platforms developed elsewhere. However, the report argues that AI shifts value creation towards ownership of foundational models rather than application-layer development. As a result, companies that depend entirely on foreign LLMs could lose their competitive advantage as AI becomes the core engine powering digital services.

Why India is behind

Bernstein says India's absence of a globally competitive LLM is largely structural rather than a deliberate policy choice. Unlike the US and China, India has historically built a services-led technology industry instead of large consumer internet platforms that generate the massive proprietary datasets needed to train frontier AI models.

The report also notes that India's AI Mission spreads resources across multiple priorities, including compute, research, startups and applications, leaving relatively limited funding for foundational model development. Unless India prioritises sovereign LLM capabilities, it risks remaining dependent on foreign AI providers for the core intelligence layer that will increasingly power businesses, governments and critical infrastructure.

MUST READ: ‘I handed Claude my salary’: Viral X post shares 7 prompts for financial freedom

A roadmap for India's AI future

Rather than attempting to compete head-on with the largest global AI companies, Bernstein recommends that India build sovereign, domain-specific LLMs using proprietary datasets in sectors such as healthcare, manufacturing, finance and defence.

The report also suggests policymakers consider encouraging localisation of AI infrastructure or incentivising domestic development of foundational models to reduce geopolitical dependence. Another option would be to require foreign AI firms to build and operate India-based AI stacks insulated from external geopolitical controls. While both approaches have challenges, Bernstein says they represent a trade-off between access to global AI systems and long-term technological autonomy.

India's digital public infrastructure, rich datasets and engineering talent provide a strong foundation, Bernstein says. But unless the country develops its own AI stack, it risks remaining one or two generations behind global leaders in AI innovation. Building an Indian equivalent of DeepSeek, the report concludes, is no longer just a technological ambition—it has become a strategic necessity for safeguarding India's competitiveness in the AI era.

MUST READ: ‘Felt like a CA sitting next to me’: Check how professional uses Claude AI to file ITR-1