India needs homegrown foundational AI models to avoid US-China gap: Bernstein
Synopsis
Global brokerage Bernstein is sounding an alarm that mainstream coverage has underplayed: India's AI gap isn't just technical — it's geopolitical. With Anthropic already restricting frontier model access for non-US citizens, the report argues India's entire AI future could be hostage to foreign policy decisions unless it builds foundational models now, not later.
Key Takeaways
Bernstein report warns India risks operating AI systems 'one or two generations' behind the US and China without domestic foundational model development.
Anthropic's restriction of frontier models for non-US citizens is cited as proof that AI access is governed by geopolitics, not free markets.
Advanced LLMs are increasingly treated as strategic national assets — comparable to nuclear energy , defence systems , and semiconductors .
India's historical focus on IT services rather than consumer internet platforms left it without the large proprietary datasets needed to train frontier AI models.
Bernstein identifies healthcare , industrials , and defence as sectors where India can build domain-specific LLMs using proprietary data.
Continued reliance on foreign AI models risks ceding control over a critical layer of future technology, the report concludes.
India must develop its own foundational artificial intelligence (AI) models to avoid falling critically behind the United States and China, according to a new report by global brokerage Bernstein released on 22 June. The report urges Indian policymakers to prioritise domestic large language model (LLM) development, warning that geopolitical forces — not market dynamics — now govern access to frontier AI.
The Geopolitical Risk to AI Access
The Bernstein report points to a concrete recent example: Anthropic's frontier AI models were restricted for non-US citizens, demonstrating that access to cutting-edge AI systems cannot be assumed. The brokerage warns that if India continues on its current path, Indian firms could find themselves operating with systems 'one or two generations' behind those available to competitors in the US and China.
Even companies with deep engineering talent would struggle to build competitive products if they are forced to rely on older AI architectures while global rivals access the latest systems. 'AI is the next fighter jet,' Bernstein stated, framing advanced LLMs as strategic national assets rather than commercially available software.
A Historical Pattern India Cannot Afford to Repeat
Bernstein draws a direct parallel between frontier AI and technologies such as nuclear energy, defence systems, and semiconductors — all of which saw access shaped by geopolitics rather than open markets. The report argues that India risks ceding control over a critical layer of future technology if it focuses solely on AI application development while depending on foreign foundational models. According to the brokerage, the country's AI future cannot be built on 'borrowed models.'
This comes amid a broader global race in which the US and China are investing heavily in sovereign AI infrastructure, while India has so far concentrated on services-layer and application-level development.
Why India Fell Behind on Foundational Platforms
The report attributes India's lag in foundational AI to its historical technology ecosystem, which was driven by IT services rather than consumer internet platforms. Unlike the US or China, India did not develop large proprietary datasets at scale — a prerequisite for training frontier models. This gap, the report notes, led policymakers and industry leaders to argue that India should focus on AI applications rather than foundational development.
Notably, foundational models require massive, domain-specific proprietary datasets to achieve competitive performance — something India's services-oriented tech industry has not historically generated at the required scale.
Where India Can Build an Advantage
Despite the structural challenge, Bernstein identifies a viable path forward. India can develop specialised, domain-specific LLMs using proprietary datasets in high-value sectors including healthcare, industrials, and defence. These verticals offer India both unique data assets and strategic incentive to build sovereign AI capability.
As global AI governance frameworks tighten and frontier model access becomes increasingly politicised, India's window to establish foundational AI capacity — rather than perpetual dependence — may be narrowing faster than policymakers have acknowledged.
Point of View
Not an anomaly; as AI becomes embedded in defence, healthcare, and financial infrastructure, dependence on foreign foundational models is a sovereignty risk, not merely a competitiveness one. India's IT-services heritage gave it global scale but left it without the consumer data flywheels that powered US and Chinese LLMs. The window for catching up on foundational models is not permanently open — compute costs are rising, talent is consolidating around a handful of frontier labs, and geopolitical AI blocs are hardening. A domain-specific LLM strategy in healthcare and defence is a credible starting point, but it requires policymakers to move from applauding AI adoption to mandating data infrastructure — something no Indian policy document has yet done with teeth.
NationPress
22 Jun 2026
Frequently Asked Questions
What does the Bernstein report say India must do on AI?
The Bernstein report urges India to prioritise building its own foundational AI models rather than depending on foreign systems, warning that geopolitical restrictions could leave Indian firms using AI that is one or two generations behind global competitors. It argues that India's AI future cannot be built on 'borrowed models.'
Why is relying on foreign AI models a risk for India?
Access to frontier AI is increasingly shaped by geopolitics rather than open markets, as demonstrated by Anthropic restricting its models for non-US citizens. If India depends on foreign foundational models, it could lose access during geopolitical tensions, leaving its tech sector at a structural disadvantage.
What sectors can India focus on for domestic AI development?
Bernstein identifies healthcare, industrials, and defence as sectors where India can develop specialised, domain-specific large language models using proprietary datasets. These verticals offer both unique data assets and strong strategic incentive for sovereign AI capability.
Why has India lagged in building foundational AI models?
India's technology ecosystem has historically been driven by IT services rather than consumer internet platforms, which meant it did not accumulate the large proprietary datasets required to train frontier AI models. This led policymakers and industry to focus on AI applications rather than foundational development.
How does Bernstein compare AI to other strategic technologies?
Bernstein draws a parallel between advanced AI and technologies such as nuclear energy, defence systems, and semiconductors — all of which became subject to geopolitical access controls rather than free-market availability. The brokerage argues frontier LLMs are following the same historical trajectory.
