India needs its own DeepSeek to avoid AI dependence, warns Bernstein

A recent report by global brokerage Bernstein highlights a pressing concern for India regarding its reliance on foreign artificial intelligence (AI) models. The report warns that without developing its own sovereign large language models (LLMs), India risks becoming permanently dependent on external technologies. While the establishment of AI applications and data centers is beneficial, the long-term competitiveness of the nation hinges on possessing the foundational AI models that support these advancements. The urgency of this issue is underscored by increasing global restrictions on access to cutting-edge AI technologies. Bernstein points to recent U.S. measures against advanced AI models, indicating that these technologies are now viewed as strategic national assets, rather than universally accessible tools. This situation raises alarms for nations that depend on foreign platforms for their AI capabilities. Bernstein emphasizes that India cannot rely on borrowed technology to forge its AI future. Drawing parallels to the defense technology sector, the report suggests that access to advanced AI capabilities is becoming tightly controlled, potentially leaving India at a disadvantage during geopolitical conflicts. If such tensions arise, Indian companies could find themselves using outdated technologies while competitors from the U.S. and China continue to advance. The report also argues for a shift in focus for India. Rather than merely creating AI-driven applications or chatbots, there is a pressing need for the country to own the 'intelligence layer'—the foundational LLMs that are essential across various sectors, including enterprise software, healthcare, finance, and scientific research. Relying solely on foreign LLMs exposes India to significant long-term risks, especially if access to these models is restricted. Moreover, Bernstein questions whether India's established IT services model can effectively transition into the AI realm. Historically, Indian tech firms have thrived by developing software on foreign platforms. However, as AI technology evolves, the emphasis is shifting towards owning foundational models rather than merely developing applications. Dependence on foreign LLMs could diminish the competitive edge of Indian companies as AI becomes integral to digital services. The report attributes India's lack of a competitive LLM to structural issues rather than a deliberate policy. Unlike the U.S. and China, India has focused on a services-led tech industry instead of building large consumer platforms that generate massive proprietary datasets needed for training advanced AI models. Bernstein also critiques India's AI Mission for spreading resources too thinly across various goals, leaving foundational model development underfunded. To address these challenges, Bernstein proposes that India should focus on developing sovereign, domain-specific LLMs utilizing proprietary datasets in critical sectors like healthcare, manufacturing, finance, and defense. Policymakers might also consider incentivizing domestic foundational model development or requiring foreign AI companies to establish operations within India, insulated from external geopolitical influences. While these strategies may present challenges, they represent a necessary balance between accessing global AI systems and achieving long-term technological independence. Bernstein notes that India's robust digital infrastructure, extensive datasets, and skilled engineering workforce lay a strong foundation for these efforts. However, without a dedicated push to create its own AI stack, India risks falling behind global leaders in AI innovation. The report concludes that building an Indian equivalent of a sovereign AI model is not just a technological goal—it is a crucial strategy for ensuring the nation's competitiveness in the AI era.