Can China's DeepSeek be a catalyst for driving India’s AI ambition?

Chinese artificial intelligence (AI) startup DeepSeek has shaken the global AI landscape by recently launching R1 LLM which has outperformed OpenAI's ChatGPT and other advanced LLMs on most benchmarks. This in turn has triggered a global selloff in AI stocks due to fears of heightened competition and investment scrutiny.

This development partly reflects a geopolitical response to US export restrictions and the cost-efficiency of this AI model for just $5.6 million — a fraction of the $100 million plus budget typical for industry leaders like OpenAI, underscores the necessity for smarter, not just larger, AI investments — and prompt experts to believe if DeepSeek's success may prompt India to consider developing its own foundational AI models.

To be sure, India has also created a few LLMs albeit on a smaller scale. Indian startups like Sarvam and Krutrim AI are making notable progress in developing LLMs tailored to the local market's diverse languages. However, DeepSeek's model boasts 671 billion parameters compared to Sarvam-1’s two billion. Hence it is imperative for India to have multiple players in the AI sector.

Deepak Pareek, founder of HTPL, a technology consultancy firm, believes DeepSeek's innovations lower barriers for advanced AI, presenting significant opportunities for Indian entrepreneurs to develop cost-effective LLMs for various applications, such as agriculture.

Pareek suggests in a LinkedIn post that Indian firms can collaborate with DeepSeek and American AI companies to create niche solutions that combine global innovation with local needs. By utilising DeepSeek's open-source tools for bias detection, Indian startups can address issues like caste, gender, and regional bias. Projects like OpenNyAI, designed to simplify legal processes, illustrate how thoughtful design can serve as a competitive edge, he said.

Other challenges include fragmented data, skill shortages in advanced machine learning, and inconsistent funding that prevent Indian companies from realising their AI ambition. For example, building sovereign AI infrastructure is crucial, as India cannot indefinitely rely on foreign cloud services. “Establishing public-private partnerships to create data trusts or shared data hubs or data centres could foster collaboration while safeguarding privacy,” he wrote.

Experts also believe India should promote the creation of Indic LLMs that reflect its linguistic and cultural diversity, leveraging cost-effective frameworks like the DeepSeek model to succeed in limited budgets.

Another need of the hour is upskilling the workforce for an AI-driven economy. According to Dhriti Prasanna Mahanta, Vice President & Business Head at TeamLease said that currently, only 2.5% of Indian engineers possess AI-related skills. “Integrating AI literacy into national school curricula can lay a strong foundation while aligning skill development initiatives with national programs like ‘Make in India’ can position India as a leader in AI-driven solutions,” said Mahanta.

Specialised training programs at Indian Institute of Technology (IITs) and Indian Institute of Management (IIMs) should focus on advanced machine learning, while collaborations with online platforms and DeepSeek's resources can facilitate upskilling across sectors.

Varun Aggarwal, founder of The Change Engine, noted on the blog the need for India to enhance AI research and develop foundational models, supported by a robust university R&D ecosystem.

The challenge for enterprises goes beyond cost optimisation, as experts believe they must establish governance and validation processes to ensure their AI solutions meet performance, security, and ethical standards. Analysts at Motilal Oswal Financial Services (MOFSL) suggest that shifting to more affordable AI infrastructure could create new opportunities for Indian IT firms.

Major Indian IT firms like TCS, Infosys, HCL Tech, Wipro, and Tech Mahindra are enhancing their AI capabilities. TCS highlighted "Agentic AI" in its Q3 results, noting its role in improving efficiency and reducing costs. Infosys is developing over 100 Generative AI agents to boost automation and productivity. HCL Tech reported an 85% drop in costs for large language models since early 2023, making AI adoption more feasible. Wipro is leveraging Agentic AI to improve customer service and supply chain processes, while Tech Mahindra focuses on creating smaller, enterprise-specific AI models.

These AI models could be game-changers for enterprise AI applications, which will align with IT firms' growing focus on cost-effective AI solutions, believes MOFSL analysts.