New Tech Debate: Should India Build its Own LLM?

As the dust of the 90-hour-work-week debate has settled, the Indian tech ecosystem is now engrossed in a new conversation. Founders and executives of different companies are voicing their opinions on whether India should focus on developing its own vernacular large language models (LLMs) or should it focus on building applications on top of existing models to address real-world problems.

The debate started when Infosys co-founder Nandan Nilekani during the Meta AI summit in October said that Indian startups should avoid building their own LLMs. He added that India’s goal should not be to build one more LLM. India should let the big boys in the Silicon Valley do it by spending billions of dollars. He added that India must use these models to create synthetic data, build small language models quickly, and train them using appropriate data.

Aravind Srinivas, CEO of Perplexity AI, took to X to share his disagreement with Nilekani’s opinion, stating that India must focus on both model training and application development in order to fully benefit from AI developments.

Also Read: India Should Design LLMs That Are 10 Times Cheaper: Microsoft CEO Satya Nadella

The Argument

The question of whether India should develop its own Large Language Models (LLMs) has elicited diverse opinions from prominent figures in the technology sector.

Narayana Murthy, another co-founder of Infosys, has added on Nilekani’s stance against India developing its own AI models. According to Murthy, while it is commendable to work towards building large language models in vernacular languages, it is uncertain how effectively these models can be utilised until we can create extensive databases. Indian languages are still in their infancy compared to English, particularly in the fields of Science, Technology, Engineering, and Mathematics (STEM).

TCS CEO K. Krithivasan also believes that India doesn't need to build its own LLMs due to the high costs and complexities. Instead, he suggests leveraging existing frameworks and focusing on training models with vast amounts of data, which is a strength of Indian IT. He also sees AI as an opportunity to create jobs rather than eliminate them.

Counter Argument

On the other hand, Manish Gupta, Director of Google Research India, believes that India would benefit from building its own foundation models. He references Nilekani's Aadhaar project as an example of starting with foundational infrastructure before developing use cases, suggesting a similar approach for AI development in India.

Vinod Khosla, co-founder of Sun Microsystems and a prominent venture capitalist is one of the biggest proponents of building national AI models. He suggests that countries like India should develop their own AI models to ensure technological independence and equitable benefits.

He remarked, "Why would any reasonably sized country bet on a foreign country's AI (which can be revoked by sanctions!) for its national security? We invested in SarvamAI for India, and each country needs to encourage and train its own regional talent through incentivised development of technology."

Also Read: Reliance Unveils World's Largest Data Centre in Jamnagar, Post Trump's Stargate Reveal

Srinivas’ View

Srinivas claimed that India is making a strategic mistake by presuming that developing AI foundation models is prohibitively expensive. "Regarding the discussion over India training its foundation models, I believe India fell into the same trap that I did while running Perplexity. "Thinking models will cost a shit ton of money to train," he stated.

He wrote about India's ability to flourish in AI, drawing parallels with the country's achievements in space. Citing the Indian Space Research Organisation's (ISRO) capacity to reach big milestones with low funding, he suggested that India should apply similar ideas to AI development.

He also cited the progress of Chinese AI firm DeepSeek as an example of what can be accomplished with a focused approach to AI research.

These varied perspectives highlight the ongoing debate among industry leaders regarding India's approach to AI and LLM development.

Ambani Launches World’s Largest Data Center

Mukesh Ambani's Reliance Group is set to build what might be the world's biggest data center by capacity in Jamnagar, Gujarat. This move comes after a flurry of global investments to capitalise on increasing demand for artificial intelligence (AI) services.

This move by Ambani has presented a compelling argument for building homegrown large language models (LLMs), countering the views of Infosys co-founders Narayana Murthy and Nandan Nilekani.