During his visit to India last June, OpenAI CEO Sam Altman declared it “totally hopeless to compete with us on training foundation models,” bluntly advising Indian engineers not to try to compete with Silicon Valley. Nandan Nilekani has emphasised focusing on simpler efforts like computing infrastructure and AI cloud services to support long-term progress. NR Narayana Murthy expressed doubt about India’s current capability to build a large language model (LLM), stating that the lack of large databases and a structured problem-solving mindset are critical obstacles.

They are all underestimating India’s capabilities — as I have seen firsthand.

Consider Tech Mahindra’s Project Indus. CP Gurnani, who was the company’s CEO, turned Altman’s scepticism into a challenge. With less than $5 million, Tech Mahindra developed an Indian LLM capable of understanding and generating content in 40 local languages and dialects. Project Indus didn’t rely on brute financial force like Silicon Valley’s behemoth models. Instead, it embraced frugal innovation to address a critical gap in AI -- language accessibility for the billion-plus Indians who speak diverse tongues.

Gurnani’s success exposed the flaws in assumptions held by Altman and others. Building LLMs isn’t about throwing vast sums of money at the problem — it’s about focus, ingenuity, and leveraging local expertise. India’s track record of achieving the improbable — whether it’s the........

© hindustantimes