Breaking India Should Shun Costly LLM Race, Says Fintech Veteran Mohandas Pai

By Simran Thakur

New Delhi — At the launch of ToneTag’s voice‑first merchant‑banking platform eKosha on Wednesday, Mohandas Pai — former chief financial officer of Infosys and current chairman of the Confederation of Indian Industry (CII) — warned that India risked “wasting resources” by chasing large‑language‑model (LLM) development on a scale comparable to that of the United States and China. Pai urged policymakers, industry bodies and venture capitalists to redirect attention toward “practical AI models” that address immediate business needs such as fraud detection, credit underwriting and voice‑based payments, rather than pouring capital into “expensive, compute‑intensive” LLM projects that may not generate domestic economic benefits.

What happened
During the inaugural event for ToneTag’s eKosha platform, Pai highlighted the firm’s voice‑biometric solution as a concrete illustration of the type of AI application he believes India should prioritize. He argued that the country already possesses a “huge talent pool” and a robust data ecosystem capable of supporting narrower, domain‑specific AI solutions. “We should focus on building AI that solves real problems for Indian businesses, especially small and medium enterprises, instead of chasing a global race that may not align with our economic realities,” Pai said.

Why it matters
Pai’s comments arrive at a moment when Indian startups are increasingly announcing home‑grown LLM initiatives and the central government is drafting a national AI strategy. The tension between ambitious, globally‑oriented research and immediate, market‑driven applications raises questions about the most effective allocation of limited public and private funds. If India were to invest heavily in LLMs without a clear commercial pathway, the opportunity cost could be significant: resources might be diverted from solutions that directly improve financial inclusion, reduce fraud in digital payments, or streamline credit assessment for micro‑enterprises.

Background and context
India’s AI policy framework, articulated in recent government releases, emphasizes talent development, research capacity and the creation of a supportive regulatory environment. At the same time, the country’s fintech sector has expanded rapidly, with voice‑based authentication emerging as a promising tool for reaching underserved merchants. ToneTag’s eKosha platform, which enables merchants to authenticate transactions through voice biometrics, exemplifies the “practical AI” approach advocated by Pai. By leveraging existing financial infrastructure and the country’s multilingual voice data, eKosha aims to improve security and inclusion without the massive compute requirements associated with training large language models.

Competing claims and uncertainty
Proponents of LLM research argue that large‑scale models can generate broad, cross‑domain capabilities that may eventually lower the cost of developing specialized AI tools. They point to the rapid progress made by leading AI labs in the United States and China, suggesting that participation in the global LLM ecosystem could position Indian firms to benefit from downstream innovations. However, Pai and other Indian technologists caution that the current cost structure of LLM training — requiring extensive GPU clusters, high energy consumption and substantial data‑center investments — may be misaligned with India’s economic realities. The debate therefore centers on whether the potential long‑term gains from LLM research outweigh the immediate opportunity costs of not deploying sector‑specific AI solutions that can be commercialized quickly.

What to watch next
Stakeholders will be monitoring several developments:

1. Policy direction – The Indian government’s forthcoming AI strategy will likely clarify funding priorities, possibly delineating separate streams for foundational research versus industry‑focused applications.
2. Investment patterns – Venture capital allocations to AI startups may shift if leading industry voices, such as Pai, influence sentiment toward domain‑specific models.
3. Corporate adoption – The uptake of eKosha and similar voice‑biometric platforms by merchants and payment aggregators will provide early evidence of the commercial viability of “practical AI.”
4. International collaboration – Partnerships with foreign AI labs could offer a middle ground, allowing Indian firms to access LLM capabilities without bearing the full cost of model development.

Conclusion
Mohandas Pai’s cautionary remarks underscore a strategic crossroads for India’s artificial‑intelligence ambitions. While the allure of competing in the high‑profile LLM race is strong, the immediate needs of India’s vast SME sector and the country’s burgeoning fintech ecosystem may be better served by targeted, cost‑effective AI solutions. The balance between long‑term research aspirations and short‑term economic impact will shape how India allocates its AI resources in the years ahead.

Sources

– The Hindu, “India should not chase costly LLM race; instead it should focus on practical AI models: Mohandas Pai,” https://www.thehindu.com/news/national/karnataka/india-should-not-chase-costly-llm-race-instead-it-should-focus-on-practical-ai-models-mohandas-pai/article71175175.ece

Story synopsis gathered from: The Hindu – National — source