Dear AI Founders from India: Do NOT listen to the Nilekanis of the world.
The next decade of Indian tech must shape global AI, not just serve it.
PS: About the title: I have huge respect for Nandan Nilekani - but honestly, I’d expect him to motivate young founders to think different and bigger..and not give in to the existing reality. Read on..👇
I have been fairly vocal about the lack of depth in Indian AI products and here is all I am seeing these days:
Infy’s Nandan Nilekani has been repeatedly saying that India should not worry about building LLMs
CEO of TCS, India’s largest IT services company recently said that there is no huge advantage in building our own LLM
The most important point to be noted here: They both come from IT services background, where a service is delivered depending on client’s requirements.
Logically, if you look at it - they are right. There is NO point in building India’s own LLMs. Apart from all the exceptional reasons, India lacks AI research talent and we are better off building AI applications (rather, services).
Service companies excel at executing client requirements.
Product companies solve problems.
Going by the same logic, there was no point in attempting to build India’s own ride hailing service or food delivery app or even local mapping solutions.
Ola, Swiggy, and MapmyIndia et al succeeded because they addressed India’s unique challenges—dense cities, fragmented supply chains, and infrastructure gaps. They proved that local problems demand local solutions.
The thing is that most IT services companies do NOT appreciate the value of building product/tech centric businesses.
Most of them want to serve clients, but not customers.
They excel at collecting client requirements—not imagining solutions for unstated needs.
Given that gentlemen like Nandan and TCS CEO etc have the loudest voice in the room, it is but natural that it will impress many young founders.
A note to ProductGEEKs out there
Ignore these gurus - they might be right (in the short-term), but your future depends on them being wrong.
Here’s why their logic fails India’s future
Foreign LLMs like GPT-4 are not trained on India’s 22 languages, regional dialects, or cultural contexts. Blindly adopting them risks excluding millions from healthcare, education, and governance. Imagine a chatbot failing to understand Tamil slang or misinterpreting Hindi idioms in legal documents.
India’s diversity cannot be an afterthought.Talent shortages are real, but fixable.
Researchers leave because India lacks ambitious AI projects, not capability. China’s Deepseek shows the path: invest in homegrown models, and talent follows.Initiatives like Aadhaar and UPI scaled because India prioritized sovereignty. AI needs similar ambition—public-private partnerships, open-source frameworks, and incentives for compute infrastructure.
Critics claim building LLMs is costly.
But the cost of dependence is way higher.Foreign models = foreign biases.
They dictate economic value and expose India to security risks (AI is a geopolitical thing as well..like it or not).
Last, but not the least - what is Indian government going to do with 10,000 GPUs? Run OpenAI? Or Llama? We need at least 10 foundation models in India - given our scale and diversity.
Time to build in India is NOW.
India’s IT giants may prefer outsourcing AI, but startups and policymakers must think bigger.
One DeepSeek from China has shaken up the entire Silicon Valley - we need to ask : why did we not think of this?
To founders and leaders out there:
The next decade of Indian tech must shape global AI, not just serve it.
If you are building ambitious AI products / infra business from India - feel free to write to me with details. I will be happy to help/connect with the right folks (my email: ashish [at] nextbigwhat dot com]