Nandan Nilekani might’ve been the architect of India’s digital stack, but his stance on AI hasn’t gone down well with social media users.
In a recent op-ed in The Economic Times, the Infosys co-founder and former UIDAI chairman argued that India doesn’t need to win the race to build the world’s most advanced AI models. Instead, he believes India’s real opportunity lies in diffusing AI broadly — across its millions of micro-enterprises, farms, and small businesses — to enhance human productivity rather than replace it.
What Nilekani Actually Said
Nilekani’s argument is built on a few interlocking ideas.
First, he draws on economic history: the Industrial Revolution generated enormous productivity gains in Britain, but workers didn’t benefit for nearly a century. Productivity gains, he argues, don’t automatically translate into shared prosperity. Without deliberate policy, gains get privatised while the costs of transition are borne by the public.
Second, he makes a structural argument about India’s economy. In the West, AI is largely being deployed inside large firms to cut headcount. India’s economy looks different — millions of small enterprises, hybrid livelihoods, multi-tasked workers. Nilekani argues this informality, long seen as a weakness, could become an advantage. Small-scale, judgment-driven work is harder to automate and easier to augment.
Third, he lays out three priorities for India: reinventing the mobility ladder for young workers displaced from IT and adjacent sectors; building AI as public infrastructure — open, interoperable platforms in Indian languages, modeled on what UPI did for payments; and enabling mass entrepreneurship by lowering the barriers to starting and scaling a business.
His bottom line: India doesn’t need to lead in building frontier AI models. It needs to lead in ensuring AI’s benefits are widely shared.
The Reaction
Nilekani’s post on X drew plenty of negative reaction on X, with more quote posts than reposts, nearly all of which were negative. “Nandan, I love you for what you have done. But, your ideas are outdated. You have typical IT services sales guy mindset. Nothing wrong with that. That greatly helped in India in 1990s and 2000s, but it will lead it in a ditch in rest of 21st century,” wrote an X user.
“I strongly disagree with this. And I cannot believe how @NandanNilekani can be so short-sighted. Data and intelligence sovereignty should be a national strategic priority. Using AI models from foreign players will involve perpetually sending them all of our data,” wrote another.
“The power of an entire nation depends on building frontier technology. This is very beta thinking. I sincerely hope students especially do not follow this. Aim high. Be the person that lands on Mars, creates new drugs and build that frontier technology. Don’t live a half life!” wrote another user.
“Nandan is wrong. India should at the very least *attempt* to lead the world in building the most advanced models. It is very difficult to distribute the benefits of something you don’t have!” said another user.
And some reactions were quite brutal. “aim for the stars land on the moon” here we have “tech luminaries” asking us to aim for sony signal,” said a user, referring to a popular intersection in Bangalore.
Why Nilekani Could Be Wrong
The criticism isn’t baseless. There are real reasons why ceding the foundation model race could cost India more than Nilekani acknowledges.
Sovereignty is not just a talking point. When a country doesn’t control the underlying model, it doesn’t control the data, the guardrails, or the priorities baked into that system. OpenAI’s IndQA benchmark — designed by a US company to become the de facto standard for evaluating AI performance across 12 Indian languages — illustrates the point precisely. If India doesn’t build its own foundational systems, it risks having its own cultural and linguistic narratives defined by foreign labs.
The commoditization argument may be premature. Nilekani has previously suggested that frontier models will commoditize, making it smarter to build applications on top of them. Former Infosys CEO Vishal Sikka disagrees sharply: “It is a very, very bad idea to abandon that and let other people build. We have to build our own foundation models.” Sikka’s point is that if future technologies are built on top of these foundational models, India would be building its future on someone else’s foundations.
India is already on the board — and pulling away is the wrong move. Sarvam 105B, trained from scratch in India using government-provisioned GPUs, has placed India sixth in the Artificial Analysis Intelligence Index. That’s ahead of Israel and Switzerland. It’s a legitimate start — and precisely the kind of infrastructure you don’t rebuild from scratch after abandoning it.
The application layer isn’t as safe as it looks. Nilekani’s “build on top” strategy assumes the application layer is where durable value accrues. But history suggests otherwise — the companies that control the platform tend to extract the most value over time. When Reliance launched its AI unit and partnered with Google and Meta, it was essentially ratifying a world where India’s largest conglomerate relies on American-made models. That’s a significant strategic concession dressed up as pragmatism.
The mobility ladder argument cuts both ways. Nilekani is right that India’s IT sector and entry-level knowledge workers face real disruption. But his proposed solution — reskilling, job-matching platforms, AI diffusion — doesn’t require abandoning model-building. Those two tracks can run in parallel. Countries like France (Mistral), South Korea (EXAONE), and UAE (Falcon) have managed to pursue both. There’s no law that says you have to choose.
Nilekani’s framework is coherent and his instincts about India’s informality-as-strength are genuinely interesting. But the conclusion — that India should be a consumer of frontier AI rather than a builder — may end up looking like the digital equivalent of outsourcing your semiconductor supply chain and then wondering why you can’t get chips during a crisis.
The real AI divide, as Nilekani himself writes, will be between societies that diffuse AI broadly and those that let it concentrate. But diffusion built on foreign-controlled foundations isn’t sovereignty — it’s dependency at scale.