India need not build a foundational AI model if we can figure out the tools that the LLM engine should power, writes Satyen K. Bordoloi
It is Mannheim, Germany, January of 1886. A three-wheeled, bulky contraption sputters down a cobbled street at ten kilometres an hour. Horses are terrified, adults are amused, and kids are delighted to see it. The man driving it is Karl Benz, who, thanks to work done in different parts of Europe in the last decade, finally has a patent (number 37435) on the first ICE or internal combustion engine. Germany had done it. Germany had won the future. Now it’d move ahead of the rest of the world. Literally and figuratively.
Except that’s not quite how it played out, is it? You’d laugh if I asked whether the automobile industry stayed in Germany. Because in under three decades, Henry Ford would build an entire industry around it in Detroit. About six decades later, Toyota would humiliate Detroit so thoroughly that the US government had to bail it out. In the middle of all this, the South Koreans would come out from nowhere with a Hyundai and conquer the world.
India, not to be left behind, saw Tata and Bajaj sell both in the subcontinent and in Africa. What about Germany? It found its swagger with BMW and Mercedes but never conquered the automobile world. That’s because today, there’s not a country of a decent size and economy without its own motor industry. Benz might have owned the patent. But nobody could own what came after.
And the world was merely following tradition. Go back further, and you find that James Watt might have improved the steam engine in 1769 – didn’t invent it, improved it – and made good money from it, but the actual fortune was made by the mills, the foundries, the railways, all the industries that took his engine and aimed it at their problems. The steam engine ushered in the modern world because, for the first time in our history, it proved that you don’t win an industrial revolution by building the engine but by figuring out what to do with it.
Apply the same logic to Artificial Intelligence, and you’ll find a lot of the expert advice, especially about how India has missed the boat, melt away.
Nine Years Old, and India Is Already Being Written Off
The transformer idea – the architecture behind every LLM, image generator and code assistant today – turned nine on June 12. Google published “Attention Is All You Need” in 2017, famously sat on it, only for others – especially OpenAI – to do the hard work of turning it into something the world wanted. Then everyone piled in: Meta, Anthropic, Mistral, even Google, till eventually Chinese labs moved so fast on the tech four years ago that “frontier model” no longer automatically means Silicon Valley. Today, open-source versions of LLMs overpopulate digital space. are everywhere. The engine is out.
And nine years in, India hasn’t built one that’s even a scratch on the US, European or Chinese models. Naturally, that became a reason to write us off. Take Bloomberg, which published a piece in May that stung precisely because it was right: India is one of the biggest losers in the great AI-driven reshuffle of global investment. India’s stock market is on the verge of dropping out of the world’s top five for the first time in years.
The foreign money that once treated India as nearly the equal of China in emerging-market portfolios has moved on to chase chips, compute and model-building, none of which India is in the business of. Foreign institutional investors have pulled roughly $27.6 billion out of Indian equities since January. The verdict of global capital markets is clear: India missed the boat, and they are punishing us for it.
Seems fair, at the face of it. The problem is that assessment is caught up in the present, which is but a moment. But the past and especially the future are a long time to make knee-jerk reactions without understanding how things shape up and pan out.
Hence, I think that Bloomberg, and indeed a lot of global verdict, are measuring the wrong race.
The Scoreboard Nobody Is Watching
Bloomberg and the FIIs are looking at one scoreboard: who built the engine. On that one, India doesn’t even make the knockout rounds. Fair enough.
But there is another scoreboard – who is actually using the thing — and on that one the picture looks very different. India and China, not the US, are the two countries putting AI to work at genuine population scale right at this moment. Around 57% of India’s working population and 58% of China’s are already using AI in some form or the other. The United States, for all its trillion-dollar valuations, congressional hearings and magazine covers, ranks 24th in the world with only 28% adoption on that same measure. Isn’t it shocking that the people who built the engine are, oddly enough, not enthusiastic about driving one?
And what are Indians using it for? Not writing college essays. Okay, maybe a few of them are. Okay, a lot more than a few. But we in India have something like the government of India’s (GoI) Kisan e-Mitra chatbot to answer farmers’ questions about welfare schemes in their own languages.
Take CropIn, a Bengaluru-based startup most people outside agritech haven’t heard of, that has digitised farmland in over a hundred countries and is working with seven million farmers, because its tools were designed for fragmented, low-connectivity, price-sensitive agriculture, aka most of Asia and Africa. Take GoI’s frameworks like SAHI to integrate AI safely, ethically and inclusively into India’s massive, chaotic public health system.
These and hundreds of thousands of other initiatives in India, both by the government and the private sector, are creating AI solutions for problems Silicon Valley hasn’t even heard of. By doing so, they’re reshaping the market for every vendor across South and Southeast Asia. None of this requires India to build GPT-5. It requires India to understand its own problems better than anyone else.
To do that which is India’s most innovative word on innovation: jugaad. Strip away the mythology, improvise on what you have till you build solutions calibrated to constraints that the Global North is ignoring because they’re busy serving richer clients.
Why the Global South Will Buy Indian
Here is the business case hiding in plain sight. A hospital director in Lagos, a farmers’ cooperative in Jakarta, a rural school district in Kenya – none of these people needs the world’s most powerful AI model. What they need is the cheapest one, but one that actually understands their situation. Western models are built for English, broadband and enterprise procurement budgets. Indian models – built for multiple languages, for 2G, for price points that make no sense in California – fit the ground realities of the Global South.
China figured this out years ago. Chinese tech’s march through Africa and Southeast Asia was paved by tools that were cheaper and close enough, not smarter and expensive. India is beginning to do this and will be arguably better in the markets that share its languages and legal traditions. Do the arithmetic on how many hospitals and cooperatives exist outside the G7. It is the larger market, uncontested, because no one in San Francisco is even thinking about it, forget designing for it.
The Talent Sitting in Plain Sight
India has one more advantage that Bloomberg manages to mention and undercount at the same time: engineers. Huge numbers of them, most of them educated in English, which happens to be the operating language of every large language model that matters.
It is not an accident that Satya Nadella turned Microsoft into the AI company it is today or that Sundar Pichai has taken Google and Alphabet beyond Page and Brin’s wildest imagination, or that Arvind Krishna – an IIT Kanpur alumnus – runs IBM. These are not aberrations. They are the predictable output of a country that has been mass-producing engineers for decades and then, for want of opportunity at home, exported them to run everyone else’s technology companies.
India’s problem is not talent. It is capital and compute – real problems, yes, but finite ones, soluble ones, far smaller than the ones the headlines describe. 47% of Indian enterprises are already running multiple AI use cases. AI investment hit $3.94 billion in a single quarter this year. The building is happening in India. It just isn’t happening in the one category – foundation model-building – that earns you a line in the Bloomberg story.
What India Actually Needs
A Ministry That Does One Job: Here is what I’d propose, and I’m aware it will sound hopelessly optimistic given what Indian ministries typically accomplish: a dedicated Ministry of Artificial Intelligence Uses and Businesses.
It shouldn’t be a ministry to build a foundational model, something that the government is spending too much to promote right now. Not a ministry to encourage those who want to compete with NVIDIA on chips. No. This should be a ministry with a single, precious mandate: to make sure India’s application-layer advantage – all those CropIns and Kisan e-Mitras that are still running on a founder’s laptop with no government customer and no export channel, doesn’t dissolve into another decade of pilots and bureaucratic red tape that never scale.
Push procurement reform so the government becomes a real customer. Build export corridors so Indian health-AI can reach Lagos as easily as Lucknow. Create skilling pipelines that turn engineering numbers into products, not just outsourced bodies doing someone else’s back-end work.
India did something like this with software exports in the 1990s – identified an advantage, built institutions around it, and created an IT sector that employs fifteen million people today. The window was clear then. Almost nobody in policy saw it. Are they seeing the same in AI today?
Nobody remembers which workshop in Detroit or Nagoya first bolted Benz’s patent onto four wheels and made something that people actually bought. What they remember are Ford, Toyota, Hyundai – names that took someone else’s invention and built industries, entire national identities around what the thing could do once it was in the right hands aimed at the right problem.
India doesn’t need to be Karl Benz. Bloomberg is right that the foundational model window is shut. But as the road Benz’s engine paved, the road of AI application is empty. It is waiting for those hungry enough to fill it with a hundred million small, useful, unglamorous solutions. That’s where India can win the LLM race, not the AI arms race, but the AI peace fixture. And we’ll do the rest of the world, particularly the Global South, a great favour by doing exactly that.
In case you missed:
- OpenAI, Google, Microsoft: Why are AI Giants Suddenly Giving India Free AI Subscriptions?
- Indian AI’s Power Problem: We Woke Up Late to AI, We’re Still Half-Asleep
- GoI afraid of using foreign GenAI in absence of Indian ones? Here’s how they still can
- The 3-Hour Countdown: India‘s New AI and Deepfake Rules Spark a Free Speech Firestorm
- 5.6 Cr Pending Cases: How India Is Letting AI Slip Through Even As World Courts Revolution
- How Can Indian AI Startups Access Global VC Funds?
- How India Can Effectively Fight Trump’s Tariffs With AI
- A Small LLM K2 Think from the UAE Challenges Global Giants
- South Korea Passes World’s First AI Laws: Visionary or Foolish?
- $725 Billion Down, Trillions To Go: Why The AI Bubble Refuses To Burst
