India Built the AI. We Are Using It to Read Our Stars.

There is a comfortable lie going around that India has produced nothing in artificial intelligence. It is wrong, and it is worth correcting before we get to the part that should actually worry us.

In February 2026, a Bengaluru company called Sarvam released two large language models built from scratch, trained to understand all 22 official Indian languages. The government’s IndiaAI Mission is funding more: BharatGen, Soket AI, Gan AI, and others, each chasing a piece of the problem. Even Krutrim, the venture from Ola’s Bhavish Aggarwal that the internet loves to mock, stopped pretending it would out-build OpenAI and turned itself into a cloud and infrastructure business that earned ₹3 billion last year. So the engines exist. India can now say it makes its own AI.

That is the good news. Here is the uncomfortable news.

Walk through the Indian app stores. Look at what ordinary people actually open every day. The breakout consumer AI hit in this country is astrology. Apps that read your birth chart, predict your marriage, time your business launch by the position of planets. These products have millions of users, real paying customers, and growing revenue. They are, by the cold logic of the market, India’s most successful application of artificial intelligence.

Sit with that for a moment. We taught machines to read 22 languages. We are using them to read fate.

This is not a joke about superstition. It is a question about ambition. A technology is defined not by who builds it but by what a society chooses to do with it. The United States built the same kind of models and pointed them at drug discovery, weather prediction, legal research, and code. China pointed its models at manufacturing, surveillance, and logistics. India pointed its best consumer engineering talent at telling a nervous 24-year-old whether Saturn is blocking his promotion.

Why does this happen? Because astrology is a perfect business. The customer is anxious, the answer can never be proven wrong, and the person comes back every time life feels uncertain, which in India is often. It is easy money. Real problems are hard money. A model that helps a farmer in Vidarbha read a soil report, or helps a weaver in Kerala price his cloth fairly, or helps a sick villager understand a prescription, those products take years, demand trust, and pay slowly. Predicting a wedding date pays today.

So the talent follows the easy money. This is the actual failure, and it is not a failure of technology. It is a failure of where we choose to spend our cleverness.

There is a deeper cost here too. Astrology AI does not just waste engineering. It sells certainty to people who deserve agency. A young person who is told the stars will fix his career is a young person quietly taught that effort is secondary to fate. Scale that across a hundred million users and you are not running a business. You are training a nation to wait instead of act.

Compare this with what the same tools could do. The single largest thing holding back the average Indian is not a lack of horoscopes. It is the inability to read a government form, fight a wrong electricity bill, understand a loan document, or get a straight medical answer in his own language. These are exactly the problems a multilingual Indian model is built to solve. The engine is sitting right there. We are using it for entertainment because entertainment is where the quick rupee lives.

None of this means the model-builders failed. Sarvam and the others did the hard part. They proved India can sit at the table. The question now passes to everyone else, the founders, the investors, the buyers. Will we build AI that makes Indians more capable, or AI that makes them more comforted?

A country that can build its own intelligence and chooses to spend it on astrology has not solved its real problem. It has only found a more advanced way to avoid it.

The machines are ready. The question is whether we are.

Comments

comments