India Slept Through the AI Alarm, and the Market Has Started Packing Its Bags
Acronyms used in this post:
AI — Artificial Intelligence, computer systems that can perform tasks involving prediction, language, pattern recognition, coding, classification, and decision support.
NSE — National Stock Exchange, one of India’s major stock exchanges.
IT — Information Technology, the broad business of software, systems, consulting, support, outsourcing, and digital operations.
GenAI — Generative Artificial Intelligence, AI that creates text, code, images, audio, or other outputs.
GPU — Graphics Processing Unit, the chip now heavily used for AI training and inference.
TSMC — Taiwan Semiconductor Manufacturing Company, the world’s most important advanced chip foundry.
R&D — Research and Development, the slow, expensive work of creating new knowledge, products, methods, and technologies.
GDP — Gross Domestic Product, the total value of goods and services produced in an economy.
GCC — Global Capability Centre, an offshore center where multinational companies run technology, finance, operations, analytics, design, and support work.
UPI — Unified Payments Interface, India’s real-time digital payments system.
India did not miss AI because Indians are stupid. India missed AI because India mistook clever clerks for industrial power, software service contracts for technological sovereignty, and English-speaking coding armies for ownership of the machine.
There. Now the room is awake.
I am writing this from the southern fringe of Calcutta, where the afternoon heat sits on the tin roofs like a fat landlord and the fan rotates above me with all the optimism of a defeated ceiling lizard. The lane outside is half dust, half drain water, and somewhere a man is shouting into a phone as if international finance will improve if only he increases his volume. I have tea going cold beside me. I have bills. I have age. I have the usual Bengali inheritance of intelligence, irritation, digestive trouble, and no useful capital.
And on the screen there is that headline: India missed out on AI, and now its run as a market darling may be over.
One wants to laugh.
Then one does not.
For thirty years India sold the world a magnificent story. We had engineers. We had English. We had mathematics. We had disciplined project managers who could turn a broken American insurance system or European bank migration into twelve thousand tickets, seventeen dashboards, and a weekly status call so long that even the chair lost faith in reincarnation. We built the outsourcing empire. We became the reliable back office of the planet. If the West had a digital mess, India had a floor of young people in ID cards ready to enter the mess with headphones, spreadsheets, Java, SQL, PowerPoint, and silent gastric suffering.
It was not nothing.
Let no fashionable fool say it was nothing.
That IT services boom put food on tables. It built apartments. It sent children to private schools. It made Mysore training campuses look like a clean republic from another species. It turned middle-class India from a nervous bicycle into a small hatchback with EMI payments and weekend pizza. It produced dignity, salaries, matrimonial confidence, airport lounges, and the first generation of Indians who could say “onsite opportunity” with the tenderness usually reserved for poetry.
But here is the catch.
The AI age is not rewarding who can supply labor.
It is rewarding who owns the bottleneck.
That is why Taiwan matters. That is why South Korea matters. That is why Nvidia sits in the global temple receiving incense from investors with trembling hands. That is why America’s hyperscalers can spend shocking sums on data centers and still be treated as if they are building the railways of the twenty-first century. The money is not merely chasing “technology.” That word has become too broad, like calling both a scalpel and a spoon “metal.” The money is chasing compute, chips, memory, power, cooling, cloud platforms, model distribution, and the ability to stand at the narrow bridge where everybody else must pay toll.
India, in comparison, built a very large bus stand.
Busy, useful, noisy, full of movement.
But not the bridge.
This is the dragon-breath part Indians do not like hearing: we became excellent at serving systems other people owned.
We customized. We maintained. We integrated. We tested. We migrated. We documented. We supported. We staffed. We optimized. We sat in calls at ungodly hours and said “let me circle back” until the phrase became a national minor art form.
But we did not own enough of the deep stack.
Not enough chips. Not enough foundational research. Not enough high-end hardware. Not enough operating systems. Not enough global platforms. Not enough serious scientific instrumentation. Not enough patient capital. Not enough universities that can breathe without political, bureaucratic, or funding asthma. Not enough laboratories where a person can fail for eight years and still be considered useful rather than defective.
We trained millions to become excellent passengers in someone else’s train.
Then AI arrived and changed the timetable.
You think AI is just another software wave?
Not quite.
The old software wave needed armies. AI eats armies. Slowly in some places, quickly in others, but the direction is not mysterious. A task that once needed twenty people may now need twelve. A coding team that once billed three thousand hours may now bill nine hundred. A support process that once justified a whole floor may now become a workflow with a human supervisor and a machine doing the drudgery like a tireless, humorless intern who never asks for Diwali leave.
This is not science fiction. This is pricing pressure.
The customer looks at the vendor and says, “But surely AI makes this cheaper?”
And the vendor smiles the smile of a man whose roof has just developed a leak directly over his wedding album.
Indian IT’s old model was built around effort. Bodies. Hours. Utilization. Bench strength. Pyramid structures. Many juniors, fewer seniors, one partner in a nice shirt saying “digital transformation” with priestly confidence. AI attacks that pyramid from below. It does not remove all work. Ignore that cartoon. Real enterprises are messy, regulated, brittle, political, and full of ancient systems that behave like retired uncles: stubborn, undocumented, and capable of causing family collapse at any moment. Humans will remain necessary.
But fewer humans may be needed for the billable middle.
And that middle was India’s golden stomach.
This is why the stock market is getting nervous. Markets are not moral philosophers. They do not care how hard your cousin studied for JEE. They care where future cash flows will grow. They care where margins will expand. They care where scarcity sits. If scarcity sits in GPUs, memory chips, data centers, power contracts, and cloud platforms, capital moves there. If your national champions look like labor arbitrage machines at the exact moment labor is being repriced, capital starts edging toward the door, pretending it only stepped out to take a call.
A country may have talent and still lose valuation.
This is the part we keep misunderstanding. Talent is not the same as ownership. Skill is not the same as leverage. Usage is not the same as control. A tea seller may understand tea better than the company that owns the plantation, the brand, the logistics, and the retail chain. Guess who gets the multiple.
India’s visible listed market has too much of the old services story and too little of the AI chokepoint story. That does not mean India has no AI future. It means the market cannot easily see one it wants to overpay for.
There is a difference.
And difference is where truth often hides, like a cockroach behind the gas cylinder.
The government has noticed. IndiaAI Mission has put tens of thousands of GPUs into a shared compute pool at subsidized rates. This is good. A student, startup, researcher, or small company should not have to sell ancestral furniture to fine-tune a model. Compute access matters. Without compute, AI becomes a club where the bouncer checks not your intelligence but your funding round.
But rented compute is not sovereignty.
Borrowing a pressure cooker does not make you the owner of the kitchen.
India also has semiconductor ambitions. Tata Electronics and ASML have moved around the Dholera fab story, and that is serious. A fab is not a small thing. It is not a software park with glass, security guards, landscaping, and one sad cafeteria dosa. It is chemistry, physics, water, electricity, process discipline, lithography, yield learning, supply chains, clean rooms, maintenance, capital, and time. It is a temple to precision where one dust particle can behave like a terrorist.
So yes, build it.
But do not clap too early.
India has a bad habit of treating announcements as achievements. We print the brochure, hold the summit, put the minister on stage, light the lamp, release the logo, and then behave as if industry has already happened. This is not industry. This is theater with air-conditioning.
Industry is when the thing works at scale for years.
Industry is when the parts arrive on time.
Industry is when the power does not blink.
Industry is when the worker knows what to do, the supplier knows what to supply, the engineer can fix the line, the regulator does not invent fresh confusion, and the bank does not behave like a nervous auntie at a love marriage.
Industry is boring competence repeated until it becomes destiny.
India needs that boredom.
Instead we are addicted to spectacle. Startup awards. AI summits. Drone shots. Founders in black T-shirts. Ministers saying “ecosystem.” Consultants drawing triangles. Everyone “leveraging.” Everyone “unlocking.” Everyone “disrupting.” Meanwhile R&D spending sits at a level that would make a serious scientific civilization wince. Around 0.6 to 0.7 percent of GDP is not enough for a country that wants to stop being merely a market and become a maker of deep technology.
A nation cannot WhatsApp-forward its way into advanced manufacturing.
It has to fund laboratories.
It has to protect scientific institutions from petty politics.
It has to let researchers fail without destroying their careers.
It has to make engineering respectable beyond coding interviews.
It has to teach children that physics is not a punishment invented by coaching centers.
It has to stop worshipping exams and start building things that survive rain, dust, heat, corruption, voltage fluctuation, and human stupidity. Especially human stupidity. That is the final boss.
Meanwhile, ordinary India is being told that AI will make everything magical. Customer care will improve. Schools will personalize learning. Hospitals will diagnose earlier. Farmers will get advice. Courts will move faster. Government forms will become less beastly. Perhaps. Some of this may happen. India has scale, language diversity, public digital rails, and an instinct for jugaad that can, under the right conditions, become innovation rather than duct tape with a slogan.
UPI is proof that India can build something astonishing when the architecture is right.
But UPI also teaches the opposite lesson. It was not built by saying “let there be fintech vibes.” It needed rails, standards, institutions, banks, regulation, adoption, and ruthless simplification at the point of use. Grandma does not care about protocol design. She cares whether the vegetable seller received the money. That is real technology: the miracle disappears into ordinary life.
AI must do that too.
Not in press releases.
In farms, clinics, factories, schools, courts, municipal offices, buses, small businesses, and the lives of people who do not know what a transformer model is and should not have to know. If AI remains a toy for urban pitch decks and English-speaking productivity bros, India will have achieved the rare feat of importing the future and using it mainly to decorate LinkedIn.
The danger is not that AI will destroy India overnight.
The danger is duller and therefore worse.
It may hollow out the very employment ladder that gave India’s middle class its upward path. Fresh graduates may find fewer entry-level coding and support jobs. GCCs may hire more cautiously. Outsourcing firms may squeeze pyramids. The old conveyor belt from engineering college to IT job to EMI to flat to school admission may start juddering. Not collapsing everywhere. Not immediately. But enough to matter.
And India is not prepared for this emotional economy.
We talk of jobs as numbers. But a job is not a number. A job is a son staying in Bengaluru instead of returning defeated to Asansol. A job is a daughter delaying marriage because she has her own salary. A job is a father’s blood pressure improving. A job is a mother telling neighbors, with carefully disguised pride, “software-e ache.” A job is a family’s tiny private independence from fate.
If AI eats the bottom rungs, where do the young climb?
That is the question.
Not “Will India use AI?”
Of course India will use AI. Indians use everything. We used smartphones before footpaths were repaired. We used digital payments while drains still smelled like medieval Europe. We will use AI in English, Hindi, Bengali, Tamil, Telugu, Marathi, and in that mysterious language used by government offices where every sentence appears to have been boiled with old socks.
The question is whether India will own enough of AI to shape its destiny, or merely consume AI while congratulating itself for being very innovative.
There is a difference between sitting at the dining table and being on the menu.
Wake up.
That is not motivational shouting. I dislike motivational shouting. It reminds me of men in tight suits who say “mindset” before asking for money. This is a technical warning wearing a lungi.
Wake up means: stop confusing service success with industrial depth.
Wake up means: stop underfunding R&D and then asking why we do not have world-class deep tech.
Wake up means: stop treating scientists like grant-begging clerks.
Wake up means: stop producing engineers who can solve exam problems but have never built, broken, repaired, soldered, measured, shipped, or maintained anything real.
Wake up means: stop believing every foreign platform with an Indian sales team is “India’s AI revolution.”
Wake up means: stop using the word ecosystem when what you have is a WhatsApp group, three consultants, and a panel discussion at a hotel.
This is not pessimism. Pessimism is saying nothing can be done. Much can be done. India has brains, scale, chaos tolerance, software depth, digital rails, hungry founders, and a young population large enough to scare any spreadsheet. India can build domain AI for agriculture, health, education, logistics, climate adaptation, law, translation, small manufacturing, and public services. India can build tools for the world’s messy majority, not only for polished corporations in California conference rooms.
But it must choose reality over applause.
The old market-darling India was a growth story. The next India has to be a capability story. Not “look how many people we have.” Not “look how many engineers we produce.” Not “look how many startups were registered.” Those are ingredients. Nobody praises a biryani because the kitchen had rice, meat, and ambition. The question is whether the final thing can stand before the public without apology.
Build fabs, yes.
Build compute, yes.
Build models for Indian languages, yes.
Build AI into public systems, yes.
But also build boring things: procurement integrity, lab funding, data governance, power reliability, vocational depth, maintenance culture, industrial supplier networks, university autonomy, scientific patience, and managerial honesty. These are not sexy. Good. Sexy has misled us long enough.
The future will not arrive as a slogan.
It will arrive as a bill.
Power bill. Compute bill. Import bill. Unemployment bill. Education bill. Climate bill. Debt bill. Social frustration bill.
Someone will pay.
Here in my Calcutta fringe, evening is coming. The crows have begun their municipal conference on the wires. A pressure cooker whistles from a nearby house. The same boy who delivered groceries yesterday is dragging his bicycle through mud with one hand and checking his phone with the other. Maybe he is watching cricket. Maybe a reel. Maybe an AI-generated video of a politician singing an old Kishore Kumar song. The future, like sewage, arrives first in small channels.
India can still turn.
But not by sleeping under the quilt of past glory.
Not by saying “we have talent” as if talent alone ever paid for a lithography machine.
Not by declaring victory because a chatbot speaks Hindi.
Not by sending another delegation abroad to announce that India is open for business. Everybody is open for business. Even the paan shop is open for business. The question is what you sell, what you own, what you know, what you can defend, and what the world must buy from you because there is no easy substitute.
That is power.
Everything else is decoration.
So yes, India may have missed the first big AI market party. The dance floor is full of chipmakers, cloud giants, memory kings, and data-center landlords. We are standing near the entrance with a good résumé and an old invitation.
Embarrassing?
A little.
Fatal?
No.
But only if we stop pretending the music is waiting for us.
P.S. References: Bloomberg reporting by Abhishek Vishnoi on India’s AI-linked market underperformance and risk of losing top-five global equity-market status;