TL;DR: Minister Vaishnaw told the IMF their Tier 2 ranking for India is wrong. He’s right—but not how he thinks. The problem isn’t infrastructure or talent. It’s that Indians won’t pay for AI. ChatGPT made just $3.6M from 29 million Indian downloads. The same people who won’t spend ₹400 on ChatGPT happily pay ₹500 for Zomato Gold. Every successful Indian SaaS—Freshworks, Zoho—makes money selling to Americans, not Indians. We produce AI talent for export and consume AI products for free. That’s not Tier 2. That’s optimizing ourselves out of the race entirely.
At Davos last week, India’s IT Minister Ashwini Vaishnaw told the IMF managing director that her organization’s AI readiness classification was wrong. India, he insisted, belongs in Tier 1—not Tier 2 where the IMF had placed us at 72nd globally with a score of 0.49.
The minister has a point. The IMF’s criteria—digital infrastructure, human capital, innovation ecosystem, regulatory framework—miss something fundamental. But not in the way he thinks.
Here’s the number that actually matters:
ChatGPT made $3.6 million from 29 million downloads in India. That’s roughly ₹12 per download.
In the same period, the average Indian Zomato Gold subscriber paid ₹500. The IMF is measuring supply-side readiness. The real problem is demand-side failure.
This isn’t about whether India can participate in the AI revolution. It’s about whether Indians will pay to participate in it.
The Infrastructure We Actually Built
The government narrative is compelling on paper. A 38,000 GPU compute facility. AI Bharat initiatives. Public-private partnerships with the usual suspects. Ministers at global forums defending our honor.
Meanwhile, two American startups—Render and Vercel—did more to democratize cloud deployment for Indian developers than a decade of government schemes. No press releases. No summits. Just products that worked, priced in dollars, used by Indians who somehow found a way to pay when the value was undeniable.
The infrastructure gap isn’t about compute capacity. It’s about what happens between announcement and adoption.
I’ve watched this pattern repeat across institutions. The government announces. Committees form. Budgets allocate. And somewhere between the press release and the product, the friction accumulates. By the time an Indian developer can actually access the “indigenous AI compute,” they’ve already shipped three projects on AWS.
This is the first uncomfortable truth: our supply-side investments keep solving problems that aren’t the binding constraint.
The ₹400 Barrier
OpenAI had to launch ChatGPT Go—a free tier specifically for emerging markets—because $20/month was too expensive. Not economically impossible. Just psychologically unacceptable.
The same professionals who won’t renew a ₹400 ChatGPT subscription pay ₹500 for Zomato Gold without blinking.
They’ll spend ₹2,000 on a single restaurant meal. They’ll buy ₹50,000 phones. But ask them to pay for software that might determine their career relevance, and suddenly they’re price-sensitive.
I’ve seen this in classrooms. Students who arrive in cars worth more than my annual salary open three Gmail accounts rather than pay ₹130/month for Google One when their Drive fills up. This isn’t poverty. This is a specific cognitive pattern around digital goods.
There’s a mental model that explains this. To understand what people actually value, watch what they pay for—not what they say they want. Indian consumers have demonstrated, with $3.6 million in revealed preferences, exactly how much they value AI tools: almost nothing.
The 65% payment failure rate for international subscriptions is often blamed on infrastructure—RBI regulations, card declines, UPI limitations. But when the same consumers want to buy a course from a foreign university or subscribe to a streaming service, they find a way. The infrastructure excuse is cope.
Who We Actually Build For
Here’s where the Tier 2 classification gets interesting.
Every successful Indian SaaS company—Freshworks at $10 billion, Zoho at $500 million plus in revenue—built their businesses by selling to Americans, not Indians. The playbook is explicit: build in Bangalore, sell in San Francisco.
This isn’t a bug. It’s a survival strategy.
When you “solve for India,” you optimize for a market that won’t pay. You build freemium models that never convert. You add payment plans that reduce already-thin margins. You localize for users who will churn the moment a free alternative appears.
The AI Bharat programs, the “Make AI in India” initiatives—they sound like nationalism but function as traps.
They redirect engineering talent toward a market that has demonstrated, repeatedly, that it will not sustain them.
The developers who build for global markets from day one aren’t unpatriotic. They’re rational. They’ve seen what happens to companies that optimize for Indian consumers: endless feature requests, minimal revenue, and eventual pivot or death.
The Trust Deficit
There’s another dimension the IMF metrics miss entirely: institutional trust in AI adoption.
Research on technology adoption consistently shows that trust determines whether theoretical readiness converts to actual usage. It’s not enough to have the infrastructure if stakeholders don’t trust the systems built on it.
In India, we have a peculiar trust configuration. Consumers trust AI enough to use it—29 million downloads prove that. But they don’t trust it enough to pay for it. They’ll use ChatGPT for homework, for emails, for quick queries. But the moment value extraction requires commitment, they exit.
High usage, low revenue. Massive adoption, minimal monetization. Exactly the pattern that makes India look like a Tier 1 market in user metrics and a Tier 3 market in business viability.
For companies trying to build AI products, this trust configuration is poison. You can acquire users cheaply. You cannot retain paying customers at any price.
What Happens Next
The minister is right that the IMF classification is wrong. But the correction goes in the opposite direction.
India’s AI readiness problem isn’t that we lack compute infrastructure or talent or regulatory frameworks. It’s that we’ve built a consumer culture that treats digital goods as fundamentally unworthy of payment.
The next wave of AI tools—the ones that will separate productive workers from obsolete ones—will not be free. They will not be freemium with generous limits. They will be subscription services that cost real money, demand real commitment, and deliver real advantage to those who pay.
The ₹400 that feels excessive today will look like the best investment never made in five years.
But here’s the question no one in Davos asked:
If Indians won’t pay for the AI tools that will determine their economic future, who exactly is all this “AI talent” building for?
The answer, as Freshworks and Zoho discovered, is Americans. Which means India’s AI future isn’t being decided in Delhi or Bangalore. It’s being decided by consumers in Texas and California who actually open their wallets.
That’s not Tier 2. That’s something the IMF doesn’t have a category for: a country that produces AI talent for export and consumes AI products for free. We’re not behind in the race. We’ve optimized ourselves out of it entirely.
Sources: IMF AI Preparedness Index 2024, Vaishnaw remarks at Davos 2025, Sensor Tower data on ChatGPT India revenue, OpenAI ChatGPT Go announcement
