The narrative coming out of diplomatic circles is as predictable as it is dangerous. We are told that India is "bringing enormous capabilities" to the global AI stage. We hear about the "talent pool," the "data advantage," and the "digital public infrastructure." It sounds magnificent in a press release. In the real world, it is a sedative.
If you look at the current trajectory, India is not building an AI superpower. It is building the world’s largest back office for other people’s intelligence.
The "enormous capabilities" being touted by officials like High Commissioner Doraiswami are largely centered on volume, not value. India has the most developers. India has the most GitHub contributions. India has the cheapest data. But in the world of Generative AI, volume is a commodity. Value is captured at the architectural level, and right now, India is barely in the room.
The Talent Myth: Quality vs. Quantity
We need to stop celebrating the "four million developers" statistic. It is a vanity metric.
I have spent a decade auditing engineering teams across Bengaluru, Hyderabad, and Pune. The "talent" being exported is predominantly trained in maintenance and implementation, not fundamental research. Most Indian developers are experts at using APIs built in San Francisco to solve problems defined in London.
When we talk about AI capability, we aren't talking about writing Python scripts to call GPT-4. We are talking about the ability to design novel transformer architectures or optimize low-level CUDA kernels.
- The Reality: India produces hundreds of thousands of engineers annually, yet a tiny fraction can contribute to the core development of Large Language Models (LLMs).
- The Gap: While the US and China fight over GPU clusters and algorithmic breakthroughs, the Indian tech ecosystem is largely focused on "wrappers"—apps that simply put a thin UI over someone else's model.
If you are building a company that relies entirely on an OpenAI API key, you don't have an AI company. You have a lease. And that lease can be terminated or priced out of existence at any moment.
Data Sovereignty is a Pipe Dream Without Compute
The "India has more data than anyone" argument is the most frequent logical fallacy in the industry. Yes, the India Stack and UPI have generated a mountain of behavioral data. But data is not oil; it is iron ore. It is useless without a furnace.
The furnace, in this case, is compute.
India lacks the sovereign compute capacity to train frontier models on its own data. While the government has announced the IndiaAI Mission with a $1.2 billion budget, it is a drop in the ocean compared to the private Capex of Microsoft or Meta.
Imagine a scenario where a country has all the world’s gold mines but doesn’t own a single refinery. That country doesn’t control the gold market; it just does the digging. That is India’s current position in the AI supply chain. We are providing the raw material (data) and the labor (annotation/labeling), but we are sending it abroad to be refined into high-value intelligence that is then sold back to us at a premium.
The "Service-First" Curse
The greatest hurdle to India’s AI dominance isn't technical; it’s cultural. The Indian tech sector was built on the "Services Model." This model prioritizes billable hours, risk mitigation, and client satisfaction.
AI requires the exact opposite. It requires "Product Thinking"—massive upfront R&D, a high tolerance for failure, and the pursuit of intellectual property.
The big Indian IT firms are currently "upskilling" millions of employees in AI. Do not be fooled. They are not training them to invent the next Sora or Claude. They are training them to automate legacy code migration so they can keep their margins high while their traditional business model evaporates. They are trying to use AI to save the past, rather than build the future.
Why "AI for All" is a Distraction
There is a noble-sounding goal frequently mentioned: "AI for the masses" or "Inclusive AI." While democratizing access is important, it is often used as an excuse for not competing at the frontier.
Focusing solely on localized applications—like AI for crop yields or vernacular translation—is vital for social good, but it doesn't create a global power player. You cannot have a seat at the geopolitical table of AI if you are only a "user" of the technology, no matter how "inclusive" your use cases are.
If India wants to be a leader, it must stop trying to be the world's "helpful assistant" and start being its architect.
The Brutal Path Forward
If I were advising the Ministry of Electronics and Information Technology (MeitY), I would tell them to stop the victory laps. The "enormous capabilities" are currently potential, not kinetic.
To actually disrupt the status quo, we need to stop asking "How can we use AI?" and start asking "How can we own the stack?"
- Kill the Wrapper Culture: Stop funding startups that are just GPT-4 proxies. If there is no proprietary model or unique data-loop, it’s a waste of capital.
- Repatriate the Brains: The best Indian AI minds are currently at Stanford, DeepMind, and OpenAI. They didn't leave for money; they left for compute and the freedom to fail. India needs to build "Compute Sanctuaries"—tax-free zones where the world’s most powerful GPU clusters are available to researchers for free, provided the IP stays local.
- Aggressive Specialization: India will not beat Nvidia at chips today. It probably won't beat OpenAI at general-purpose LLMs this year. But it can dominate "Vertical AI." Whether it's AI-driven drug discovery or autonomous systems for chaotic urban environments, pick a niche and over-index on it.
The High Commissioner is right that the world is watching India. But they aren't watching in fear of a new competitor. They are watching a massive market of 1.4 billion consumers and a workforce of millions ready to be plugged into their platforms.
Being the world's best "user" of AI isn't a badge of honor. It’s a surrender.
True capability isn't measured by how many people in your country use a chatbot. It’s measured by whether the rest of the world has to ask your permission to use yours. Right now, India isn't even in the conversation for the latter.
Stop celebrating the scale of the crowd and start looking at who owns the stage.
