Billions are going into NVIDIA's chips right now, data centres are coming up everywhere, and compute is being called the new oil.
India needs that infrastructure too. But I do not think spending the most at that layer is our only path to winning.
Here's how I actually think about this. There are three things happening in AI.
→ Someone has to build the chips and data centres that make it all run. → Someone has to build the models on top of that. → Someone has to actually use all of it to solve a real problem for a real person.
The first two layers demand extraordinary capital, research talent and time. India is investing in both, but our most immediate advantage may sit in the third.
The application layer.
The place where local language, local behaviour, distribution and context matter more than benchmark leadership.
A model can be built anywhere.
But knowing what a fisherman in Bengal needs, how they ask for it, what data they trust and what interface they will actually use, that understanding has to come from much closer to the problem.
That is where Indian companies can build an edge today.
Not by pretending infrastructure does not matter. By recognising that we do not need to own every underlying layer before we start creating value on top of it. India will still need its own compute capacity and sovereign-model capability. The IndiaAI Mission is already investing in both.
But while that foundation is being built, there are millions of problems in agriculture, healthcare, finance, logistics and public services that global model companies will not solve for us by default.
The ground available to win today is where global intelligence meets local context.
The question is not whether infrastructure matters. It does.
But: where can India create disproportionate value without first having to outspend everyone underneath it?