What does India need to do to catch up in the AI race?

In today’s Finshots, we explain why the AI race is not just about building the smartest model.

But before we dive in, if you’re someone who loves keeping tabs on the world of business and finance, hit subscribe if you haven’t already. And if you’re already a subscriber, thank you! Maybe forward this to someone who’d enjoy our stories but hasn’t discovered us yet.

Now, onto today’s story.

The Story

Last week, NVIDIA unveiled the RTX Spark, a new generation of AI chips designed for laptops with up to 1 PetaFLOP of AI compute.

In simple terms, NVIDIA is trying to bring serious AI capability out of the data centre and onto a developer’s desk. With 128 GB of unified memory, these systems can run some AI workloads locally, including prototyping and fine-tuning large models.

So the company at the centre of the AI revolution is now trying to make powerful AI infrastructure feel almost personal.

And yet, as impressive as NVIDIA’s breakthroughs are, they reveal an even more fascinating story.

While NVIDIA gets the headlines, there is another player powering almost every major AI breakthrough on the planet.

It’s not OpenAI, Anthropic or Google. In fact, it isn’t even an AI company, but a small island nation, off the coast of China, with a population smaller than that of Mumbai. We’re talking about Taiwan.

Now, Taiwan didn’t build ChatGPT, and doesn’t even have a single major AI model. Yet without Taiwan, the entire AI revolution would come to a grinding halt.

And that makes you ask, “If Taiwan became indispensable without building the world’s best AI models, what exactly does India need to do to catch up in the AI race?”

To understand this, let’s first discuss what they did right.

Taiwan occupies a strange position in the global technology industry. Most people would struggle to name a Taiwanese software company. But if Taiwan’s chip industry stopped functioning tomorrow, the AI world would be in serious trouble.

Taiwan’s genius was that it did not chase the most glamorous layer of technology, but captured the layer everyone else had to depend on ― chips.

The world’s most advanced AI chips are designed by NVIDIA. Yet they are manufactured thousands of kilometres away by Taiwan Semiconductor Manufacturing Company, better known as TSMC. And TSMC is not just another chipmaker. In Q4 2024, it controlled about 67% of global foundry (basically factories that make chips) revenue. More importantly, it dominates the advanced chipmaking capacity that companies like NVIDIA depend on.

That’s what makes Taiwan’s story so interesting. Sometimes, it is the country that controls the invisible layer without which the entire system cannot function.

And that brings us to India. Whenever discussions about India’s AI ambitions emerge, the conversation usually starts with talent, startups, digital public infrastructure, and population scale.

These are real strengths. India has a large engineering base, a growing startup ecosystem, and public digital rails that already touch hundreds of millions of people.

Yet despite all these strengths, India has not produced a frontier AI model capable of competing with GPT, Claude, Gemini, or DeepSeek.

The instinctive response is to conclude that India is falling behind. But that conclusion only makes sense if we assume the goal is to replicate what countries like the US or China are doing.

The reality is that AI resembles a supply chain far more than a single product.

At the bottom sits energy. Above that sit chips. Then comes infrastructure such as data centres, cloud platforms, networking, and compute clusters. Only after that do we arrive at models. And finally, at the top, sit applications, the tools that businesses and consumers actually use.

Seen this way, the AI race starts to look very different. No country dominates every layer equally. Taiwan is indispensable in chips. China is strong in energy and increasingly competitive in models. The US still leads in frontier models and consumer-facing AI applications.

Each country has found a layer where it can build real leverage.

This is why India does not need to think of the AI race only as a contest to build the next ChatGPT. The smarter question is: which layer of the AI supply chain can India become unusually good at?

Look, India’s position across these layers is uneven.

The country trails the US and China in advanced semiconductor manufacturing. It trails them in frontier-model research and large-scale compute infrastructure. But it possesses something that many other countries do not: an extraordinary ability to deploy technology at scale.

So you could say that India’s strongest advantage may be deployment. Over the last fifteen years, India has shown that it can take complex digital systems and push them across a massive population. UPI and Aadhaar are the strongest examples. 

That capability may matter more than many people realise.

After all, even if India somehow acquired unlimited GPUs tomorrow, another challenge would immediately emerge: research culture. 

Frontier AI models are not simply a matter of having enough hardware. They require a willingness to spend billions of dollars on uncertain outcomes, support long-term scientific research, tolerate repeated failures, and invest for years before commercial returns appear. Silicon Valley spent decades building universities, research labs, venture networks, and talent ecosystems before companies like OpenAI emerged.

This is why focusing exclusively on building the next GPT may be the wrong objective.

History suggests that countries rarely capture value by simply copying the current leader. The bigger opportunity is often to identify one layer of the value chain, become world-class at it, and then build relentlessly around that advantage.

India may need a similar approach.

That does not mean abandoning foundation-model research. It certainly does not mean giving up on sovereign AI. But it does suggest that India’s realistic AI strategy should focus on three things.

First, India needs sovereign compute infrastructure so that its startups, researchers, universities, and public institutions are not fully dependent on foreign AI platforms.

Second, India needs high-quality Indian datasets in areas where it has natural depth, such as healthcare, agriculture, law, education, finance, and regional languages. If India wants AI systems that work for Indian users and institutions, it cannot rely only on datasets built for Western contexts.

And third, India needs to focus aggressively on adoption. This is where the country already has a proven advantage. The country may not lead every foundational layer of AI, but it knows how to deploy technology at population scale.

That may be India’s real opening.

So yeah, the AI race is not just about who builds the best model. It is about who captures the most economic value from AI. And India’s opportunity lies in strengthening its compute and data foundations while using its biggest advantage: large-scale technology deployment.

Because in the end, the AI race is often described as a competition to build the smartest model. But that is probably the wrong way to think about it.

Taiwan proved that the most valuable position is not always the most visible one. 

The real question, therefore, is not whether India can build the next GPT. It is whether India can identify the part of the AI value chain where the rest of the world eventually cannot function without it.

Until then…

Liked this story? Share it with a friend, family member or even strangers on WhatsApp, LinkedIn and X.

Message to all the Breadwinners

You work hard, you provide, and make sacrifices so your family can live comfortably. But imagine when you’re not around. Would your family be okay financially? That’s the peace of mind term insurance brings. If you want to learn more, book a FREE consultation with a Ditto advisor today.