India's AI infrastructure vulnerable to geopolitical shocks, says M2P Fintech co-founder

In the run up to the Global Fintech Fest 2026, Madhusudanan R, co-founder of M2P Fintech, spoke to The Economic Times Digital about the new frontiers of challenges for India in the age of AI, how fintechs have to walk a tightrope and why India is a better market than the US.

India’s vulnerability to geopolitical shocks such as those witnessed during the ongoing Israel-Iran-US conflict are no longer limited to oil prices or supply chain bottlenecks. Foreign-owned large language models (LLMs) and the underlying graphic processing units (GPUs) form the core infrastructure of India’s artificial intelligence (AI) infrastructure. At least, a significant part of it.

M2P Fintech’s cofounder Madhusudanan R believes this could be exposed next if India does not reduce its over-reliance on foreign technology. “We are still very dependent on international tech. What the war has shown us in the last few months is that nobody is a friend. We could become a victim of some adversary somewhere else, which might not be of our own doing. That's a real threat,” he told The Economic Times Digital.

He also warned against cheaper alternatives based on open-source models, especially those coming out of China, questioning their operating environment.

“You don’t know what the Chinese models are, who's behind them, what support they're getting. They say they're open weights and all that, but there's still a lot of suspicion around how they operate in an open model,” he added.

At present, a handful of companies like Sarvam AI, BharatGen, AI4Bharat, CoRover and Gnani AI are building frontier models at scale in India. Since 2024, when the Cabinet approved the IndiaAI Mission, with a budget outlay of ₹10,371.92 crore over five years, India has achieved 38,000 GPUs from an initial target of 10,000 GPUs.

Though India has increased its computing prowess by roughly four times, the entire process still sits on chips manufactured in our countries such as the US – an underlying problem that needs long-term solutions. Competing with behemoths like Anthropic, OpenAI and Google is another challenge.

Madhusudanan opined: “This (competition) is an under-appreciated risk, and there isn't a quick fix for it. Companies like Sarvam are trying to build a model out of it, but the investments required for us to get to that level of sophistication are far, far greater. This (chips) can become an instrument that can be used against us as well.”

Fintech is a heavily-regulated space as the Reserve Bank of India keeps a close eye on the guardrails placed within innovation. With AI coming into the mix, more regulatory dimensions are expected to be added across the layers. In addition, adherence to the Digital Personal Data Protection (DPDP) Act, 2023, with respect to data localisation, further adds responsibilities on fintechs.

So, how do fintechs operate in such a dynamic landscape, where policies are still catching up to the speed of AI-backed innovation? Understanding the spirit of the regulation is critical, as per Madhusudanan, whose business was impacted by RBI’s clampdown on non-bank wallets and pre-paid cards from loading their credit lines into these platforms in 2022.

“When the regulations came, the regulator took a very adverse view and shut it down in 2022. But when we rolled it (PPI programmes) out, it had passed every compliance test — every bank agreed with it. Those programs were live for five, six years, and banks went through several regulatory audits. Yet the regulator took the stance that it was never approved. Those scars continue to steer the way we operate,” he said.

Even with AI in the mix, as per Madhusudanan, M2P Fintech applies certain principles in product development to answer questions such as what's the regulatory test here, how might these regulations change, and how do they make sure the platform adapts?

He said: “We always assume that even though the regulation isn't clear, what is the spirit behind it? What would the rule be if this were to get regulated? Then you start building. Explainability is the most critical thing. We have a couple of large NBFCs that have adopted our back-office and middle-office products to manage their loan portfolios. Those products have to explain themselves.”

He further claimed that if a customer is priced differently for a loan, the system will show that the customer belongs to a particular cohort. “... That cohort has consistently behaved in a certain way, the loss rates are at a certain level, and therefore the customer is priced this way. That explainability is by design. That's the principle we operate on, and the regulations will catch up over time. “

Madhusudanan also made an interesting assertion, contrary to conventional wisdom, that India offers a clearer regulatory landscape and operating environment than other markets such as the US. The difference? India and other emerging markets work on a rule-based framework, while the US operates on a principle-based framework.

“When we first went to the US three or four years ago, we assumed India's rules were the strict ones we had to follow to the letter. But under principle-based regulation, it's a judgment call. You can be taken for a ride, or the regulator can take an adverse view based on how they read the situation, because nothing is clearly spelled out. Establishing the reasoning behind a principle is even harder,” he suggested.

In India, the regulator usually spells out the finer details. “For instance, when you have to send the MITC, it has to be sent in a font size, within a particular timeline. So, we came to feel India is actually the better regulatory environment: as long as nothing goes wrong, it's fine. Those [US-style] markets are very litigious, and a single class-action suit can take the company down,” he explained.

However, irrespective of where one operates, the questions on AI and its integration in fintech continues to be a hotly-debated topic. Challenges with data privacy, regulatory hurdles, data localisation, access to AI agents, among others, still need comprehensive solutions. Still, Madhusudanan believes AI will continue to stay relevant; it might face pricing pressures and slow down.

Comparing the eras of the internet and AI, he explained, “Everybody refers to this as the internet bubble of 2001. But back then, you had thousands of companies without any underlying business model. This time it's a handful, probably a few hundred companies focused in the [Silicon] Valley, and a lot of money has been poured in. I don't think it will bust the way the naysayers expect. It will slow down.”

While Madhusudanan doesn’t worry too much at the moment, he does caution that there's certainly some heating up happening, and in some form or shape a consolidation will play out over the next two-three years. “All the signs are there, but nobody's pulling the plug yet.”