AI infrastructure in India needs to be decentralised, says Vishal Dhupar of NVIDIA

NVIDIA India MD Vishal Dhupar believes that decentralisation of AI infrastructure beyond cities like Bengaluru should be of utmost priority for the nation. Speaking with Fortune India, Dhupar asserted that while such hubs attract talent with their prosperity and diversity, over-centralisation leaves other regions overlooked and underdeveloped.

“AI factories have to come up not in one place, but in multiple places, so that people can get access to those factories closer to them and work,” said Dhupar.

He explains that AI factories will transform India’s raw assets—data, culture, and skills—into computer intelligence. These factories will then provide specialised computing infrastructure and organisational frameworks designed to develop, deploy, and scale AI solutions efficiently.

"Bengaluru has played a pivotal role in this growth, currently housing over 30% of Global Capability Centres (GCCs) and 35% of the GCC workforce. The city is also home to numerous AI and technology startups, fostering an ecosystem that promotes innovation, provides financial support, and nurtures businesses,” S Anjani Kumar, partner, Deloitte India, tells Fortune India.

While Bengaluru dominates the AI startup landscape, even housing 43% of all GenAI startups in India in 2024--according to an October Nasscom report--Dhupar believes the next phase of growth must be decentralised.

India is home to a vast tech talent pool that has been instrumental in driving the global IT industry to its $3 trillion scale. A GitHub report from October highlighted that India has over 17 million software developers, growing at an annual rate of 28%. The AI market in India is projected to reach $20-22 billion by 2027, growing at a CAGR of 30%, according to Deloitte. This strong foundation positions the country well for the AI era. And with machines now able to understand human language, all 1.4 billion Indians have the potential, he says, to participate in the AI economy, making it significantly more inclusive.

Analysts echo Dhupar’s perspective on the importance of decentralising technological growth away from established hubs such as Bangalore.

“It is tempting to concentrate all our AI horsepower in a few giant ‘AI factories,’ but India’s diversity demands a more distributed approach. By extending compute power beyond the big tech hubs into multiple cities, we can cut down latency and broaden access to AI services,” Nipun Kalra, India leader, Boston Consulting Group X, tells Fortune India.

Adding to Dhupar’s point, Kalra says that by establishing local data centres and compute facilities, data sovereignty is not only enhanced, but the benefits of AI—spanning sectors from healthcare to agriculture—can also reach every corner of the country.

“This will require creative public–private collaboration​, but if we get it right, a network of regional AI hubs can transform India’s AI ecosystem into a resilient, inclusive powerhouse,” Kalra adds.

Establishing AI factories across states and cities would allow local talent to contribute from their home regions, Dhupar says. This would not only boost regional economies but also ease pressure on urban centres, fostering balanced and inclusive growth nationwide.

“These advancements have, however, brought challenges such as congestion and traffic. Decentralising GCCs and other businesses can alleviate these issues while continuing to bolster AI infrastructure, which is crucial for innovation,” Kumar adds.

Taking AI to the world

With inclusive AI growth across regions, India can emerge as a global exporter of what he calls, “manufactured intelligence”. This is because by solving population-scale challenges with diverse, locally trained foundational models, the country has the potential to offer uniquely robust AI frameworks to the world.

“[There are] government policies, startup communities are already in play and taking advantage of [translating] their ideas into reality, and academia, and if I combine all this, no doubt we will not only consume intelligence but will be a net exporter of intelligence. I wouldn't be surprised [if] sooner than later, we will be called the capital of intelligence,” Dhupar said.

And building these AI systems and their infrastructure will not only maintain India's strength in services but also create large-scale employment, like traditional manufacturing.

“We have been a services economy. We've been able to generate a certain number of jobs, but when manufacturing comes into it, we generate more jobs. Now we have an opportunity to create solutions for population-scale problems,” Dhupar said.

Building population-scale solutions

Amid this massive potential, Dhupar points out that the current progress remains insufficient. While data centres are being built to support the AI vision, a truly scalable infrastructure remains absent. Scalable data centres are being built to grow and adapt to changing workloads without compromising on performance or reliability.

“There are data centres being built for AI factories across the nation, and talking about scalable data centres, [these] are second to none,” he said.

As of February 2025, India had 153 data centres, with Mumbai leading at 38, followed by Bengaluru with 21, according to Statista.

Scalability is not only a challenge for data centres in processing growing volumes of data but also for AI research, which needs to tackle larger, more complex problems.

“We have the startups, and the infrastructure is already [there] in the country. I see a lot of work happening on research, we now need to start seeing the scale coming up and that's a motion that we are into,” Dhupar said.

Addressing population-scale challenges requires robust research capabilities and the development of the right tools and frameworks.

He agrees that although many AI startups may fail in this journey of building population-scale solutions, together they will drive the ecosystem.

Nvidia’s role in India’s AI story

Nvidia, for now, is investing in upskilling talent in deep learning, partnering with both government and private sector players to enhance AI-related skills.

“NVIDIA is focussing on building talent and helping private and government players to endure their talent, [which] goes on to build population-scale applications and help people with recipes to be sovereign,” says Dhupar.

While Dhupar did not provide specifics, he mentioned that Nvidia will continue to make small, tactical investments in startups. The company is also focussing on improving its Hindi large language model (LLM) to serve as a template for other Indian languages, ensuring that the intellectual property for these LLMs remains within India.

“Hindi LLM out of 23 national languages is one way to show that we don't need to get inundated, that we can do it. Here is a methodology, here is a purpose, this is how we can do it. If you can do Hindi, and then there are so many languages, we can make those. We help people with templates and blueprints, so that the IP belongs to the nation,” Dhupar said.