India's AI data centre race moves beyond traditional technology hubs

Details: Shuriah Niazi; Category: Data Centres & Networks; 05 August 2026; 299 views

The $2 billion investment by the Spanish company Submer Group to establish a semiconductor facility and a 1 GW AI-ready data centre campus in Madhya Pradesh in central India marks a significant shift in the country’s digital landscape.

Announced at the region’s Tech Growth Conclave 3.0, this project demonstrates how data centre development is expanding beyond traditional hubs like Mumbai, Chennai, and Hyderabad.

Given the specific requirements of AI, operators are now prioritising reliable power, ample land, access to renewable energy, and swift regulatory approvals over proximity to major metropolitan financial centres or subsea cable landing points. Emerging states are now becoming key destinations for hyperscale AI infrastructure.

"AI is fundamentally changing how data centre locations are selected," said Rajeev Gandhi, Group CTO at UST MobileComm, which provides engineering and technology services for telecom operators, equipment vendors, and enterprises.

"Traditional cloud data centres were designed to minimise latency, so they were naturally built close to enterprise customers, major cities and international cable landing stations. AI training workloads have different priorities. They require reliable power and large campuses capable of supporting high-density GPU infrastructure, while being less sensitive to a few milliseconds of network delay."

Shifting priorities

Driven by shifting priorities, data centre developers are expanding beyond traditional hubs like Mumbai—which accounts for nearly half of India's capacity—towards emerging cities such as Visakhapatnam, Nagpur, Bhubaneswar, Jaipur, Lucknow, Kochi, and Raipur to cater to future AI investments.

"States with abundant renewable energy, lower land costs and room for long-term expansion are becoming increasingly attractive. Rather than transmitting electricity over long distances to major cities, it often makes more economic sense to build AI centres closer to where power is generated”, says Gandhi.

Madhya Pradesh's ability to make swift decisions proved crucial in attracting Submer. According to reports, land for the proposed project was allocated within just a few days of the company's request.

"What distinguished Madhya Pradesh here was the speed of approvals," Gandhi said. "Every state in India now offers capital subsidies and duty exemptions. Very few can compress decisions into a week, and that is increasingly what investors are buying."

This investment is also significant because 'Submer'—a global leader in immersion and direct-to-chip liquid cooling for AI—is now evolving into a developer and operator of AI-ready campuses.

"Submer sells immersion and direct-to-chip cooling systems, and what they announced at Bhopal is an AI-ready campus phased towards 1 GW," said Arif Gasilov, Partner, Natural Resources & Built Environment at Gasilov Group, an international advisory firm that provides strategic analysis on infrastructure, energy, water, urban development, and sustainability. "A cooling vendor is putting US$2 billion behind building facilities where its own product goes inside."

Submer's strategy reflects the broader changes in the industry. As traditional air cooling is no longer effective for high-heat-generating AI systems, liquid cooling is rapidly becoming the preferred thermal solution for hyperscale facilities.

"Operators at GRI's India data centre forum put around 75% of new project pipelines on liquid cooling," said Gasilov, adding that the Madhya Pradesh investment reflects growing confidence that demand for AI infrastructure will continue to expand over the next decade.

Unlike mature markets in Europe and North America, where operators are spending billions to retrofit ageing facilities for high-density GPU computing, much of India's AI infrastructure is still in the planning and development stage.

"India's relatively young data centre market has become an advantage," Gandhi said. "Since much of India's AI capacity is yet to be built, operators can design facilities around liquid cooling, higher power densities and energy-efficient architectures from day one."

He believes AI-ready campuses will soon become an essential component of India's national digital infrastructure.

"The real opportunity for India is not simply building larger data centres but building smarter and more sustainable AI infrastructure," Gandhi said. "AI-ready data centres will become an extension of national digital infrastructure, supporting cloud services, telecom networks, AI inference and advanced digital services."

While the opportunities are immense, industry experts agree that the biggest challenge lies in ensuring infrastructure can support the explosive growth of AI workloads. The issue is no longer whether India can generate enough electricity, but whether reliable power can be delivered quickly to emerging AI clusters.

"Power availability has become the single most important consideration. Investors look for firm grid commitments, resilient transmission infrastructure and the ability to scale over many years. Delays in power availability can have a much greater financial impact than differences in electricity tariffs."

India has substantially expanded its electricity generation capacity over the past decade, crossing 500 GW of installed capacity, with nearly half coming from non-fossil fuel sources.

Gasilov believes transmission has become India's defining infrastructure challenge.

"India is building power plants faster than it is building the lines to move the power," he said, noting that renewable energy is already being curtailed in some regions because transmission capacity has failed to keep pace with new generation.

Digital connectivity

Beyond electricity, digital connectivity is emerging as the second critical pillar of AI infrastructure. Hyperscale campuses require multiple independent fibre routes, carrier diversity and seamless interconnection with cloud providers and internet exchanges to support distributed AI workloads. According to Gandhi, India's connectivity outlook is encouraging.

"India is actively strengthening its international network capacity through multiple new subsea cable projects. Equally important is the emergence of Visakhapatnam as a third international connectivity gateway alongside Mumbai and Chennai, reducing concentration risk and improving national network resilience,” Gandhi said.

He cautioned, however, that inland AI campuses would require sustained investment in long-haul domestic fibre networks to ensure emerging data centre hubs enjoy the same resilience as established coastal markets.

Water availability is also becoming an increasingly important factor in investment decisions. A significant proportion of India's existing data centres are located in water-stressed regions, raising concerns about the long-term sustainability of conventional cooling systems as AI deployments continue to expand.

Gasilov said this is another reason why liquid cooling is rapidly gaining acceptance across the industry, while developers are increasingly favouring locations that can support both sustainable cooling practices and renewable energy integration.

Both experts believe government policy will play a decisive role in determining whether India can translate major investment announcements into operational AI infrastructure.

"The highest priority is ensuring reliable power infrastructure is available before major AI data centre investments are announced," Gandhi said. "Dedicated AI and data centre zones with pre-planned power and connectivity can significantly reduce project delays and improve investor confidence."

He also called for a consistent national framework governing renewable energy access, faster approvals and the completion of the long-awaited National Data Centre Policy.

Gasilov believes policy reforms should go further by addressing structural issues in infrastructure planning and transparency.

"Federal and state governments should start with rules on who pays when a large customer needs new lines and substations," he said. "They should also require every facility to report water withdrawn and water consumed, and build transmission lines ahead of demand rather than behind it."

India’s advantage

Despite these challenges, experts believe India is well positioned to emerge as one of Asia's leading AI infrastructure markets. Unlike Singapore, Malaysia and Indonesia, where much of the recent expansion has been driven by regional cloud demand, India's growth is supported by one of the world's largest domestic digital economies.

"India's competitive advantage lies in its scale," Gandhi said. "The country combines one of the world's largest and fastest-growing digital economies with a deep pool of engineering talent, competitive construction costs and expanding renewable energy capacity. India's data centre growth is driven primarily by domestic demand."

He added that India's strategic location between the Middle East and Southeast Asia could eventually establish it as a natural interconnection hub for regional AI and cloud infrastructure.

Gasilov believes India's opportunity also stems from the gap between the country's massive data consumption and its relatively modest data centre capacity.

"India is consuming roughly one-fifth of the world's data while hosting less than five per cent of global data centres," he said. "That means Indian facilities are primarily serving Indian demand rather than someone else's."

However, he cautioned that the country's long-term ambitions would ultimately depend on its ability to deliver supporting infrastructure on time.

For Gandhi, the Submer investment represents far more than a single project—it signals the direction of India's next phase of digital infrastructure development.

"Ultimately, investors value reliable infrastructure, predictable policies and faster execution even more than financial incentives," he said. "Those will determine India's long-term competitiveness in AI infrastructure."