As India races to position itself as a global artificial intelligence powerhouse, a stark reality emerged at Nasscom’s pre-summit dialogue on January 12, 2026: the nation’s AI ambitions could be derailed not by lack of innovation, but by the unprecedented energy demands threatening to overwhelm its infrastructure.
The high-level event, held as an official precursor to the AI Impact Summit 2026, brought together regulators, hyperscalers, and policy experts who painted a sobering picture of the challenges ahead. According to data presented at the conference, AI will become the most significant driver of energy consumption growth, with global demand set to more than double by 2030 to around 945 terawatt-hours (TWh).
“If you take out power and telecom, there’s no AI. It is as simple as that,” a Nasscom representative stated bluntly during the opening session, framing the infrastructure challenge in stark terms.
The Sustainability Paradox
The event’s first panel on sustainable AI infrastructure revealed a complex balancing act between growth and environmental responsibility. Vrushali Gaud, who leads Climate Operations for Google globally, outlined the tech giant’s “24/7 carbon-free energy” philosophy, emphasizing the challenge of powering AI data centers with renewable sources around the clock.
Google’s recent announcement of a gigawatt-scale data center in Visakhapatnam—a departure from traditional Mumbai-Chennai clusters—signals a strategic shift toward regions with more substantial renewable energy potential. Andhra Pradesh, with 50% of its installed capacity from renewable sources, emerged as a model for what states must offer to attract AI infrastructure investments.
However, Prasanto Kumar Roy from FTI Consulting highlighted a critical gap: India lacks clear policy mandates on Power Usage Effectiveness (PUE) standards for data centers. While Singapore set an aggressive PUE target of 1.3 six years ago, India’s data centers still operate closer to 2.0—meaning they consume twice the power for the same computational output.
Water Wars and Cooling Conundrums
Microsoft’s Kavickumar Muruganathan addressed the elephant in the room: water consumption. With average Water Usage Effectiveness (WUE) numbers of 0.3 liters per kilowatt, Microsoft has reduced water use by 39% since 2021. The company is pivoting toward liquid cooling approaches that could achieve zero-water evaporative cooling, but the transition comes with its own trade-offs.
“The fundamental issue is heat,” Muruganathan explained, noting that higher compute densities in GPU infrastructure are forcing a paradigm shift from traditional HVAC systems to direct-to-chip cooling solutions.
The Democratization Dilemma
The second panel tackled AI democratization, revealing tensions between hyperscaler dominance and local innovation. Sheetal Srikant from Bharat Cloud represented indigenous providers serving small and medium businesses—a market segment largely untouched by global giants.
“We are literally taking digital transformation to the offline,” Shikha said, emphasizing the role of local cloud providers in reaching India’s “real” businesses through vernacular language support and localized service models.
Intel’s Sanjay Krishen challenged conventional wisdom about GPU scarcity, arguing that India’s obsession with GPUs overlooks readily available alternatives. “If you think of GPUs as luxury cars, we’re asking if we can build enough luxury cars to satisfy the country’s transportation needs. You don’t need to,” he stated, advocating for diverse chip architectures including CPUs and NPUs.
Infrastructure Fragility Exposed
Perhaps most alarming were revelations about network resilience. Vishwanathan Subramaniam from Lightstorm disclosed that while global subsea cables experience only 200-250 cuts annually across 600+ cables, India’s 30,000-kilometer network suffers 8-10 cuts daily—a catastrophic vulnerability for AI workloads requiring 99.999% uptime.
The Telecom Regulatory Authority of India (TRAI) acknowledged infrastructure gaps, including the need for India’s own cable repair ship and the better utilization of railway fiber networks, though service-level agreement challenges remain.
The Road Ahead
As India prepares to host the AI Impact Summit 2026, the Nasscom pre-summit laid bare an uncomfortable truth: the country’s AI infrastructure exists in a precarious state between ambition and capability. With data center capacity currently at just 1.4 gigawatts and projections calling for dramatic expansion, the next 12-18 months will prove whether India can build the sustainable, resilient, and accessible infrastructure its AI dreams demand—or whether those dreams will be throttled by the very wires and power lines meant to carry them forward.