Why India may win the AI race by outgrowing the startup obsession
India talks about artificial intelligence constantly. We talk about startups, funding rounds, talent pipelines, innovation rankings. These markers dominate conferences, policy panels, and headlines. Yet the most consequential question about AI is still asked too rarely and too softly: where does intelligence physically live? The answer is inconvenient for a software-first narrative. Intelligence today lives where infrastructure can support it.
Despite its digital image, artificial intelligence is no longer just code floating in the cloud. It depends on power grids that do not fail, cooling systems that work through heatwaves, land that can be secured at scale, fibre networks with low latency, and long-term contracts that guarantee uninterrupted operations. In practical terms, AI behaves less like an app economy and more like heavy industry. This is not just India’s problem.
Around the world, Governments and companies are discovering that modern AI systems require continuous, industrial-grade operations. Large AI data centres run day and night. Even brief disruptions — power instability, cooling failures, network delays — can lead to serious losses. Reliability, not novelty, has become the defining constraint.
What makes India’s situation different is timing. India is still building much of its core infrastructure. Electricity demand is rising steadily. New industrial corridors are being planned. Cities, transport networks, and digital backbones are expanding. Many advanced economies, by contrast, are trying to insert AI into systems designed decades ago — grids stretched to capacity, land locked in regulatory delays, water and cooling already contested. That gives India a genuine opening, if it chooses to use it deliberately.
India already has three things in abundance: skilled engineers, ambitious startups, and a massive user base. The strategic question now is whether India can convert its infrastructure expansion into AI-ready systems — systems that are reliable, predictable, and designed from the outset to handle large-scale computing. If it does, India’s role in the global AI economy could change in fundamental ways.
One reason infrastructure matters so much is speed. For many AI applications — financial systems, logistics, language services, industrial automation — even small delays matter. When AI systems rely on servers located overseas, data must travel longer distances, adding latency that degrades performance and reliability. This is why companies increasingly prefer to run AI close to users —but only where local infrastructure can be trusted. Globally, this helps explain a striking imbalance. Developing countries account for a large share of internet users, yet host only a small fraction of the world’s AI data centres. The constraint is not demand or talent. It is the ability to host complex systems reliably over long periods of time.
This is where India’s heavy emphasis on startups, while understandable, begins to look incomplete. Startups build applications and services. But the deepest economic power in AI lies lower in the stack: in hosting the infrastructure itself — the servers, facilities, and long-term systems on which everything else depends. When those are located elsewhere, a significant share of value creation flows elsewhere too.
A country can produce world-class engineers and still remain dependent on foreign infrastructure for its most critical digital functions. India is not destined for that outcome. But it becomes more likely if infrastructure planning fails to keep pace with ambition. The logic is simple. Every system grows until it hits its weakest point. For AI today, that weak point is no longer algorithms or data. It is whether systems can run smoothly, continuously, and predictably —without improvisation. AI does not reward jugaad. Small failures compound quickly.In many advanced economies, these weaknesses are already visible. Power connections take years to approve. Industrial land is scarce. Permitting processes move slowly. Local opposition delays large facilities. Faced with uncertainty, companies quietly redirect investments to jurisdictions where timelines are clearer and coordination faster.
India, for now, still has room to plan ahead. Large AI facilities also place heavy demands on cooling and water. A single large data centre can draw tens to hundreds of megawatts of power — comparable to the consumption of a mid-sized city and can place significant stress on local water systems if cooling is poorly designed.
Newer technologies can reduce these pressures sharply, but only when power, water, and land are planned together from the start. Countries that coordinate these systems are far more attractive destinations for AI investment. This is not an unsolvable problem for India. It is an engineering and planning challenge — areas in which India has long experience.
Infrastructure also signals commitment. Large AI facilities require billions of dollars in upfront investment and operate for decades. Once built, they anchor supply chains, skills, and secondary industries. Hosting AI infrastructure creates long-term economic gravity, not just short-term excitement.
This shift is already changing industrial policy globally. In earlier technology waves, Governments competed mainly through incentives. In the AI era, companies look first for certainty: clear rules, reliable utilities, coordinated institutions. Where that certainty exists, investment follows. AI also has a practical role to play inside infrastructure itself.
Around the world, AI systems are improving power grids, transport networks, and utilities by predicting failures, managing demand, and reducing downtime. For India, this suggests a priority often overlooked: infrastructure agencies should see AI not as an external demand they must accommodate, but as a tool to strengthen their own systems first. India has a quiet advantage here. The country trains large numbers of engineers who understand physical systems — electricity, machinery, networks, logistics. AI at scale depends as much on this systems knowledge as on software skills.
So where do startups fit into this picture? Exactly where they work best — on top of strong foundations. In the AI era, the most important platform is not an app store or a funding ecosystem. It is the infrastructure beneath everything else. When that
foundation is stable, startups scale faster and more sustainably. When it is weak, even the best ideas remain dependent on systems outside the country.
India does not need to abandon its startup ecosystem. But it does need to stop treating startup counts as the primary measure of AI leadership. The real competition is about who can host intelligence reliably and at scale. Very few countries have India’s combination of size, growth, engineering talent, and policy flexibility. Whether India merely uses AI — or becomes a place where AI truly lives — will depend less on code, and more on the systems that quietly support it every day.
Author is a theoretical physicist at the University of North Carolina at Chapel Hill and the author of the forthcoming book Last Equation Before Silence; views are personal
