As global leaders gather at the AI summit in Delhi, the familiar script plays out yet again: The US versus China, trillion-dollar valuations, compute races, and speculation about machines that may one day rival human intelligence. India appears in this narrative, but usually in a supporting role. It is described as a large market, a vast engineering workforce, and a services hub, rarely as a country that will shape the architecture of Artificial Intelligence (AI) itself. That framing reflects a hierarchy and level of arrogance that places Silicon Valley and Beijing at the centre of technological destiny and treats everyone else as peripheral.
Over the past three years, Silicon Valley has delivered a steady stream of grand AI pronouncements: AI will replace most human jobs within a decade; it will achieve general intelligence; it poses existential risks if not tightly controlled. And now, according to former Google CEO Eric Schmidt, the constraint may not be algorithms or talent but electricity itself. Schmidt has argued that the US may need roughly 92 gigawatts of additional power to sustain AI’s trajectory, the equivalent of about 60 nuclear power plants.
That projection is framed as proof of scale and inevitability, but it also carries a subtle message: Only those who can marshal nuclear-scale power grids and limitless capital truly belong in the AI race. For a country like India, still balancing development priorities and energy transitions, the suggestion is that the decisive moves are being made elsewhere.
It is not just Schmidt. When Nvidia CEO Jensen Huang declined to attend the summit in India, citing “scheduling conflicts”, he sent a clear message. He would not miss a similar gathering in China, where compute and semiconductor strategy are treated as instruments of national power. The contrast revealed how parts of Silicon Valley still view India — consequential, but not central.
India must reject that view.
The central flaw in the dominant AI narrative is the assumption that scale equals progress. It assumes tomorrow’s systems will simply require more GPUs, more data centres, and more electricity. History shows that linear extrapolation collapses at every major inflection point.
In the early 20th century, forecasters warned that expanding telephone networks would require millions of operators, because they assumed the future would simply be a larger version of the present. Technology rapidly erased that constraint. In 1943, IBM’s president predicted a global market for five computers. Today, billions of devices shape how we live, work, and think. As exponential technologies advance, they become smaller, faster, and cheaper, and their architecture changes.
AI will be no different.
Already, specialised AI chips are reducing energy consumption per operation. Model compression and distillation techniques are shrinking large systems into smaller, efficient versions that retain core capabilities. Edge computing is moving intelligence closer to devices rather than concentrating it in hyperscale facilities. Open-source models are being optimised to run on standard hardware. The architecture is evolving rapidly.
The competition will not be won by those who build the largest models, but by those who design intelligence that works under real-world constraints and solve consequential problems.
This is where India’s strengths lie.
India does not have the deepest venture capital markets or the most advanced semiconductor fabs. What it does have is unmatched experience in building digital systems at scale under constraint. It built Aadhaar to give digital identity to more than a billion people. It built UPI to process billions of transactions at negligible cost. It launched ONDC to prevent monopolistic capture of digital commerce. These were architectural solutions built for India’s context, not copies of Silicon Valley platforms.
AI should follow the same logic.
India does not need trillion-parameter models trained on the largest GPU clusters in the world. It needs models optimised for Indian languages, agricultural realities, health care gaps, and climate vulnerabilities. It needs systems that can run efficiently on accessible hardware and operate in low-bandwidth environments. It needs AI that strengthens farmers, doctors, teachers, and public servants.
A mid-sized model optimised for rural health care diagnostics may deliver more social value than a giant chatbot optimised for advertising engagement. An agricultural AI tuned to soil patterns and monsoon variability may double incomes without requiring nuclear-scale power expansion. Smaller, specialised systems designed for real constraints can outperform bloated general systems in human impact.
Silicon Valley’s fixation with scale reflects its incentive structure. Venture capital rewards dominance, public markets reward grand narratives, and size becomes a proxy for success. Bigger is treated as better because that is what the system pays for.
India operates under different incentives. It must solve for inclusion, affordability, and resilience. Its advantage lies in building systems that work for 1.4 billion people, not in chasing someone else’s benchmark.
Technology does not simply advance; it converges. Chips improve, sensors multiply, algorithms become more efficient, and architectures shift. India does not need to replicate today’s centralised infrastructure at national scale. It can leapfrog toward distributed, efficient intelligence designed for its own realities.
India also does not need to match the spending of the US or China. It needs to invest strategically, build indigenous capability, embed governance from the beginning, and focus on deploying the most useful systems rather than the largest ones. If it stays grounded in solving real problems and expanding opportunity, it will shape AI in a way that uplifts humanity rather than merely inflating valuations.
Vivek Wadhwa is CEO, Vionix Biosciences. The views expressed are personal
