Building India’s control over its digital future

India should use the best processors, software, models and platforms the world has to offer. But where technology becomes part of critical national digital infrastructure, control over its deployment and operation must remain with us. There should be no external “kill switch” that can disable strategic capability. Continuity must be protected through access to know-how, technology transfer where feasible, adequate spares and components, diversified supply chains and progressively greater Make-in-India capability.Sovereignty also cannot exist without scale.

AI is becoming essential to economic and national infrastructure. India therefore needs very large compute capacity within the country - capacity that our enterprises, startups, researchers and government can access and direct towards Indian priorities, from healthcare, education and agriculture to governance, financial inclusion, Indian languages, science and defence. 92-crore IndiaAI Mission was an important intervention because it recognised compute, indigenous models, datasets, skills and responsible AI as national priorities.

Its next phase should focus more on sustained compute availability for mass scale inferencing (instead of just model training), sovereign token factories serving open source and sovereign models, affordability, multi-billion private capital mobilisation and effective utilisation in terms of useful AI output created.For an AI economy, GPUs and megawatts are essential inputs, but not the ultimate measures of success.

We should increasingly measure how efficiently infrastructure converts energy and capital into intelligence - tokens per kilowatt, tokens per dollar and, ultimately, economic and societal value created from compute.Compute alone, however, does not create AI sovereignty. India also needs sovereign models for strategic, critical and population-scale applications. At the same time, we should exploit the extraordinary global innovation happening in open-source AI by running, fine-tuning and optimising such models within trusted sovereign environments where we retain control over infrastructure, data and deployment.

Global frontier models offer remarkable capabilities, and Indians should benefit from them for numerous personal and enterprise applications where strategic dependence is not a concern.The choice, therefore, is not Indian versus foreign. It is where to use what. For every important workload, we should ask: How strategic is it? Who controls the infrastructure and model? Can access be withdrawn? Do we possess alternatives if geopolitical, technological or commercial circumstances change?Finally, India must move higher up the AI value chain - not merely build data centres and host imported technology, but build AI clouds, models, inference platforms, token factories and agentic systems.

We should produce intelligence competitively for India and eventually export AI capacity and services to the world.Sovereign AI is ultimately about ensuring that India has the scale, capability and control to harness AI for its own development - while retaining the freedom to choose its technology partners, its priorities and, ultimately, its own digital future.The writer is co-founder, CEO and MD, Yotta Data Services