Enabling Innovation Without Dependency: India’s Approach to AI Compute Access

Artificial intelligence has the potential to be one of the most transformative forces of our timedriving economic growth, improving public services, and accelerating solutions to complex global challenges. From enabling early disease detection and precision agriculture to strengthening disaster preparedness, climate modelling, and welfare delivery at scale, AI is already reshaping how societies function. As with earlier technological revolutions, however, it is also redefining competitive advantage, often shaped by access to critical resources.

Today, these differences are increasingly geographic, with uneven access to compute, data, and infrastructure influencing who can innovate at scale. India’s response has been clear and constructive. Through the IndiaAI Mission, India is working to democratise access to AI capabilities by treating compute, data, and models as digital public enablers rather than concentrated assets. By building open and scalable digital infrastructure, India aims to ensure that AI advances inclusion, supports sovereign innovation, and delivers shared progress.

This approach is underpinned by a comprehensive, end-to-end vision. At the World Economic Forum in Davos,  Union Minister Ashwini Vaishnawspoke about India’s five-layer AI framework, recognising that meaningful AI deployment requires coordinated progress across the energy layer, compute and infrastructure, chips and hardware, models, and applications.  Anchoring AI development in reliable energy systems and domestic compute capacity ensures resilience, while investments in chips, models, and applications enable innovation that is scalable, secure, and aligned with national priorities. Together, these layers form the foundation of India’s AI ecosystem, linking policy ambition with operational readiness.

What distinguishes India’s approach is its architecture of access. India has created a common compute platform where 38,000 GPUs  are accessible through a unified portal at an affordable rate of INR 65 per hour compared to the global cost of $2.5 to $3 per hour . The Government, under the IndiaAI Mission, now offers 100 per cent  subsidies for foundational model training and 40 per cent subsidies for inference workloads, ensuring that cost does not become a barrier to innovation. This transforms compute from a commodity into an enabler, allowing domestic startups like Sarvam AI, Soket AI, and Gnani AI to build indigenous large language models without the capital intensity that typically favors Big Tech.

Building indigenous large and small language models (LLMs and SLMs) is central to India’s broader objective of algorithmic sovereignty. India’s sovereign AI modelexpected to be unveiled at the India AI Impact Summit (16-20 February 2026)is being trained on Indian datasets and hosted on domestic infrastructure. By grounding these models in India’s linguistic and cultural diversity, the effort also seeks to reduce bias and ensure AI systems are more representative, and inclusive in nature.

This approach ensures that data residency, model behaviour, and governance frameworks remain aligned with national priorities, citizen rights, and India’s constitutional values.It reflects a comprehensive strategy for technological self-reliance and strategic autonomyalong withthe capacity to collaborate globally while retaining control over critical digital systems.  The India AI Impact Summit will shift the global AI conversation from abstract safety debates to measurable impact.

Its seven thematic pillars; democratising AI resources, enabling inclusive development, harnessing AI for planetary resilience, ensuring trustworthy systems, developing contextual multilingual AI, empowering Global South voices, and bridging the AI divide, directly address the dependencies that have historically constrained developing nations.

India’s approach offers a replicable case study for the Global South. For regions facing constraintsranging from infrastructure gaps to large informal economiesit demonstrates how openness and innovation can be balanced with sovereignty. Rather than prescribing a single model, India’s experience illustrates how nations can build AI ecosystems that are aligned with local realities while remaining globally interoperable.

India’s sovereign AI models, being developed as open-source resources, are designed for adaptation across diverse contexts. Countries can draw lessons from this approachadapting model architectures, governance frameworks, and capacity-building strategies to suit their own linguistic, cultural, and developmental priorities.

The emphasis is on ecosystem-building, where shared standards, open tools, and collaborative learning accelerate impact without creating new dependencies. As AI becomes increasingly central to economic competitiveness and national resilience, the question is no longer whether developing nations should invest in AI capabilities, but how they do so responsibly and sustainably.

The Author is Joint Secretary, Ministry of Electronics & Information Technology