The debate around AI infrastructure in India is shifting, and that change is overdue. Instead of focusing solely on scale, speed, or global competitiveness, attention is now being directed toward access. That shift was clearly reflected in a recent white paper released by the Office of the Principal Scientific Adviser (PSA) to the Government of India, which outlined a vision for making AI infrastructure broadly available across regions, institutions, and use cases.
At its core, the document argues that AI infrastructure should be treated as a shared national resource. This framing matters. When access to computing power, datasets, and model ecosystems is limited to a narrow set of actors, innovation tends to cluster. Over time, those clusters harden into structural barriers. By contrast, wider access allows experimentation to occur where local needs are best understood.
AI Infrastructure in India and the Case for Shared National Resources
According to the PSA, democratizing AI infrastructure in India means ensuring that core building blocks- compute, data, and tools are affordable and reachable for a broad group of users. That includes startups, academic institutions, public agencies, and innovators outside major urban hubs. In this model, AI development is not confined to a few global firms or metropolitan centers.
This approach has been framed as especially important for India, given its linguistic diversity and varied development needs. Local-language tools, assistive technologies, and region-specific applications are unlikely to emerge at scale if access remains centralized. Therefore, AI infrastructure is being positioned as an enabler of inclusion, rather than a reward for scale alone.
Physical Constraints Facing AI Infrastructure in India
The white paper also highlights a critical imbalance. While India generates nearly 20 percent of the world’s data, it accounts for only about three percent of global data center capacity. That gap creates dependency risks and limits domestic experimentation.
Physical AI infrastructure, including data centers, GPUs, TPUs, and specialized processors, remains unevenly distributed. Mumbai and Navi Mumbai dominate capacity, followed by Chennai, Bengaluru, Hyderabad, Delhi NCR, Pune, and Kolkata. As a result, geographic concentration continues to shape who can realistically access high-performance compute.
To address this, planned expansion under the IndiaAI Mission was referenced, including the creation of a secure GPU cluster featuring 30,000 next-generation units. These resources are intended for sovereign and strategic applications, but they also signal recognition that domestic compute capacity must expand rapidly.
Digital Public Infrastructure as a Path Forward for AI Infrastructure in India
Beyond hardware, the PSA places strong emphasis on digital public infrastructure, or DPI. Under this framework, AI systems are treated as digital public goods. Access to data, compute, and foundational models can be enabled without physical proximity, provided the right digital layers exist.
Notably, the paper does not advocate a single centralized platform. Instead, a modular design is recommended. Public-good enablers, such as registries, metadata standards, access protocols, and directories, are proposed as early building blocks. These lighter components are viewed as practical entry points that can be implemented quickly.
Over time, attention is expected to move toward more complex systems. Consent-based data sharing, structured data access mechanisms, and coordinated compute exchanges are highlighted as next steps. This phased approach reflects an understanding that governance and trust must evolve alongside technical capacity.
Why Timing Matters for AI Infrastructure in India
Importantly, the white paper stops short of proposing formal policy changes. That restraint appears intentional. AI infrastructure in India is still developing, and early design choices tend to persist. If access considerations are addressed at this stage, future expansion may occur with fewer structural corrections.
From a policy perspective, this matters because retrofitting inclusivity into mature infrastructure is difficult. Regional gaps, once entrenched, are costly to reverse. By contrast, early coordination around standards and access norms can reduce friction later.
A Commentary on Direction, Not Just Deployment
Taken together, the PSA’s vision signals a strategic pause. Instead of rushing toward scale alone, a question is being raised about who benefits from AI growth. AI infrastructure in India is being framed as foundational civic capacity, not merely industrial input.
That framing deserves serious attention. As AI systems become embedded in governance, healthcare, education, and public services, the infrastructure behind them shapes outcomes. Access patterns influence whose problems are addressed and whose voices are reflected in systems.
The white paper does not claim to have final answers. However, it establishes a direction. If followed through with execution and coordination, this approach could help ensure that AI infrastructure in India supports innovation across regions rather than reinforcing existing divides.
