Digital infrastructure as foundation of India's AI ambitions

This article is authored by Ankit Goel, founder, Constl & chairman and founder, Space World Group.

India is undergoing a decisive shift in its AI journey as we are no longer content with being a consumer of AI built elsewhere. Our national agenda is transitioning towards building the infrastructure required for India to emerge as a global hub for AI innovation and deployment. This is not a subtle shift in policy emphasis; it represents a fundamental rethinking of where India sits in the global technology order. The question driving our national conversation is no longer whether India will be an AI economy, but what kind of AI infrastructure will actually power that economy and whether we are building it fast enough.

AI

The government’s actions reflect this shift with initiatives such as the IndiaAI Mission, backed by a budget outlay exceeding ₹10,300 crore over five years and the deployment of 38,000 GPUs nationwide. This reflects a broader shift in India’s approach to AI. The conversation is no longer limited to technology adoption; it is increasingly focused on creating the infrastructure required to support AI innovation, deployment, and long-term digital growth at scale.

According to Omdia, India's data centre market has expanded nearly tenfold over the past five years and is expected to exceed 5 GW by 2028, with estimates by Jefferies projecting growth to nearly 8 GW by 2030. These numbers signal that India is rapidly becoming a critical node in the global AI ecosystem.

India's expanding data centre footprint is more than just a measure of the country's digital growth; it reflects the scale of the infrastructure being built to power the country's AI future. Behind every AI workload lies a vast physical infrastructure layer and data centres form its foundation, providing the compute, power, storage, and connectivity that turn AI from concept into reality.

However, the data centre infrastructure that powers AI is not simply about adding more capacity. Unlike traditional digital applications, AI workloads place unprecedented demands on infrastructure. The ability to support high-density compute environments, move vast volumes of data efficiently, and deliver reliable performance is becoming as important as the intelligence of the models themselves.

Moreover, as AI models become larger and more sophisticated, data centres are evolving from isolated facilities into interconnected environments that must operate in close coordination. This makes the quality and performance of the connectivity infrastructure between them just as important as the compute and power resources. In the AI era, real-world performance will depend not only on what happens within a data centre, but also on the efficiency of the Data Centre Interconnect (DCI) networks linking them.

The models being trained and deployed today comprise hundreds of billions of parameters and demand vast compute and power requirements that standalone data centres are often unable to handle. As AI models grow in scale and complexity and inference demands multiply across billions of daily requests, workloads will increasingly become complex and distributed across multiple data centre facilities operating in harmony across geographies. With AI clusters scaling across locations, the volume of East-West traffic--the data exchanged between interconnected AI systems and data centres--will increase significantly, placing unprecedented demands on network infrastructure.

Consequently, AI workloads are fundamentally reshaping network requirements and raising the performance expectations placed on digital infrastructure. They demand extremely low latency, sustained high bandwidth, and near-perfect reliability - simultaneously and without compromise. AI is driving rapid adoption of 400G and 800G interconnect speeds, with hyperscalers already preparing for the transition to 1.6T networking.

Latency, in particular, has graduated from a technical benchmark to a business-critical variable. In distributed AI environments, even milliseconds of delay compound into performance degradation. The network infrastructure is no longer a utility that forms the foundation of AI, it is now a strategic enabler of AI outcomes at scale.

The next phase of AI growth will be determined by dense, high-capacity networks designed to support hyperscale and distributed AI workloads. As AI systems continue to scale, optical fibre remains the only medium capable of meeting these growing demands of bandwidth, latency, and reliability.

This is where India faces a gap that deserves honest acknowledgment. Legacy fibre infrastructure in India was built to support consumer broadband, mobility services, basic backhaul requirements, and traditional enterprise applications. While these networks have successfully enabled the country's digital transformation, they were not designed to meet the performance, scale, and traffic demands of the AI era.

Simply upgrading legacy infrastructure for hyperscale and AI applications is not a viable path. The architecture simply cannot be retrofitted to meet modern demands. Instead, we need to fundamentally rethink how these networks are designed, implemented, and managed.

Several design principles are emerging as non-negotiable for networks that genuinely serve AI workloads. Networks must be purpose-built with high-capacity fibre paths that follow the shortest possible routes, supported by diverse routing architectures that minimize the risk of outages. They must also leverage low-loss fibre to preserve signal integrity over long distances, while incorporating the scalability needed to support future growth from the outset.

These are not incremental improvements to existing networks. They represent a new class of greenfield fibre infrastructure built for hyperscale and distributed AI--not derived from architecture designed for a previous generation of digital requirements.

As AI scales, infrastructure will increasingly become a differentiator. The nations and organisations that invest in resilient, high-capacity digital networks today will be better positioned to unlock the full potential of AI tomorrow.

India's AI ambitions are no longer a question of intent; they are increasingly a question of execution. The infrastructure decisions being made today will shape what our AI economy looks like in 2035 and beyond. We are investing in compute, attracting investments by global hyperscalers, and creating the foundations of a vibrant AI ecosystem. The next big aspect of India's AI success will be whether the network layer keeps pace with the scale and complexity of AI-driven demands.

The decisions we make today will influence far more than network performance. They will shape India's ability to innovate, compete, and lead in an AI-driven world. Building AI-ready infrastructure is therefore not simply a technology imperative, it is a strategic investment in the country's digital future.

(The views expressed are personal)

This article is authored by Ankit Goel, founder, Constl & chairman and founder, Space World Group.