The future of AI isn’t just software — it’s infrastructure

Everyone is talking about what AI can do. Very few are talking about what AI needs.

Behind every large language model, every autonomous agent, and every real-time AI decision sits an enormous physical apparatus. Power. Cooling. Compute. Connectivity. The conversation around AI has largely centred on capabilities and product development, while the harder, less visible, and more consequential work happens in data centres, substations, and building projects. It is both a hardware and software revolution, and the physical layer is finally getting the attention it has always deserved.

The gap in infrastructure nobody talks about

When you speak to most business executives regarding the requirements for scaling AI at their firm, they’ll talk about skills, tools, and money. But power needs are barely mentioned.

Power is fast becoming the greatest limiting factor for AI adoption worldwide. A single AI data centre can use tens of megawatts of power. This increases every time a new user, app or process gets added. The demand for power at data centres is expected to increase by 165% by 2030. The computational appetite of modern AI is not a future problem. It is a present one, and it is growing faster than the infrastructure built to support it.

Data centres are being planned at a pace not seen since the early days of cloud computing. Global data centre investment is projected to reach $1 trillion annually by 2027, driven almost entirely by the infrastructure demands of AI. Chip manufacturers are in a supply crunch that has pushed delivery timelines beyond a year for specialised AI hardware. Power purchasing agreements for renewable energy are at an all-time-high. This is AI’s foundational build out moment, and the companies that control power and physical capacity will shape the next decade of the industry.

Why India cannot afford to stand idly watching

India happens to be one of the fastest-growing markets when it comes to using AI technology. The demand signal is clear and accelerating. But it is unclear if the physical foundation will be able to match up.

India aims to become a developed country by 2047, according to the Viksit Bharat mission. The scope of this goal deserves some thought. The vast majority of the built environment that India will need by then is yet to be constructed. Every campus, industrial park, government facility, and urban development that breaks ground between now and then is an opportunity to embed foundational technology from day one, not retrofit it in later at three times the cost.

India has a genuine structural advantage that most mature economies do not. Ageing infrastructure in Europe and parts of the US requires expensive modernisation to handle high-density compute workloads. India can design its next generation of facilities with AI

infrastructure baked in from the start. That window is open now. It will not stay open indefinitely.

The built world is the next AI frontier

There is a dimension to this story that goes well beyond data centres and cloud capacity. AI has also started to make its presence felt in the physical surroundings, and perhaps nowhere is this more evident than in the building structure within which we spend our days.

Just think about the possibilities today. A commercial campus in Bengaluru that will take care of occupancy, water management, air quality, security, and energy through a single layer of intelligent management. The government campus that works without manual intervention at every stage, be it Parliament House in Delhi, the memorial of Sardar Patel, or any coastal or civic infrastructure. Factories where equipment communicates with management systems before a fault becomes a failure. Hospitals where resource allocation across power, water, and space is handled dynamically in real time.

In each of these settings, AI is not being added onto existing infrastructure as an afterthought. It is being integrated into how the environment functions at its core. There are far-reaching consequences that go beyond energy reduction. With automation comes greater reliability, occupant satisfaction, enhanced security, and tangible efficiency improvements for all the resources consumed by a building. The growth rate of the Indian market in intelligent buildings is an indication of just how quickly this is happening, with an expected CAGR of over 24% and reaching $109 billion by 2033.

The intelligence is not abstracted away in the cloud. It is present in the systems that run the spaces people inhabit. The most consequential AI deployments of the next five years will not look like applications. They will look like buildings.

The strategic asset most organisations are underpricing

There is a tendency to treat infrastructure as a cost centre. Something to be minimised, outsourced, or deferred. In the context of AI, that instinct is a liability.

Organisations building or securing access to intelligent infrastructure today are not just solving an operational problem. In three years’ time, their competitive advantage will not be easy to emulate because of the high disparity between supply and demand. In addition, 78% of businesses are currently applying AI in at least one of their functions. The organisations that control the physical layer underneath that adoption will hold a structural advantage over those that do not. Infrastructure in AI is not overhead. It is leverage.

The same principle also holds at the national level. Those countries who own the layer of the infrastructure of AI will define its access, governance, and commercialisation. That is not a technology question. This is a sovereignty issue.

Building the foundation

The next wave of value in AI won’t be driven by a smarter model or merely by a better user interface. It will rather come from the organisations and nations that did the unglamorous work of building the foundation: the compute, the connectivity, and the physical intelligence layer that makes AI operational rather than theoretical.

India has the rarest of advantages right now. A massive construction cycle ahead. A national vision that demands it. And the technology available today to capture it.

Software defines what AI can imagine. Infrastructure determines what it can actually do.

(The author is Co-Founder and CEO, Enlite. Views are personal.)