The Power Equation Behind the Data Centre Race

                The AI Infrastructure Economy                                             The Hidden Input – Part 2

Editor’s note

As the world competes to attract hyper-scale AI investments, electricity is emerging as more than a utility. The reliability, capacity and responsiveness of the power grid could increasingly influence where the next generation of AI infrastructure is built. However, the power requirements of AI rarely dominate discussions in the public domain.

The Hidden Input- Part 1 examined the water needs of large data centres. As the global AI race gathers pace The Hidden Input- Part 2  looks at another critical requirement: POWER. The accompanying decoder examines why data centres need dedicated substations.

The New Power Equation

The first generation of data centres primarily required reliable internet connectivity and uninterrupted electricity. The next generation of AI infrastructure requires something much more demanding.

Training large artificial intelligence models requires tens of thousands of high-performance processors operating continuously for weeks or even months. These facilities consume electricity on a scale that was once associated primarily with heavy industry.

Electricity is no longer just another operating cost.

As AI investments accelerate across the world, electricity is no longer just another operating cost. It is becoming one of the factors shaping where AI infrastructure can realistically be developed. This represents an important shift. For decades, electricity policy focused largely on meeting residential demand, industrial growth and economic development.

Artificial intelligence is now the new consumer.

Hyper-scale AI data centres require not only enormous quantities of electricity but also exceptionally high levels of reliability. Even brief interruptions can disrupt operations, damage expensive equipment or delay AI training workloads. The conversation changes from simply generating more electricity to delivering electricity differently.

Capacity matters, reliability matters, transmission capacity matters, substation availability matters. The ability to connect large consumers to the grid within reasonable timeframes also matters.

Several countries are already confronting this challenge. In some regions, proposed data centre projects have reportedly faced delays because transmission infrastructure or grid capacity could not be expanded quickly enough. Elsewhere, governments and utilities are accelerating investments in transmission networks, substations and new generating capacity to accommodate AI-driven demand.

India may increasingly face similar questions.

The country has made substantial progress in expanding renewable energy capacity while simultaneously strengthening transmission infrastructure. Yet AI infrastructure introduces new planning requirements.

The Meta–Reliance data centre project in Jamnagar illustrates how renewable power is becoming closely linked to new AI infrastructure, with the project backed by renewable energy arrangements. At the same time, India is adapting its existing power infrastructure to support grid stability as the electricity mix changes. Nine thermal power units, eight coal- or lignite-based and one gas-based, have been identified for possible repurposing as synchronous condensers to support grid stability. At the state level, Karnataka’s new data-centre policy also links the expansion of data-centre capacity with renewable energy and supporting infrastructure. 

It is no longer sufficient to ask whether electricity is available

It is no longer sufficient to ask whether electricity is available. A more relevant question may be whether electricity can be delivered at the scale, quality and reliability required by hyperscale AI facilities.

This also alters the way states compete for investment. Land and fiscal incentives will remain important. However, states capable of offering dependable power infrastructure, faster grid connections and long-term capacity planning have an advantage and may become more attractive destinations for large AI investments.

Power Equation Beyond Generation

This perspective also broadens the economic opportunity, because the AI infrastructure power grid story is not confined to power generation alone. Transmission companies, substation developers, switch gear manufacturers, transformer producers, grid technology providers, battery storage companies and power management firms all become part of the wider infrastructure ecosystem – the hidden inputs supporting AI.

The discussion around artificial intelligence is very often dominated by breakthroughs in models and computing power that can deliver at an amazing pace and scale. The ability to deliver reliable electricity at an unprecedented scale may prove equally important.

As the AI economy expands, the power grid infrastructure may gradually evolve from being an invisible utility into one of the defining competitive advantages for regions seeking to host the next generation of digital infrastructure.

Decoding The Hidden Input

Why Do Data Centres Need Dedicated Substations?

When a new office building or shopping mall is constructed, it is usually connected to the existing electricity distribution network. But for hyper-scale AI facilities, simply connecting to the local electricity network is often not enough. Many require their own dedicated substations to ensure a reliable, high-capacity power supply.

Large AI data centres operate on a very different scale. Their electricity demand can be so high that the local distribution network may be unable to supply sufficient power reliably. Instead, many large facilities are connected through dedicated substations designed specifically to serve their operations.

What is a substation?

A substation is an electrical facility that receives power from the transmission network and transforms it to the voltage required by consumers. It also performs critical functions such as switching, protection, monitoring and controlling the flow of electricity. Without substations, electricity generated at power plants cannot be delivered safely and efficiently to industries, businesses or homes.

Why can’t data centres use ordinary connections?

The answer lies in both scale and reliability. Modern AI data centres may require hundreds of megawatts of electricity—far more than most commercial buildings. Supplying this level of demand through conventional urban distribution infrastructure may not be practical. Dedicated substations provide a direct, high-capacity connection capable of supporting continuous operations.

Reliability matters as much as capacity

A brief power interruption at home may be an inconvenience. In a hyper-scale data centre, even a short disruption can interrupt computing workloads, affect digital services and trigger complex recovery procedures. For that reason, operators design multiple layers of electrical resilience. Dedicated substations form one part of that resilience strategy.

Are substations enough?

No. Electricity reliability is achieved through several layers. A typical hyper-scale facility may include: dedicated substations, multiple incoming transmission lines, backup diesel or gas generators, battery-based uninterruptible power supply (UPS) systems, intelligent power management systems. Each layer is designed to reduce the risk of service interruption.

Why are substations receiving more attention now?

Traditional enterprise data centres already required dependable electricity. Artificial intelligence changes the scale. As AI facilities become larger and more power-intensive, substations are becoming critical enabling infrastructure rather than routine utility assets.This is one reason why utilities, transmission companies and state governments are increasingly involved in planning large data centre developments.

What does this mean for states?

When states compete to attract AI investments, the discussion extends beyond land availability and investment incentives. the AI Power infrastructure grid—including transmission capacity and the ability to develop dedicated substations within reasonable time frames—can influence how quickly major AI projects move from announcement to operation.

The takeaway

Substations rarely attract public attention, yet they perform one of the most important functions in modern digital infrastructure. They ensure that enormous quantities of electricity can be delivered safely, reliably and continuously to facilities whose operations depend on uninterrupted power. As AI infrastructure grows, substations are likely to become an increasingly familiar part of the conversation—not because the technology has changed, but because the scale of electricity demand has.