Data centres, which already account for an estimated 2–3% of India’s electricity consumption, are expanding rapidly, with capacity expected to grow by nearly 5x double by 2030, per S&P Global. As AI workloads increase energy intensity, rack densities that traditionally operated at 8–12 kW are now moving toward 100 kW and beyond. This is pushing the country’s energy grid to its limit.
"The industry is moving from a compute race to a power race," observes Ankit Saraiya, Director and CEO of Techno Digital.
As digital infrastructure scales alongside rising household demand, it calls the need for a unified, digital, and intelligent energy network that can be dynamically managed, with energy traded and optimised in real time.
A Unified Energy Interface (UEI) is designed to bring interoperability and coordination to a rapidly decentralising energy ecosystem. It creates a universal platforms that connects enables electricity demand to interact seamlessly with sources.
While no Indian state has formally implemented a full-scale UEI, Bengaluru has become a key hub for UEI-linked EV charging and energy coordination pilots, with platforms such as Bharat e-Charge using Beckn-based open-network standards for interoperable charging discovery and payments. Industry groups linked to the UEI Alliance are also working with DISCOMs on EV charging and distributed energy coordination systems.
“Much like how UPI transformed payments by creating a common interoperable layer, a similar digital interface for energy could enable households and businesses to track generation, trade surplus power, and interact with the grid seamlessly in real time,” says Karthik Raju, Executive Director at Atria Renewable Private Limited, in conversation with AIM.
As rooftop solar adoption accelerates, he argues, consumers are no longer just users of electricity—they are becoming “prosumers,” capable of generating, consuming, and even selling power.
Overhauling Outdated Energy Systems
This, however, introduces a level of complexity that traditional grid systems were never designed to handle. For decades, electricity flowed in one direction: from large, centralised power plants to end users. Grid operators could plan generation and distribution in advance, relying on predictable demand patterns.
India’s energy transition is not just about moving from coal to renewables; it is also about decentralisation. “We are moving to a world that generates electricity from wherever they sit, such as putting solar panels on our rooftops,” explains Prof Praphul Chandra, Director, Atria Centre of Excellence in AI & Decentralised Technologies. “There might be times during the day when we are generating more electricity than we need.”
This creates a fundamental challenge. Unlike other commodities, electricity cannot be easily stored at scale. “At any point in time, the amount of electricity being pulled out from the grid must be equal to the amount of energy being put into the grid,” he notes. “Now, this is a very difficult problem to solve in a chaotic environment where you have energy generators everywhere, electric vehicles plugging in at any time, and data centres whose demand can spike unpredictably.”
This is where UEI and AI can help. These can offer real-time demand–supply balancing, automated decision-making in peer-to-peer energy markets, dynamic pricing, predictive maintenance of assets, and grid stability management.
"The most forward-thinking operators are no longer asking how to consume power more efficiently. They are asking how to control their own power environment entirely," Saraiya says, pointing to the growing adoption of renewable energy, battery storage, and AI-driven energy management systems.
In a decentralised grid, supply and demand are inherently variable. Solar generation fluctuates with weather conditions, while consumption patterns vary by time, location, and user behaviour. AI systems can ingest vast amounts of data, from weather forecasts to historical usage patterns, and make real-time decisions to optimise the system.
“Whether it is your solar panel, your battery system, or a data centre, each of these needs to forecast its supply and demand, and then interact with others using the protocol,” Prof Chandra says.
Managing Flexible Loads
A promising aspect is the emergence of AI agents embedded at different points in the grid. These agents can forecast energy generation and consumption, decide when to store or use power, and even participate in energy trading.
“AI agents can analyse historical consumption patterns, weather forecasts, solar generation data, and grid behaviour to accurately predict energy production and demand,” Raju explains. “Over time, they will function as digital energy managers, continuously optimising energy flows across decentralised networks.”
In practical terms, this could mean a household solar system automatically selling excess power to a nearby building, or an electric vehicle delaying charging to off-peak hours based on price signals. Prof Chandra describes this as a system of “flexible loads.”
“If I plug in my electric vehicle at 5 pm and it is going to stay plugged in till 7 am, I am absolutely fine if you delay my charging,” he says. “But a commercial taxi driver may not have that flexibility.” AI can differentiate between such scenarios, optimising energy usage without compromising user needs.
UEI + Data Centres
Data centre operators are also increasingly responding by building their own energy ecosystems. According to Techno Digital, data centres in some of India's largest digital infrastructure clusters already account for around 15% of grid capacity, a figure that could rise to 30–35% by 2030.
Rather than relying solely on the grid, many facilities are integrating renewable energy, battery storage systems, and captive power generation into microgrids that can dynamically manage supply and demand.
“In data centres, some workloads must be attended to right then and there, but others can potentially be delayed,” Prof Chandra explains. Training an AI model, for instance, can often wait a few hours, while real-time applications cannot. By aligning non-urgent workloads with periods of high renewable energy generation, sAI can help reduce strain on the grid.
“The fundamental problem is that in a country like India, we have huge energy during the daytime because of the sun, but much less from renewables at night,” he says. “Can data centres automatically manage their load based on energy being generated? That’s the problem.”
For data centre operators, that challenge is increasingly shaping how computing workloads are distributed. Today, cloud systems allocate workloads based on latency and compute availability. In the future, energy availability may become just as important. "Workloads will be optimised based on renewable energy availability, grid stress conditions, cooling efficiency, and the carbon intensity of supply," says Saraiya.
Large AI training workloads, which are less latency-sensitive, could be scheduled around periods of abundant renewable energy, while inference workloads remain closer to users. In this model, energy becomes an active operational variable rather than a passive utility input.
Building such a unified energy interface requires coordination across multiple stakeholders, from regulators and utilities to technology providers and consumers. Here, AI can be both a major energy consumer and a critical tool for managing energy.
“Such a system would make energy markets more transparent, efficient, and participatory,” says Raju. “It would unlock peer-to-peer trading, improve grid responsiveness, and empower consumers.”
Your reaction