AI forcing a rethink on data centre needs, say experts at BS Manthan

GPU-led AI workloads are increasing data centre power use up to tenfold, forcing companies to rethink energy sourcing, efficiency, and long-term renewable power strategies

Artificial intelligence (AI) is rapidly changing the energy profile of India’s data centres, sharply increasing their power consumption and forcing companies and policymakers to rethink how the sector will be powered in the coming decade. The shift is being driven by the growing use of graphics processing units, or GPUs, which consume far more electricity than traditional computing systems, industry executives and energy experts said at BS Manthan in New Delhi on Tuesday.

 

“From 6 kW from earlier, it has turned into 60 kW now,” said Sunil Gupta, co-founder, managing director and chief executive officer of Yotta, referring to the power consumed by a single rack in an AI-enabled data centre. He further said that GPUs consume at least 10 times more power than conventional processors, significantly increasing both electricity use and cooling requirements.

 

 

How has AI changed the energy demand of data centres?

 

Data centres have always consumed large amounts of electricity, but artificial intelligence has accelerated that trend. Gupta said demand was already rising due to the spread of digital services such as UPI, video streaming, and cloud computing.

 

“Anything we are using digital in our lives, at the backend is going to some datacentre,” he said.

 

Historically, data centre racks consumed 6–7 kilowatts of power, a level that remained stable for about 25 years.

 

Artificial intelligence has disrupted that pattern.

 

AI workloads rely heavily on GPU-based systems, which dramatically increase power density. This has also raised cooling needs, making energy efficiency a central concern for operators.

 

The surge in power demand is also changing long-term projections. India’s data centre capacity, which stood at about 200 megawatts in 2016, has grown to 1.4 gigawatts today. Earlier projections estimated capacity would reach 3 gigawatts by 2030. With AI, estimates now range between 6 and 12 gigawatts.

 

Can renewable energy meet the rising power demand?

 

The experts said renewable power can play a central role in supporting this expansion.

 

Mohit Bhargava, country head, India Energy & Climate Center at UC Berkeley’s Goldman School of Public Policy, said large facilities can already run entirely on renewable energy. “A 200 megawatt data centre operating on a round-the-clock basis can be easily powered,” he said.

 

He added that renewable plants can be built in less than two years and at lower cost than coal, gas, or nuclear power. Long-term contracts can also provide stable electricity prices for 15-20 years, reducing exposure to fuel price fluctuations.

 

India also has structural advantages, including land availability and faster transmission deployment compared to countries such as the United States, Bhargava added.

 

What role will policy and infrastructure play?

 

Nikhil Dhingra, chief executive officer of ACME Solar, said renewable companies require aggregated demand and clear planning to build distribution infrastructure. He added that renewable energy already accounts for 25-30 per cent of consumption for many companies.

 

Bhargava said regulators can support the transition by ensuring easier connectivity to transmission networks and maintaining policy clarity.

 

Will data centres strain India’s power system?

 

Despite rapid growth, data centres still account for a small share of national electricity consumption. Gupta said data centres currently consume about 10 terawatt hours of electricity, compared to 1500 terawatt-hours consumed nationwide.

 

By 2030, data centres could account for about 3 per cent of total power consumption.

 

Gupta said India’s main constraint is unlikely to be generation capacity, but last-mile delivery. “The last mile distribution availability would be the real requirement,” he said.