In an interview with CNBC-TV18, Sridhar emphasized that the rapid pace of AI advancements necessitates immediate action in developing the energy systems essential for data centers.
“The rate at which AI is progressing indicates that if India delays for even a few years, the economic fallout of trying to catch up could be quite serious,” he stated.
Sridhar noted that India currently uses approximately 2,000 terawatt-hours of electricity each year, in contrast to China’s 10,000 terawatt-hours. He argued that India must significantly enhance its power generation in a short time frame to stay competitive.
He advised against depending solely on renewable energy sources for the AI-driven economy.
“No single technology — particularly solar and wind, due to their land requirements and competing needs for land and water — can suffice on its own,” he remarked.
Sridhar pointed out that solar and wind power are intermittent and are thus not ideally equipped to meet the steady and concentrated power demands of large data centers.
He proposed an “all-of-the-above” strategy that encompasses nuclear energy, hydro power, solar, wind, and natural gas.
“Is nuclear necessary? Yes. Should we use solar, wind, and hydro? Absolutely,” he confirmed.
Sridhar mentioned that data centers should increasingly depend on dedicated on-site generation since traditional grids are not designed for massive, concentrated, and variable loads.
He cautioned that compelling existing transmission systems to manage such demands could raise electricity costs for other consumers and create reliability issues.
The Bloom Energy leader also underscored the significance of energy security, suggesting that India could diversify its natural gas supplies through friendly nations, including Australia, Canada, the US, and countries in the Middle East.
“Currently, natural gas is a reliable resource,” he said, adding that green hydrogen has the potential to serve as a future alternative.
Sridhar emphasized the necessity for rapid execution in the AI age, highlighting Bloom Energy’s recent project for Oracle, where the company delivered 55 megawatts of power infrastructure in just 55 days, despite an initial commitment of 90 days.
“Our aim is to reduce this to one day, creating a deployment model that is as simple as Lego blocks, allowing for quick implementation and payment only as growth occurs,” he explained.
He added that the capacity for rapid scaling will be increasingly vital as AI infrastructure proliferates worldwide.
