US data centres could consume one-fifth of electricity by 2035 as the rapid expansion of artificial intelligence (AI) infrastructure reshapes the nation’s energy landscape, according to a new BloombergNEF report. The analysis projects that U.S. data centres will require nearly 200 gigawatts (GW) of capacity within the next decade, with approximately 20% of the country’s total electricity generation dedicated to powering these facilities by 2035. Nearly half of this capacity is expected to support AI training and inference, reflecting the explosive growth of generative AI technologies.

BloombergNEF also forecasts that the United States will host 64% of the world’s AI chip power demand by 2033, reinforcing its position as the global hub for AI computing. However, this unusual growth is expected to place enormous pressure on electricity grids, increase energy costs, and accelerate investments in power generation.

AI Is Driving an Unusual Surge in Power Demand

Artificial intelligence has rapidly evolved from a specialized computing application into one of the world’s most energy-intensive industries.

• Training and operating large AI models require vast computing resources, prompting technology companies to invest billions of dollars in building larger and more powerful data centres.
• BloombergNEF’s latest outlook represents an 83% increase compared to its electricity demand forecast published in December, highlighting how quickly AI adoption has exceeded previous expectations.

Other industry organizations have reached similar conclusions.

• The Electric Power Research Institute (EPRI) has more than doubled its projections since 2024, while S&P Global has increased its forecasts by over one-third within just a few months.

Regional Power Grids Face Mounting Pressure

The rapid expansion of AI infrastructure is expected to strain electricity networks that already have limited spare capacity.

• The PJM Interconnection, which manages electricity across parts of the eastern and midwestern United States, could see 34% of its electricity consumption coming from data centres by 2035.
• Similarly, ERCOT, the grid operator serving most of Texas, is projected to allocate approximately 22% of its generating capacity to data centres over the same period.

PJM has already faced significant challenges in accommodating new electricity demand, including a four-year suspension of new grid connection applications while it worked through a backlog of projects.

Also Read: France’s Floating Igloo Embarks On Historic Polar Exploration Mission

Electricity Prices Are Already Rising

Growing demand is beginning to affect electricity markets.

BloombergNEF reports that wholesale electricity prices within the PJM market have increased by 76% over the past year, while data centres accounted for 38% of bids in the grid operator’s latest capacity auction.

These trends suggest that expanding AI infrastructure will require substantial investments in power generation, transmission networks, and grid modernization to maintain reliability.

Also Read: Environmental CSR Spending By Indian Companies Crosses ₹17,000 Crore In 10 Years

AI’s Energy Challenge Goes Beyond the United States

While US data centres could consume one-fifth of electricity by 2035, the energy demands of AI are becoming a global concern. Under an aggressive AI growth scenario, BloombergNEF estimates that worldwide data centres could generate an additional 1,935 terawatt-hours (TWh) of electricity demand by 2033, almost equal to India’s current annual electricity consumption.

As AI continues transforming industries, governments, utilities, and technology companies will increasingly face the challenge of balancing innovation with sustainable energy development. Expanding renewable energy, improving grid infrastructure, and developing more energy-efficient computing technologies will be essential to supporting the AI revolution while limiting its environmental impact.