AI Boom to Quadruple US Data Center Power Use by 2035

The rapid buildout of generative AI infrastructure is turning electricity supply into a critical constraint for the technology industry. Data centers require large volumes of reliable, around-the-clock power, placing additional strain on public grids, transmission networks and the pace of new generation. The expansion is also sharpening concerns over who pays for grid upgrades and whether costs associated with serving technology companies could raise electricity bills for households and other businesses.

As grid connections become harder to secure and power-cost disputes intensify, more data center operators are considering on-site fuel cells and natural-gas generation to improve energy independence and reduce delays. Such systems could give operators greater control over reliability, but they also raise questions about capital costs, emissions and dependence on fossil fuels. The available event information does not identify specific companies, investment amounts or project dates, but signals that the AI infrastructure race increasingly hinges on access to power.

U.S. electricity consumption is rising again after years of stagnant growth, driven primarily by data center expansion for AI training, cloud computing and cryptocurrency mining, alongside demand from manufacturing and electrification. The U.S. Energy Information Administration (EIA) said these facilities, which fall within the commercial sector, are reshaping a power-use mix long dominated by households while straining grid expansion and electricity costs.

In its Short-Term Energy Outlook released on July 7, 2026, the EIA projected total electricity consumption would rise from 4.195 trillion kWh in 2025 to 4.269 trillion kWh in 2026 and 4.399 trillion kWh in 2027, setting consecutive records. Commercial consumption is expected to reach 1.550 trillion kWh this year, surpassing residential consumption of 1.508 trillion kWh for the first time. Solar power's share of generation is projected to rise from 7% to 9% in 2027, while wholesale electricity prices are estimated at $45 per megawatt-hour this summer.

France’s Regulatory Authority for Electronic Communications, Postal Affairs and Print Media Distribution, or Arcep, said the continued expansion of generative AI training, inference and cloud services has made data centers a major source of new electricity demand. As facilities grow to hyperscale, their power consumption and carbon emissions are affecting regional grid development, energy allocation and France’s environmental sustainability policies.

Arcep’s latest survey showed that electricity consumption by French data centers rose 38% over the past three years, with large AI data centers in particular creating new pressure on energy planning. Available information on the event did not specify the report’s publication date, total electricity consumption or investment amount, but the 38% increase underscores that AI infrastructure is expanding far faster than overall power demand.