AI Data Centers Spark a Power Surge, Prompting Technical, Regulatory and Political Responses

The rapid expansion of artificial-intelligence (AI) workloads is forcing data centers to become major electricity consumers. The Department of Energy projects that U.S. data centers could use up to 9 percent of national generation by 2030, up from 4 percent in 2023. Lawrence Berkeley National Laboratory estimates a 9.5-15 percent share of total load by decade’s end. In response, the Federal Energy Regulatory Commission (FERC) issued orders to shorten interconnection reviews to 90 days, a drastic cut from the multi-year timelines that previously slowed AI-driven builds.

Clean Energy Resources to Meet Data Center Electricity Demand

AI’s rise has outpaced Moore’s law; transistors can no longer shrink below 4-7 nm, so chip designers are lowering operating voltages to 0.5-1 V. Lower voltage reduces heat but, per Ohm’s law, requires higher current, increasing conversion losses in point-of-load (POL) converters that feed GPUs.

New technology would boost energy efficiency at data centers

• DOE: data-center electricity demand could rise 15-20 percent over the next decade.

Clean Energy Resources to Meet Data Center Electricity Demand

• LBNL: 9.5-15 percent of U.S. electricity use by 2030.

Clean Energy Resources to Meet Data Center Electricity Demand

• FERC: 90-day target for grid interconnection; 48-V DC bus cuts ohmic loss 16-fold.

US Acts to Speed Up Power Grid Hook-Ups for AI Data Centers

• Binghamton prototype: 10-12 percent higher efficiency across loads, slew-rate doubled.

New technology would boost energy efficiency at data centers

• Bloom Energy: 1.2 GW of fuel-cell capacity deployed for AI sites.

Fuel cell maker Bloom Energy surges 40% over six trading days as AI data centers drive power demand surge

Researchers at Binghamton University have patented a single-stage POL converter that steps 48 V down to 1 V within 5 mm of the chip, delivering 10-12 percent higher efficiency and faster power delivery. Bloom Energy’s solid-oxide fuel cells provide on-site, low-CO2 electricity that can be installed in 90 days, bypassing grid bottlenecks. GE Vernova supplies natural-gas turbines sized for AI farms, while Redwood Materials repurposes used EV batteries into a 12 MW/63 MWh micro-grid that powers a 2 000-GPU AI cluster in Nevada.

Old EV batteries are quietly becoming the cheapest way to power AI data centers

FERC Chair Laura Swett declared, “This is the biggest priority our country is facing at the moment.” The bipartisan Ratepayer Protection Act, introduced by Rep. Kathy Castor (D-FL) and Rep. Gabe Evans (R-CO), would require utilities to allocate upgrade costs to large-load customers rather than to residential ratepayers. State operators such as PJM and NYISO are under pressure to justify tariffs for megawatt-scale users.

Florida Politics

A bipartisan backlash has emerged: Texas Gov. Greg Abbott and New York Gov. Kathy Hochul have each advocated stricter local controls on AI data-center siting. Environmental groups cite unpermitted natural-gas turbines at xAI’s Mississippi site and a federal push to revive coal plants as evidence of regulatory leniency for high-power users.

Grist

A working paper posted to arXiv argues that “data centers caused average rates to fall modestly, not rise,” contradicting political narratives that blame AI farms for higher bills. The study notes that future supply constraints could reverse this effect, highlighting a need for transparent cost-allocation mechanisms.

The study that blows up the biggest argument against AI data centers

Grid operators are drafting new interconnection standards, while companies continue to field on-site power solutions and explore orbital data-center concepts. Legislative debate over the Ratepayer Protection Act and further FERC rulemaking will shape how the nation balances AI-driven compute growth with affordable, reliable electricity.

US Acts to Speed Up Power Grid Hook-Ups for AI Data Centers