US electricity consumption is on track to set consecutive annual records as AI data centers add to demand pressure from the electrification of heating, transportation, and industry.

According to the EIA's latest projections, US electricity demand will rise from a record 4,195 billion kilowatt-hours in 2025 to 4,288 billion kWh in 2026 and 4,356 billion kWh in 2027. As Reuters has reported, the expansion is being driven by AI data centers and cryptocurrency, along with growing use of electricity for heating and transportation. The EIA revises these outlooks through the year, so subsequent updates will show whether data-center demand is tracking ahead of, or behind, current expectations.

Commercial demand sets a record as prices climb

Commercial power sales, a category that includes data centers, are projected to reach a historic high of 1,549 billion kWh in 2026, while residential sales are expected to reach 1,541 billion kWh and industrial use 1,055 billion kWh. A record in the commercial category underlines how much weight data-center construction now carries in the national power picture.

Electricity is also getting more expensive. The EIA expects wholesale prices to average $52 per MWh in 2026, an increase of 11% over 2025, largely due to extreme weather conditions. Regional divergence is stark: PJM prices are forecast to climb 41%, while Mid-Columbia prices are expected to decline 23%. PJM is the regional grid operator spanning parts of the mid-Atlantic, Midwest, and South, while Mid-Columbia is a wholesale power-trading hub in the Pacific Northwest, and the spread between them means companies planning power-hungry facilities face sharply different cost environments depending on where they build.

The pressure is being felt worldwide. According to Gartner's projections, electricity consumption in data centers will increase 26% in 2026, from 447 TWh in 2025 to 565 TWh, and may exceed 1,200 TWh by 2030. For scale, that 565 TWh figure is roughly 13% of projected US electricity demand for the same year. AI-optimized servers are expected to consume 31% of data-center electricity this year, and global data-center power demand is projected to rise from 132 GW in 2026 to 290 GW by 2030.

Grid limits could decide where AI gets built

That makes access to electricity a competitive advantage. According to Berkeley Lab's estimates, US data centers could account for up to 11.8% of the country's total electricity consumption by 2030, within a scenario range of 9.5% to 15.3%; the reference case reaches 649 TWh.

The IEA's mid-year report stated that electricity demand worldwide will rise 3.6% in 2026 and 3.8% in 2027, higher than the 3% recorded the previous year.

For the AI sector, chips are no longer the sole hurdle. Grid connections, energy contracts, energy generation capacity, and permitting now shape how and where new computing facilities get built.

Why Bitcoin miners sit in the middle of this

That shift is spilling into crypto infrastructure. As Cryptopolitan reported, Bitcoin miners already control energized sites and grid connections that AI developers need. The US interconnection queue — the waiting list of power projects seeking grid access — stands at around 2,600 GW, a backlog that helps explain why already-energized sites are drawing attention from AI developers.

Valuations reflect the change: CoinShares pegged stabilized AI facilities in one recent transaction at roughly $27 million per megawatt, compared with below $3 million for miners' energized but unleased capacity. The gap shows how much of the value in this market has shifted toward ready access to power.

The economics are moving fast. In the AI infrastructure race, the scarce asset may not be computing power itself, but the electricity connection needed to switch it on.