Introduction
After 15 years of flat electricity demand, artificial intelligence is once more growing the grid. Data center electricity consumption is projected to triple by 2030, reaching 400 to 600 terawatt-hours (TWh) annually, roughly 8 to 12 percent of total U.S. demand. Unlike prior false starts, this demand surge is backed by the world's most creditworthy companies. Driven by AI opportunity, Microsoft, Google, Amazon, and Meta have collectively committed approximately $400 billion in capital expenditure for 2025 alone.
The grid cannot currently serve this demand, and a mismatch in timelines is the culprit. Data centers can be built in 18 to 24 months; connecting them to the grid takes four to five years; building the transmission to deliver power often takes seven to twelve years. Fortunately, these delays are not primarily technical. They reflect permitting regimes, interconnection procedures, and planning processes designed for an era of slow growth. This is compounded by decades of institutional atrophy that have occurred since the regulatory structures that once enabled rapid buildout were dismantled in the 1970s.
Inaction will mean missing out on a substantial opportunity to relieve households that have been increasingly burdened by rising electricity prices. When large loads cannot connect to the grid, they build dedicated generation and cease contributing to shared infrastructure costs, making it harder to justify the price of transmission that could deliver energy to where it is needed and removing the opportunity to spread maintenance costs beyond existing ratepayers. Strain is already reaching consumers: capacity prices in the mid-Atlantic Regional Transmission Organization (RTO) PJM increased tenfold between 2024 and 2025, adding roughly $20 per month to bills in affected regions, with transmission and generation costs soon to rise as well.
Reforms can be split into two categories: those available under existing authorities and those requiring new statutory language. States should protect ratepayers through rate structures that allocate attributable infrastructure risk to data centers rather than captive customers, following Virginia's template of minimum demand obligations and upfront collateral requirements. The Federal Energy Regulatory Commission (FERC) should enforce interconnection penalties without exception, encourage adoption of automated queue processing tools, require evaluation of grid-enhancing technologies before approving network upgrades, and mandate parallel processing of cross-boundary studies. RTOs should improve long-range planning through probabilistic demand forecasting and develop clear procedures for large load interconnection. But the binding constraints are federal: the National Environmental Policy Act (NEPA) applies to any project with a federal nexus, and interstate transmission requires coordination mechanisms that no state can create unilaterally. Congress should pass legislation expanding FERC backstop siting authority for nationally significant projects, creating categorical exclusions for transmission upgrades within existing rights-of-way, and consolidating environmental review to eliminate duplicative processes. Addressing the seven-to-twelve-year construction timeline requires all of the above.
The legislative template for congressional reform already exists. The Manchin-Barrasso permitting and transmission bill passed the Senate Energy and Natural Resources Committee 15–4 in 2024, and bills addressing similar issues are currently pending. The SPEED Act passed the House in December 2025 and awaits Senate action, while the SPEED and Reliability Act addresses transmission siting directly. The relevant committee chairs have identified permitting reform as a priority, industry support continues to build, and a reliability event in Midcontinent ISO (MISO) or PJM could concentrate legislative attention rapidly. The window is open. Whether it stays open depends on choices yet to be made.
