Texas Grid Buckles Under AI Data Center Surge as Power Prices Jump 9x
ERCOT faces 226 GW in pending interconnection requests—mostly from data centers—while capacity auction prices hit federal limits and wholesale rates climb 45% year-over-year.
Texas’s electric grid operator received 226 gigawatts in new large-load interconnection requests through November 2025, nearly quadrupling the 152 GW requested across the entire 2022-2024 period, as AI data centers overwhelm a power system that cannot expand fast enough to meet demand.
The surge transformed ERCOT’s capacity auction prices from $29 per megawatt-day in 2024-25 to $270 per megawatt-day in 2025-26—a 9.3x increase that hit the federal price cap of $329/MW-day in December 2025 for the second consecutive year, according to Seminal Analysis. Wholesale electricity prices at the ERCOT North hub jumped 45% year-over-year in Q1 2026, reaching $50+ per megawatt-hour from pre-pandemic lows near $20/MWh, per Energy Ogre.
Data Centers now account for 77% of ERCOT’s interconnection queue by request volume, totaling roughly 165 GW. A single three-month period in Q2 2025 saw 78 applications requesting over 70,000 MW—more capacity than the entire state consumed during its 2024 summer peak of 86 GW. The grid operator projects total electricity demand will climb from 87 GW in 2025 to 145 GW by 2031, with data centers driving 46% of that growth, rising from 29.6 GW in 2024 projections to 78 GW by 2030.
generation queue cannot match demand profile
While ERCOT’s generation interconnection queue totals 432 GW as of October 2025, 77% consists of solar and battery storage—intermittent sources that cannot provide the constant baseload power data centers require. Natural gas projects total just 48 GW, creating what analyst Dave Friedman termed a “dispatchable power mismatch” between the 165 GW of data center requests and available firm generation.
The state added 22.5 GW of new generation capacity between January 2024 and early 2025 and plans roughly 100 GW in solar and battery additions by 2030. But that represents only 11% of the current large-load queue, and battery storage provides hours of backup rather than the 24/7 uptime hyperscale facilities demand.
“We have outgrown the process that was established for reviewing these large loads,” Kristi Hobbs, ERCOT’s vice president of system planning and weatherization, told industry analysts. The average consumption per site runs only 49.8% of requested megawatts, indicating substantial uncertainty around which projects will materialize—what grid planners now call “phantom loads” in the queue.
Texas operates the only major U.S. grid largely disconnected from interstate networks, limiting its ability to import power during shortages. ERCOT serves 90% of the state’s load across a deregulated wholesale market where generators compete directly. The model attracted data center developers with fast permitting and historically low prices, but grid expansion requires multi-year transmission builds that cannot keep pace with demand growth measured in months.
residential ratepayers face cost pressure
Texas residential electricity rates hit 15.87 cents per kilowatt-hour in Q1 2026, with wholesale prices driving estimates that household bills could rise 25-70% over the next five years. The U.S. Energy Information Administration projects average wholesale electricity prices at the ERCOT North hub could reach 78.9% above the baseline $47.39/MWh forecast by 2027 under high data center demand scenarios.
Large-load customers, primarily data centers, consumed 54 billion kilowatt-hours in 2025—a 60% increase from 2024. Their capacity requests grew from 41 GW in December 2023 to 63 GW by late 2024, then exploded to over 165 GW by November 2025. Areas with concentrated data center development saw electricity prices spike 267% compared to baseline rates, analysis from Bloomberg of Northern Virginia markets found.
“Without mitigation, the data centers sucking up all the load is going to make things really expensive for the rest of Americans.”
— David Crane, CEO of Generate Capital and former Biden energy official
Texas lawmakers responded with Senate Bill 6 in June 2025, requiring large loads to pay $100,000 interconnection fees, disclose duplicate requests, maintain 50% onsite backup generation, and accept remote curtailment during grid emergencies. The legislation shifts some costs to data center operators but does not address the underlying capacity deficit.
hyperscalers bypass the grid
Interconnection backlogs now exceed 10 years for new projects, compared to historical timelines under two years. Nearly 50% of planned U.S. data center builds in 2026 faced delays or cancellations due to power infrastructure shortages, per Tom’s Hardware reporting in May 2026. Gas turbine manufacturers carry 5-7 year backlogs, creating bottlenecks even for projects pursuing onsite generation.
OpenAI’s Stargate campus in Abilene reached 1.2 GW capacity by 2026—equivalent to powering 840,000 to 1.2 million homes—while Microsoft adds roughly 1 GW of global data center capacity every three months. These facilities increasingly pursue behind-the-meter generation rather than wait for grid interconnection, fundamentally altering the compact between utilities and their largest industrial customers.
| Metric | 2024 | 2030 Projection |
|---|---|---|
| Total Consumption | 183 TWh | 426 TWh |
| Share of U.S. Electricity | 4.1% | ~9-10% |
| Growth Rate | — | +133% |
Wood Mackenzie analyst Ben Hertz-Shargel described the dynamic as a “land grab” where companies believe “access to more capacity for compute will be necessary to win the future battle over AI services.” The Center for Strategic and International Studies warned that electricity supply constraints could undermine U.S. competitiveness in AI development if projects shift to regions with available power.
competitive paradox emerges
Texas attracted AI Infrastructure investment with deregulated markets, fast permitting, and historically cheap electricity. That competitive advantage now inverts: the same features that enabled rapid deployment prevent the coordinated, decade-scale grid expansion hyperscalers require. ERCOT’s load forecast shows summer peak load could reach 368 GW by 2032 under current request trends, but the Public Utility Commission rejected the initial projection as inflated.
The uncertainty extends beyond Texas. PJM Interconnection, the nation’s largest grid operator serving 13 states, faces similar strain with a 241 GW backlog. “The current situation is not tenable,” PJM CEO David Mills said in congressional testimony. American Electric Power CEO Bill Fehrman warned that without process reforms, “we could still be having these same conversations in 10 years.”
National data center electricity demand is projected to grow 133% by 2030, from 183 terawatt-hours in 2024 to 426 TWh. A single AI-focused hyperscale facility now consumes electricity equivalent to 100,000 households annually, according to Pew Research Center analysis.
- ERCOT’s interconnection queue vastly exceeds available dispatchable generation, with 77% of new supply coming from intermittent solar/battery rather than baseload power
- Capacity auction prices surged 9.3x in one year while residential electricity rates face 25-70% increases over five years as data centers drive half of Texas’s projected load growth
- Interconnection backlogs exceeding 10 years force hyperscalers toward onsite generation and out-of-state alternatives, bypassing the utility model entirely
- U.S. grid constraints could shift AI infrastructure investment offshore if domestic power supply cannot match compute demand growth
what to watch
ERCOT’s revised long-term load forecast awaits final Public Utility Commission approval after initial projections were rejected as inflated. The realization rate—how many of the 226 GW in requests actually connect—will determine whether Texas faces managed growth or acute capacity crisis. Current data suggests roughly 50% of requested megawatts materialize, but that still implies 110+ GW of new load against 48 GW of firm generation in the pipeline.
Monitor Senate Bill 6 implementation through 2026-27 to assess whether cost-shifting and backup generation requirements slow data center applications or simply push projects to pursue fully independent power systems. Capacity auction results for the 2026-27 service period will signal whether prices stabilize near federal caps or retreat if speculative projects cancel.
Broader grid operator reforms at PJM and other RTOs will test whether interconnection backlogs represent a temporary bottleneck or structural constraint on U.S. AI infrastructure expansion. If natural gas turbine backlogs extend through 2030 and permitting timelines remain at decade-scale, expect accelerating divergence between regions that can deliver power and those that cannot—with direct implications for where the next $200 billion in hyperscale capital deploys.