The Scale of the Problem
The artificial intelligence revolution is rewriting the fundamentals of US electricity demand. After two decades of essentially flat load growth, grid operators across the country are confronting a structural shift: data centers now represent the single largest source of new electricity demand, and the pace is accelerating. The EIA projects that data center power consumption could grow from approximately 4% of US electricity use today to between 6.7% and 12% by 2028 — a range that reflects deep uncertainty about how quickly AI workloads will scale.
For commercial and industrial electricity buyers who share the grid with these facilities, the consequences are direct: higher capacity auction clearing prices, accelerated transmission investment (socialized across ratepayers), and localized congestion premiums in zones targeted for data center clusters.
ERCOT (Texas): The Interconnection Scramble
Texas remains the epicenter of data center development, attracted by deregulated markets, low base prices, and abundant land. But the scale of demand has overwhelmed ERCOT’s interconnection process:
- 40+ GW interconnection backlog: ERCOT’s queue of pending data center requests now exceeds the entire installed capacity of many US states. The “Batch Zero” priority system selects projects that can demonstrate grid-readiness within 18–24 months.
- 10–15% annual load growth: Baseline forecasts project 10% annual demand growth for 2025–2027, with high-demand scenarios reaching 15%. This is unprecedented in a market that saw near-zero growth for a decade.
- Price risk: EIA analysis suggests wholesale electricity prices in ERCOT could increase up to 79% under high-demand scenarios by 2027 if capacity additions fail to keep pace.
- New cost-allocation rules: ERCOT is implementing rules for facilities exceeding 75 MW to ensure data center developers bear appropriate interconnection costs rather than socializing them across existing ratepayers.
| Metric | ERCOT | MISO | CAISO |
|---|---|---|---|
| DC Interconnection Queue | 40+ GW | 82 GW* | 4.9 GW† |
| Annual Load Growth | 10–15% | 5–8% | 1.8 GW by 2030 |
| Transmission Investment | TBD | $12.3B (MTEP) | Under review |
| Key Risk | +79% wholesale | 1.9 GW gap | Pulsating loads |
*MISO long-term survey through 2044. †CEC forecast through 2040.
MISO (Midwest): The Reliability Imperative
MISO is approaching the data center challenge from a reliability-first perspective, and the numbers are staggering:
- 82 GW potential load: MISO’s long-term load forecasting surveys indicate up to 82 GW of new data center demand by 2044 — a figure that would roughly double the current system peak.
- $12.3 billion MTEP investment: The 2025 Transmission Expansion Plan (MTEP25) allocates $12.3 billion for infrastructure buildout, with $3.1 billion earmarked specifically for data center load growth regions in the South and West subregions.
- Capacity gap: The 2026/27 Planning Resource Auction opened with a 1.9 GW capacity deficit. If data center load materializes faster than modeled, the gap widens and clearing prices rise — directly impacting commercial capacity charges across Illinois, Indiana, Michigan, and Wisconsin.
- Fast-track interconnection: MISO has processed 24 GW through three fast-track cycles, prioritizing gas-fired generation to close the capacity gap before the 2027 PRA.
CAISO (California): The Pulsating Load Challenge
California’s data center challenge is qualitatively different. While the absolute GW numbers are smaller than ERCOT or MISO, CAISO faces a unique operational risk from AI workloads:
- “Pulsating” loads: CAISO research has identified that AI data centers can swing power demand by 80–90% within seconds — as GPU clusters complete large computational tasks and abruptly shed load before the next batch begins. This creates frequency and voltage instability risks that traditional demand forecasting cannot capture.
- Forecast growth: The California Energy Commission projects data center load on the CAISO grid to increase by 1.8 GW by 2030 and 4.9 GW by 2040.
- Battery saturation risk: California’s 25.6 GW battery fleet was built to arbitrage solar overgeneration, not to buffer high-frequency data center load swings. Whether these assets can serve both functions simultaneously is an open question.
What This Means for Commercial Electricity Buyers
The “data center tax” on commercial electricity bills manifests through three channels:
- Capacity charges: Higher auction clearing prices (PJM already saw a 262% spike in 2025) flow through to all load-serving entities. Commercial buyers in MISO and PJM territory should expect capacity cost increases of 15–40% on their next procurement cycle.
- Transmission riders: The billions in new transmission investment required to serve concentrated data center loads will be socialized across ratepayers through transmission cost allocation mechanisms. ERCOT North and MISO South are particularly exposed.
- Congestion premiums: Localized data center clusters create transmission bottlenecks. Nodal price spreads in the Dallas-Fort Worth corridor, Northern Virginia (PJM), and central Ohio have already widened 20–35% year-over-year.
Procurement Action Items
- Lock capacity costs now: If your contract renews in 2026–2027, negotiate capacity charge caps or fixed-price structures before the next round of auction results lands.
- Evaluate demand response revenue: Grid operators are increasingly desperate for flexible load. Commercial facilities that can curtail during stress events are now more valuable than ever — MISO and ISO-NE are paying premium rates.
- Monitor transmission rider filings: State PUC dockets for transmission cost allocation changes are the leading indicator of future bill increases.
- Consider geographic diversification: For multi-site operations, shifting load to zones with less data center concentration can reduce congestion exposure.
Source: EIA; ERCOT Publications; MISO MTEP25; CAISO Large Load Studies; PJM Independent Market Monitor 2025 State of Market Report; California Energy Commission.