If you operate an AI data center, Bitcoin mining facility, or high-performance computing campus in Texas, you have almost certainly encountered a line item on your electricity bill labeled something like "4CP transmission charge" or "coincident peak charge." For many operators, this charge represents hundreds of thousands — sometimes millions — of dollars per year. And the frustrating part is that it is determined by just four fifteen-minute intervals spread across an entire summer.
This article explains exactly what ERCOT's 4CP mechanism is, how it is calculated, why it disproportionately affects AI data centers, and what the real financial exposure looks like at different scales of operation.
4CP stands for Four Coincident Peaks. It refers to the four highest demand peaks that occur on the ERCOT grid during the summer months — specifically June, July, August, and September. These four peaks are identified retroactively at the end of the summer season, and they form the basis for calculating transmission charges for every large electricity consumer connected to the ERCOT grid for the following twelve months.
The term "coincident" is important. It does not mean the four highest peaks at your specific facility. It means the four highest peaks for the entire ERCOT grid — the moments when total system demand is at its maximum. Your facility's demand at those exact moments is what determines your transmission cost allocation for the following year.
In other words: what you are consuming during four fifteen-minute windows across four summer months will determine a significant portion of your electricity costs for the entire next year.
The calculation process works in four steps. First, at the end of each summer, ERCOT identifies the single hour in each of the four summer months during which total grid demand was highest — typically hot weekday afternoons between 3 PM and 7 PM Central Time. Second, ERCOT measures your facility's demand during each of those four peak hours using interval meter data, typically as a fifteen-minute average in kilowatts. Third, your four demand measurements are averaged to produce your "4CP demand" figure for the year. Fourth, that figure is multiplied by the applicable Transmission Cost of Service (TCOS) rate, which currently ranges from approximately $50 to $90 per kW per year in most ERCOT regions.
The Formula:
Annual 4CP Charge = Average 4CP Demand (kW) × TCOS Rate ($/kW/year)
For a 100 MW data center at full capacity: 100,000 kW × $70/kW/year = $7,000,000 per year
Seven million dollars. Determined by four fifteen-minute windows.
Traditional electricity consumers — office buildings, retail stores, manufacturing plants — often have natural load variability. AI data centers are fundamentally different. They have several characteristics that make 4CP exposure severe.
Near-Constant Load Profile. AI training clusters, inference serving infrastructure, and HPC workloads run around the clock at near-maximum capacity. Unlike an office building that empties at 6 PM, a GPU cluster running large language model training does not care what time it is. This means AI data centers are almost always at or near full load — including during summer peak events.
Massive Connected Load. A modern AI data center housing NVIDIA H100 or B200 clusters can draw anywhere from 50 MW to 500 MW or more of power. A 500 MW facility running at full load during four summer peak events could face $35 million or more in annual 4CP transmission charges.
Rapid Growth in Texas. ERCOT has received over 220 gigawatts of large load interconnection requests as of early 2026, with more than 70% from data centers. Texas has become the epicenter of AI infrastructure buildout in the United States, making the 4CP mechanism increasingly significant for the industry.
Limited Awareness Among New Entrants. Many AI infrastructure companies entering the Texas market come from cloud computing backgrounds. When they receive their first annual reconciliation bill showing millions in 4CP transmission charges, it is often a significant and unpleasant surprise.
| Facility Size | Connected Load | 4CP Demand (90%) | TCOS Rate | Annual 4CP Charge |
|---|---|---|---|---|
| Small AI cluster | 10 MW | 9,000 kW | $70/kW | $630,000 |
| Mid-size data center | 100 MW | 90,000 kW | $70/kW | $6,300,000 |
| Large HPC campus | 500 MW | 450,000 kW | $70/kW | $31,500,000 |
These numbers assume the facility is running at 90% utilization during all four peak events. In practice, many AI data centers run at 95–100% utilization continuously, making the actual exposure even higher.
4CP events are not announced in advance. The four peaks are only identified after the summer ends, based on actual grid demand data. However, there are strong historical patterns:
Experienced operators monitor ERCOT weather forecasts, day-ahead and hour-ahead load forecasts, real-time LMP price spikes, and ERCOT emergency alerts to predict likely peak events.
4CP charges appear through Transmission Cost of Service (TCOS) charges — pass-through charges from your Retail Electric Provider (REP) or directly from ERCOT if you are a direct market participant.
Critically, 4CP charges are backward-looking. The charges you pay in 2026 are based on your demand during the four peak events of summer 2025. This creates a one-year lag between your operational decisions and their financial consequences — a lag that makes proactive management essential.
Manual Curtailment. Operators monitor ERCOT conditions manually and reduce load when they believe a peak event may be occurring. The challenge is that manual monitoring is imprecise, requires human judgment at inconvenient hours, and often results in unnecessary curtailment on non-peak days.
Financial Hedging. Some operators work with energy trading firms to purchase financial instruments that hedge their 4CP exposure. These products limit downside risk but do not eliminate the underlying charge.
Load Scheduling. Operators with flexible workloads can attempt to schedule their most energy-intensive jobs to run during off-peak periods — nights, weekends, and cooler months.
Automated Demand Response. The most sophisticated approach involves automated systems that monitor real-time ERCOT LMP prices and grid conditions, and automatically reduce facility load when conditions indicate a potential 4CP event. These systems can respond in seconds — far faster than any human operator.
On-Site Generation. Some large operators are installing on-site gas turbines or battery storage that can reduce their net ERCOT grid draw during peak events.
There is an important flip side to the 4CP story: ERCOT's Alternative Demand Response (ADER) program compensates operators who can demonstrate verified load reductions during grid stress events.
This means an AI data center with effective demand response faces a dual financial benefit:
1. Reduced 4CP charges — by curtailing during peak events, you reduce the demand measurement that determines your annual transmission cost
2. ADER revenue — by participating in ERCOT's demand response programs, you receive direct compensation for your verified load reductions
The combination of these two benefits can transform 4CP from a pure cost center into a net revenue opportunity for operators with the right infrastructure.
The operators who understand 4CP mechanics deeply, and who have the infrastructure to respond to peak events automatically and reliably, will have a significant structural cost advantage over those who do not.
Table of Contents
LIVE ERCOT DASHBOARD
See real-time ERCOT LMP prices, grid signals, and verified savings data.
VIEW DASHBOARD →