Keeping the Lights On #8
The Transmission Service Provider adds a new set of data into the ERCOT Adjusted Load Forecast, connecting generation via transmission to load demand for reliable low-cost electric power for Texas.
ERCOT Planning Process – Demand, Capacity, Transmission
ERCOT was in the news after the April 8, 2025 Board Meeting presentation. A key topic, the first inclusion of the Transmission Service Provider supplied TSP - Load Demand Data was integrated into the overall 2025 ERCOT Adjusted Load Forecast.
New to this annual data set of the TSP Load-Demand Forecast is the inclusion of “TSP Officer Letters” Load-Demand. This refers to new demand likely to connect to the grid, even if no agreements are currently in place. With these Officer Letter inputs, the TSP Demand Forecast for 2030 demand more than doubled from the 85,508 MW peak demand record set August 10, 2023.
Such a dramatic increase makes good media headlines, while consumers worry about losing their power as the hot summer overloads AC systems. Don’t panic, ERCOT plans for the long term and there is time to “grow into the real demand.”
All this data is included in the 2025 Long-Term Demand and Energy Forecast (LTDEF) for the ERCOT region. The LTDEF is comprised of six forecast components: Economic Base Load Forecast, Electric Vehicle Forecast (EV), Behind the Meter Rooftop Photovoltaic Forecast (PV), Large-Flexible Load Forecast (LFL), Large Load Contracts, and Large Load TSP - Officer Letters. The LTDEF uses the waterfall method to combine each forecast to create the ERCOT Net Forecast.
Figure 1: TSP Provided Load Forecast Process
Figure 2: ERCOT received TSP Provided Large Load Breakdown
ERCOT considers the “TSP Officer Letters” Load-Demand inputs in comparison to current load, forecasts and actual performance in order to best estimate future demand in ERCOT Demand Forecast.
For data center loads specifically, ERCOT's adjustment methodology modifies the TSP- Load Demand Data including:
· Delaying in-service dates by 180 days for all new large loads, based on historical experience showing actual energization dates average 220 days later than projected
· Reducing new data center demand to 49.8% of requested capacity, reflecting observed utilization patterns
· Decreasing officer-attested loads to 55.4% of projected demand, based on performance data from facilities energized by February 2025
Figure 3: ERCOT Adjusted Large Load Forecast Methodology
These adjusted forecasts guide crucial planning activities, including the Capacity Demand and Reserves (CDR) Report, Regional Transmission Plan development, and Resource Outage Scheduling.
The 2025 ERCOT Adjusted Demand Forecast shows an 8% increase in 2030 demand over the prior year estimate. More insight into this methodology later.
Figure 4: ERCOT’s adjusted summer peak forecast after Adjustment to Large Load Forecast
ERCOT’s December 2024 CDR report (published in February 2025) incorporating the ERCOT Adjusted Load Forecast shows two potential reserve margin issues: “a -6.2% planning reserve margin for summer 2026 in the net peak load hour and a 5.2% planning reserve margin for summer 2026 in the peak load hour.”
Spoiler Alert: There are sufficient power generation and reserves as well as load demand reduction incentives for the ERCOT grid to reliably deliver electricity through summer of 2026.
Typically, ERCOT provisions supply, demand and reserves in the year-ahead market with ongoing adjustment through the day of delivery. In a later newsletter we will detail the multiple mechanism to ensure grid reliability.
Why include forecast not under contract with commitments to delivery of electric power? ERCOT is responsible under guidance, regulation and law to plan delivery of reliable electric power in an open market with fair and open processes and data-driven transparent processes. ERCOT needs to understand how large the market can be and the TSP has the closest supplier relationship with the load demand.
Reserves for Reliability, empowered by real-time monitoring and responsiveness
ERCOT maintains grid reliability, first by sampling frequency every 5 seconds, expecting system-wide conformance at 60.00 Hz. When load exceeds supply, frequency will drop and the under-frequency measurement triggers supply to the grid from on-call frequency up generation or storage.
As a load goes off-line such that generated power exceeds the load demand, more power is generated than needed. More power than demend increases the frequency, tripping limits and automatically bringing on-line Battery Energy Storage Systems (BESS) to store the power for later usage, thereby dropping the frequency as generator outputs reduce in response to the decrease in overall demand. BESS systems are very quick to respond; while thermal generation may need several minutes to increase or decrease the power generated due to the inertia of the large mechanical generators supplying power.
Every 5 minutes (288 times per day), real-time, ERCOT dispatches resources based on economics and reliability to meet system demand. This Security Constrained Economic Dispatch (SCED) system allows real-time market evaluation of offers to produce a least-cost dispatch of online resources. SCED calculates location marginal prices (LMP) using a two-step methodology that applies mitigation to resolve non-competitive constraints.
ERCOT's demand response framework provides critical flexibility for managing system peaks and unexpected constraints. These programs allow loads, including data centers, to "reduce or modify electricity use in response to instructions or signals" from ERCOT. Data centers can participate directly in ERCOT markets or indirectly by voluntarily reducing energy usage during high wholesale prices.
Large data center operators often negotiate specific load management agreements, allowing ERCOT or the contracted TSP to count on predetermined load reductions during grid emergencies. This capability helps "preserve system reliability, enhance competition, mitigate price spikes, and encourage the demand side of the market to respond better to wholesale price signals".
Open Market, Data Transparency
Eight load regions, almost 2,000 market participants in generating and transmitting power to about 1,000 large load customers, another million industries and commercial businesses have meters and eight million meters at homes.
Weather is accumulated through the eight geographic load regions, with hourly records maintained back to 2008. All of this data is maintained and available for real-time analysis, data decisions on operating efficiently, as well as trend analysis.
ERCOT is on the forefront of managing such system complexity, rich with real-time data, while efficiently connecting generation to load cost-effectively and reliably.
Demand for AI, requires much larger data centers
There is unlimited demand for more computing because of the competitive advantages as AI achieves ever-elusive breakthroughs, e.g. curing cancer, sustainable fusion. Keeping track of the data centers is of value for the development of a quality long range forecast that guides investment strategies for cost-effective reliability.
In times of scarcity, end demand may have multiple engagements points at any one time, trying to find the “best” way to market with favorable economic factors. Not all demand will be realized, not a valid business case, or speculative with no lasting connection to demand.
For example an entity building Data Centers may work in two different cities, with different TSPs and local governments in the negotiation to secure the best match to requirements.
As companies push to add capacity quickly in response to their opportunities and needs, any forecasting process encourages duplicity of demand solutions. Multiple solutions are pursued in parallel, as the Data Center works with multiple parties all with inputs that make or break the deal.
Finally, the “winning” deal with the highest confidence from the “first to operation” planning marathon emerges. Now the data center decides about the current view of their future demand. Considerations are time to market, and ability to serve the demand.
If demand is still strong, the second (or third) site is still relevant, needing resources. If this supplier thinks market demand is slowing, or too risky for their investors, they would drop an alternate site in the next round of forecast updates.
Again, in times of scarcity, hoarding is a natural response and requires significant involvement to accurately assess demand, albeit constraints for toilet paper, eggs, computer chips or electric power.
As such the ERCOT “adjustments” of the TSP-load demand is a realistic “sizing” of real demand, requiring hands-on experts reviewing the data and validating “resizing” decisions for best results.
In parallel to the need for performance, the underlying technologies included in data centers are scaling up to delivery more performance, albeit with additional demand for electricity.
Conditions that create the demand in specific geography, e.g. land, power access are enduring – if the first deal doesn’t work, there will be another that will.
Changing Data Center Trends
2022 data center area averaged about 10,000 square feet, with server racks requiring 4 – 17 kW each, Total site power was between 10 – 100 MW. Benefiting for proximity to large numbers of users, fast internet connections, and electric power drove most of Texas 330+ data centers to urban areas.
Today’s best servers offer 30-50 the performance, but require four times as much power each. Thus, a data center requires 300 MW to 1,200 MW for a facility with 250,000 and 1,000,000 square feet of floorspace. Land is too expensive in most urban locations and power demand possily subject to “peak-demand” load management curtailments.
Rural land parcels with fast internet and nearby 138 kV substation access enable timely addition of critically needed computing capacity. Otherwise, demand is delayed waiting for transmission and transformer availability. Urban area difficulty in adding transmission capacity starts with negotiations for right-of-way access to power and land. These additional delays lengthen the time and thus risk to the project success.
Typical costs to connect are one million dollars per mile from an existing substation that has available capacity for the demand. If the substation needs expansion or a new site, delays and costs can mount quickly. High voltage substation transformers have multi-year lead times and can cost over a million dollars each.
Summary
Overall, the quality of ERCOT Adjusted Large Load Breakdown forecast and likelihood of adequate future power to support demand improves with the ongoing evolutions in the ERCOT forecast methodology. The positive is a more inclusive data set with a more complete view on coming demand needs for electric power. The downside is many of these may still be speculative and not have concluded all decisions necessary to break ground and start the program.
ERCOT takes the TSP Demand Forecast inputs from the Transmission Service Providers serving Texas. Then ERCOT adds historical agreements, and in consideration of key economic trends adjusts load for EV (Electric Vehicles) and Roof-top Solar (Utility grade Solar included in Capacity Demand Reserves report)
Annually ERCOT builds a Long-Term Load forecast. The Long-Term Load forecast is an hourly forecast for the next 10 years. It is based on forecasted economic data and historical weather from 2008–2022.
Adding the Regional Transmission Plan, ERCOT then aggregates the nearly 2,000 participants engaged in generation, delivery and insuring power is available with a twice a year release into a snapshot showing Capacity, Demand and the Reserves necessary reliability of the ERCOT grid.
Below are these planning report by date of most recent release and report name:
April 7, 2025 ERCOT Demand Forecast report (annual)
February 13, 2025 ERCOT Capacity, Demand & Reserves report (2 x per annum)
January 2025 ERCOT Regional Transmission Plan (RTP) 345-kV and Texas 765-kV Strategic Expansion Plan Comparison (RTP released annually)
Together all three plans define ERCOT strategy of providing reliable energy in an open market to power the Texas economy.
These adjusted forecasts guide crucial planning activities, including the Capacity Demand and Reserves (CDR) Report, Regional Transmission Plan development, and Resource Outage Scheduling.
We will next explore responsiveness to outages and then follow with a discussion about AI and why the demand for hyper-data centers requires more than one gigawatt of power.
Paid subscriptions fund the resources necessary for quality coverage. I appreciate every reader and ask you to share the newsletter with interested parties as well as consider becoming a paying subscriber.
May 6, I will join ERCOT’s Innovation Summit to get the in person update key for getting the right information to subscribers.
Thank you for your time and interest.
References
1. 2025 ERCOT System Planning, Long-Term Hourly Peak Demand and Energy Forecast, April 08, 2025, accessed from internet 04/20/2025, https://www.ercot.com/files/docs/2025/04/08/2025-LTLF-Report.pdf
Accessed April 10, 2025 https://www.texastribune.org/2025/02/13/texas-power-grid-ercot-energy-forecast/\
Accessed April 8, 2025 https://www.ercot.com/calendar/04072025-Board-of-Directors-Meeting
Specifically Agenda item 8.1 https://www.ercot.com/files/docs/2025/04/07/8.1-Long-Term-Load-Forecast-Update-2025-2031-and-Methodology-Changes.pdf