This is the second article in Yes Energy’s summer demand series, which covers demand-side market complexities. In the first article, Alex Bennitt, Product Manager for Yes Energy, breaks down macro-level demand trends, load projections, and market uncertainty that’s impacting traders, asset developers, and resource planners. In this installment, he focuses on skyrocketing demand and projections in ERCOT, with themes that resonate with ISOs across the United States.
Power markets are entering an era of complex load growth after two decades of near-flat demand, driven by data center buildout, and messy interconnection queues. Add in a rapidly changing generation mix and things start to get spicy. Check out the first installment of the series for a high-level overview on load growth, data center capacity, and generation capacity buildout across the United States.
For the second installment of the series, we’ll focus on demand forecast projections and demand growth in ERCOT, which is at the extreme end of the spectrum compared to other regions across the United States.
The Texas grid, the majority of which is operated by ERCOT, is a frontier when it comes to rapidly changing grid dynamics. On the supply side of the equation, favorable wind and solar economics combined with plentiful land options have made ERCOT the leader in grid-scale wind, solar, and solar-complementing battery capacity. On the demand side of the equation, ERCOT also leads the way, with population growth, commercial and industrial (C&I) growth, oil and gas (O&G) growth, and data center growth sending load to higher and higher levels over the last eight years. Peak load has grown 18 GW over a 5-year period.
Factor in ERCOT’s market design, lack of FERC oversight, and other macroeconomic factors unrelated to energy markets, and it’s no wonder ERCOT leads the way in almost any electricity market specific view. Am I biased? Maybe… I’m from Austin, TX, and it’s the market I know the best. Outside of my ATX roots, I think the data speaks for itself. For an energy market enthusiast, it’s the most exciting market to follow. The frontier occurring in ERCOT on both sides of the market equation is a lens into the future for other ISOs that are bound to see similar market dynamics showing up soon.
Rapid change like this presents a major challenge. How do you forecast demand growth when a new space race is taking place with respect to AI compute? ISOs have long been forecasting growth across base case scenarios, electrification, BTM PV, C&I, O&G (where applicable), etc., but now large load interconnection has placed increasingly wide error bars on forward-looking forecasts. In ERCOT alone, there’s more than 400 GW of large load interconnection requests (~4X peak load). ERCOT’s adjusted forecast is 50 GW of peak demand growth in the next 5 years (30 GW of which is data center driven).
Adding to the load growth uncertainty is federal, state, and local policy. In Texas specifically, the Texas legislature has entered the game with Texas Senate Bill 6 (SB 6), which gives ERCOT the authority to forcibly curtail large loads during an Energy Emergency Alert 2 (EEA 2), altering their firm peak load forecast in the Capacity, Demand, and Reserves (CDR) report by ~40 GW with this scenario.
It’s not far-fetched to say that these forecasts are capturing a potentially significant moment. A new space race has kicked off when it comes to AI compute, and hyperscale data centers are a significant driving force behind hard-to-comprehend forecasts. Is this a potential bubble? Will there be a boom and bust cycle? The amount of uncertainty is a big opportunity for companies to position their forward-looking view accurately, and seize a market opportunity as a result.
ERCOT is the most extreme example of current load growth happening in the United States. From 2011-2021, peak load slowly crept up in 1-2 GW increments, rising 9% in 10 years. In 2019, the new record (74.5 GW, +1.2 GW) pushed ERCOT operations to the limit, and prices reached the (at the time) $9000/MWh cap on two separate days in August 2019.
What was “slow” growth in ERCOT from 2011-2021 is relative to the levels seen the last couple of years. In 2022 and 2023, back-to-back years of ~6 GW peak load growth tested ERCOT operations once again, with prices spending 80 hours above $1000/MWh in 2023, including 5 hours of prices pinned at the (at the time) $5000/MWh price cap. +11 GW of all-time peak demand served in just a two-year period.
2026 continued the load growth trend with another ~6 GW increase to all-time demand set July 22, 2026 with a ~91 GW peak. Yet… changing supply-side dynamics barely tested ERCOT operations. Prices have barely flinched in 2026, with only 10 hours of real-time prices above $100/MWh, not even remotely getting close to the current $2000/MWh real-time price cap, despite operating at levels above the previous all-time demand peak for over 200 hours this summer! (A topic worthy of a separate discussion in its own right).
The strong load growth observed in ERCOT over the last decade was very much in line with the load projections put together in the CDR report. Looking at the demand forecast evolution from the CDR report helps shed light on projections vs actuals.
From 2016-2023, ERCOT CDR forecasted an average of ~6 GW peak load growth over a 5-year time horizon. For 2022, that resulted in an average forecasted peak of 79.2 GW vs observed actual of 79.8 GW, almost spot on. For 2023, observed actuals of 85.4 GW came in +4 GW strong vs the forecast of 81.3 GW.
During this period, the slope of the forecast remained largely the same, even after the strong load growth observed in 2022 and 2023. The December 2023 CDR did show the first indication of what was potentially on the horizon, with the 10-year forecast effectively tripling the long-term growth rate.
ERCOT’s December 2024 CDR kicked off the era of hard-to-comprehend load projections. The report projected ~141 GW peak load by 2029, representing an 8x increase over the previous decades 5-year forecasts (5-year growth of 48 GW vs 6 GW in previous versions). In 2025, the trend continued, albeit pushed back by a year. 2026 should see another strong projection, but potentially at a slightly lower rate as 2025 saw a reduction in slope.
An 8x increase in a demand forecast year-over-year (YOY) is an astonishingly high number, representing a major shift in how ERCOT is treating large load interconnection. This shift isn’t unique to just ERCOT. ISOs across the United States are sending a strong signal to the market that the load growth era is here. Over the next 5 years, most ISOs are projecting all-time peak load to grow 5-25%. That’s all-time peak load! Considering some ISOs are having operational challenges even getting close to previous all-time levels, this is surely a concern.
Part of the problem is that interconnection queue data is messy. In ERCOT, their adjusted load forecast has 62% of the growth consisting of data centers and crypto (~30 GW). The ISO currently has ~10 GW of nameplate data center capacity, half of which is crypto. Data center buildout took off in 2024 and 2025, with 2-3 GW of nameplate data center capacity brought online each year. This doubled data center capacity in just a two-year period. The forecast would see construction rates increase 2-3X to 6-9 GW of nameplate capacity brought online each year, sending nameplate capacity to 39.7 GW by 2030.
What’s lacking in most ISO queue data is out-of-the-box explainability with regard to turning 400+ GW into 30 GW projected over the next 5 years. ERCOT is by a wide margin the most transparent ISO when it comes to market data and explainability, but even then, there’s a lot of manual work needed to recreate their numbers. In other ISOs, you may not even get full access to the queue itself, and only get a simplistic explanation of their load forecasts.
Thankfully, our in-house queue data at Yes Energy provides out-of-the-box value with demand center queues, with projected capacity easily obtained and allowing for filtering out the noise in the queue. Our view is a little more bearish in ERCOT with respect to nameplate interconnected in the next 5 years by 3-4 GW. Yes Energy’s view uses a different methodology vs ERCOT that can be applied across ISOs. Our infrastructure team has tracked a lot of projects, and helped us project online dates and assign a percent chance a facility connects to the whole queue, vs ERCOT looking at TSP-provided contracts and letters, then scaling those facility amounts down by ~50% and delaying interconnection by ~18 months.
Note: Data center historical and projected capacity provided via our in-house infrastructure tracking capabilities
To focus more closely on the near term, ERCOT’s 2026 peak load forecast sat at 94 GW. Once again, ERCOT has a lot of transparency on how they reached this value, yet one thing is missing! We don’t have the ERCOT model itself to run different scenarios. 94 GW is a static number provided to the market. It represents an aggregated number, using normal weather inputs from the last 17 weather years, across base case, EV, large flexible loads (LFLs), BTM solar, and large load additions (via TSP contracts and officer letters; full methodology can be found on ERCOT’s load forecasting page under the 2025 Long-Term Load Forecast Report tab).
What about percentiles? Or non-normal weather? Our in-house demand forecasting at Yes Energy helps layer in more explainability with respect to demand forecasts. Plus… we run 14x more weather scenarios through our model! Shifting historical weather by +/- 7 days to account for holidays, weekends, or extreme weather events that occurred potentially a couple days before. Both these views allow us to strip out weather effects and assign percentile ranges to the model outputs, yielding temperature departure from normal load impacts, and load level percentile a year out in advance. To put this summer into context, our P90 value in ERCOT sat at ~90 GW, and a 94 GW load level would need to bring the heat with a +6-degree-F departure from normal!
ERCOT's demand forecasts have shifted from a decade of steady, unremarkable projections to the fastest-accelerating numbers in the country, driven largely by data center growth. Making sense of any single figure, though, requires context — from generation queue probabilities to the weather and percentile assumptions baked into the forecast itself.
With load projections climbing and uncertainty running high across markets, having accurate data to guide your strategy matters more than ever. From Infrastructure Insights to EnCompass’ comprehensive forecasting models, Yes Energy gives you the tools to navigate what's ahead.
If you want to learn more about how Yes Energy's datasets can help you scale your demand projections and position your business, watch the full webinar or request a demo.
Next up in our summer demand series: How hyperscale data centers actually operate on the grid and a crypto mining case study.