7x Better than ERCOT: How Amperon Maintained Accuracy in Unusual Conditions

On July 22, 2026, ERCOT load hit 91.1 GW, absolutely shattering its previous record of 85.5 GW from 2023. This type of load growth cannot be explained by temperature alone. As Amperon has documented, multiple factors, from population growth and electrification to data centers and industrial activity, have been driving Texas' load structurally higher.

Turns out, ERCOT itself hasn't fully accounted for these changes. Even as 100+ degree temperatures scorched Texas, the grid operator was forecasting 89 GW for July 22. This would have still been record-shattering, to be sure. But the gap between the ISO's forecast and the new reality of the Texas grid remains substantial.

Thanks to AI-powered forecasts that continuously retrain on changing grid dynamics as well as weather, Amperon was able to maintain impeccable accuracy during the massive load event. And given the forces at play, it's likely that the 91.1 GW record won't stand for long, so market participants would be well served to stay abreast of market changes.

ERCOT load July 22 2026 vs Amperon and ISO forecasts — *Amperon’s day-ahead ERCOT grid demand forecast predicted the July 22, 2026 peak with a Mean Absolute Percentage Error (MAPE) of 0.32%, compared to the ISO’s 2.31%.*

Here's what made the situation unusual compared to historical records:

Weather played a limited role in ERCOT's record load

Yes, triple-digit temperatures were doing real work. HE17's 101.6°F is genuinely hot, and ERCOT load tracks cooling demand closely on days like this. But compare the July 22 numbers to the scatter plot below: ERCOT total load against temperature for every hour above 90°F since 2020. At 100-101°F, prior years in this sample topped out around 85 GW. This week pushes that ceiling well above 90 GW at essentially the same temperature.

Scatter plot of historical temperatures vs load in ERCOT — *A scatter plot of historical temperatures vs. load in ERCOT shows that prior days over 100 degrees have produced load around 85 GW, as opposed to 90+ GW on July 22, 2026.*

It's not that the weather was unusually hot for Texas in July; triple digits happen most summers. It's that the same heat which used to produce an 85 GW day is now producing a 90+ GW day. The upward drift in the scatter shows the ceiling on ERCOT load climbing year over year at a fixed temperature. Weather sets the shape of the curve; something else has been raising the floor underneath it.

The forces driving record ERCOT demand

Heat still matters. Triple-digit July afternoons push residential and commercial AC hard, and ERCOT's summer peak has always tracked cooling demand closely. That hasn't changed.

Population and housing growth compound it. Texas keeps adding people and rooftops faster than almost any other state, and each new household adds AC load that didn't exist five years ago. This is a steady, unglamorous driver, but it shows up in the baseline every summer.

Data centers are the new marginal load. This is the piece that's structurally different from 2020. Hyperscale campuses, AI training clusters, and crypto-mining operations have been landing across Texas at a pace the grid hasn't seen before, and unlike AC load, this demand doesn't cycle off overnight or ease up in the shoulder seasons. It's flat, large, and growing. That's what's pulling the whole load duration curve upward, not just the summer peak.

The market isn't panicking

Here's the part that should reassure people watching from the trading side: day-ahead prices at the North hub didn't blow out. HE21 settled at $237.83/MWh, a real scarcity premium, but nowhere near ERCOT's $5,000/MWh cap, and nowhere near the kind of pricing we've seen in past tight-reserve events. The market priced in a stressed hour, not an emergency.

That's a meaningful signal. It suggests the incremental generation and reserve margin ERCOT has brought online is, so far, keeping pace with this new baseline of demand. Prices moving up modestly rather than spiking to the cap is the market telling us this looks manageable, for now.

What this means going forward

90 GW likely won't stand as a record for long. If data center buildout continues at its current pace and population growth holds, we should expect ERCOT to test new highs again this year and set a materially higher one next summer.

The question worth tracking isn't whether the next record comes. It's whether reserve margins and new generation keep growing fast enough to keep day-ahead prices behaving the way they did this week.

‍