Last year, ERCOT expanded what it includes in its load forecasts. Had the grid operator stuck with the same methodology this year, it would have led to shocking predictions as to how much energy the grid would need in the next five years. Instead, ERCOT last week revised its methodology to provide more realistic forecasts. ERCOT CEO Pablo Vegas called it a “sanity check.”

Context: House Bill 5066

Following the enactment of House Bill 5066 in June 2023, ERCOT started including in its long-term load growth forecast all proposed development in the grid’s interconnection queue, even projects without signed contracts or interconnection agreements.

While this change allowed ERCOT to plan for new load farther in advance, it also likely inflated the total load growth forecast by including speculative projects that may not ultimately get built.

Indeed, as we reported last fall, 80% of the market participants we surveyed said ERCOT’s load growth forecasts were too high.

"Adjusted" Load Forecast

The updates announced last week seek to improve the accuracy of the forecasts by introducing a dose of skepticism to projects, specifically data centers, in the queue.

ERCOT looked at interconnection data from the past two years and concluded that:

• Projects come online about 6 months later than they request
• Only about half of the data center load shows up
• Only about half of the load reported by transmission companies shows up

ERCOT is now including these findings in what it calls an “Adjusted Load Forecast,” and the implications are significant.

Using the previous, unadjusted methodology, ERCOT would have forecasted 2030 peak load at 208 GW, up from 148 GW last year. Instead, the newly introduced adjustments decrease the 2030 peak load forecast by 10 GW, to 138 GW.

What accounts for the massive discrepancy? According to Vegas, its developers racing to secure sites as they compete to serve the big tech companies looking to build data centers.

“They are scouting for and shopping for locations and sites recognizing there is going to be a huge load demand,” Vegas said at a Texas hearing. “If they can have site-ready sites to then sell to an end customer, they’re going to be well positioned to gain as much of that growth as possible.” 

In other words, developers are submitting multiple interconnection requests in ERCOT – and likely beyond it too - for what ultimately may be a single data center project they each hope to build.

Follow the Yellow Brick Load

Where does ERCOT believe this adjusted load growth will materialize?

The North weather zone, where there’s been a surge of data center development activity, is predicted to see the steepest increase, from 2 GW this year to 8 GW in 2031, or a 26% compound annual growth rate. The West weather zone comes in second, with a compound annual growth rate of about 23%.

Remaining Uncertainty

While the adjusted forecast adds clarity to how ERCOT views the future, it leaves important questions outstanding.

First, is it reasonable to consider the past two years of interconnection data as representative of the most likely load growth in the future? Should we assume the pattern of the last two years - that only 50% of data center load materialized – will continue, especially if data center demand increases?

Second, unlike other large loads, data centers have the potential to be responsive to price, meaning they can scale back their consumption during peak hours. Given this, how much of data center load should be included in peak load forecasts?

Third, as more and more data centers get added to the grid and increase the price of electricity for everyone, will we see a greater proportion of data centers seeking behind-the-meter solutions? How should this be factored into the load forecast?

ERCOT did a “sanity check” last week, but it still has much to figure out.