Texas AI Buildout: The Queue Survived - Even Thrived - with SB6
Texas wrote the toughest large-load law in the country, but the interconnection queue got bigger. That fact alone says a lot about data center demand.
Last June, Governor Abbott signed Senate Bill 6 into law. The stated goal: make data centers pay their own way, protect ratepayers, and give ERCOT a kill switch for emergencies. The bill applies to any load over 75 MW, which means it applies to essentially every AI data center project in the state.
SB6 is not symbolic. A 1 GW data center campus now faces $50 million in upfront financial security before a single interconnection study begins. Add the $100,000 screening fee, mandatory site control documentation, disclosure of parallel queue positions, and a remote disconnect mechanism that lets ERCOT cut your power during firm load shed.
The reasonable expectation: this slows things down. The “bragawatts” are supposed to exit. Speculative queue-stuffing stops. The pipeline contracts to something closer to reality.
That last part happened. The first part did not.
The filter worked but the pipeline didn’t shrink.
Oncor told the Texas House Committee on State Affairs it was tracking 650 large-load requests totaling 273 GW. That was the raw queue: every developer who filed paper, serious or not.
On April 15, ERCOT filed the 2026 Preliminary Long-Term Load Forecast in PUCT Project 58777. The filtered number, applying SB6-adjacent criteria (site control, financial commitment, realistic ramp schedules), came in at 110 GW for Oncor alone and 243 GW statewide.
So SB6’s filter removed 163 GW of phantom load from Oncor’s queue. It worked exactly as designed.
What remained: 243 GW of filtered demand against an 85 GW grid. Nearly 3x peak. After the filter.
ERCOT’s own cover letter to the filing says the quiet part out loud. The grid operator has “concerns with using the preliminary load forecast values” for its own reliability assessment. It is asking the PUCT for permission to discount the numbers further using “actual historical realization rates.”
They basically said ‘we filtered the queue, and we still cannot plan around what is left’.
Why the queue grew anyway
SB6 did exactly what it was designed to do. It separated serious developers from bragawatts. It removed 163 GW of phantom load from Oncor’s pipeline alone.
And yet, the queue got bigger.
It didn’t grow because the filter failed, but because the economics of building in ERCOT are so compelling that new serious entrants replaced every speculative one that left.
Texas has cheap land, abundant gas, terrific solar and wind resources, a market structure that lets large loads contract bilaterally without capacity market overhead, and faster permitting timelines than most other states.
SB6 raised the cost of getting in line. It did not change the reason people are standing in it.
The result: 243 GW of filtered, financially committed demand against an 85 GW grid, after the filter. That is wild.
The kill switch is the CHR in legislative form
The most revealing provision in SB6 is the mandatory remote disconnect. Any large load interconnecting after December 31, 2025 must give ERCOT the ability to curtail it during firm load shed.
For a “free” market, that’s pretty interesting. The Texas Legislature looked at data center demand and concluded that the price signal cannot do this job.
This is the Compute Heat Rate™ (CHR)1 made tangible. The CHR quantifies the maximum electricity price an AI data center can profitably sustain: approximately $6,350/MWh blended across workload types. ERCOT’s systemwide offer cap is $5,000/MWh. The ceiling the grid operator can charge is lower than the price the buyer can pay.
A rational AI data center operator will never voluntarily curtail at virtually any price ERCOT can send. The economics do not permit it. SB6’s companion provision confirms this: its voluntary demand response program explicitly excludes loads that curtail solely in response to wholesale prices. The statute was drafted with the understanding that this demand will not respond to price.
So they built a kill switch.
That is not a policy choice you make when you think the market will self-correct. That is a policy choice you make when you have measured the buyer and concluded the buyer is bigger than the market.
The pattern, not the exception
Other states are watching. Virginia, Ohio, Georgia, Oregon, and at least six more are drafting SB6-style frameworks. They will produce the same result: cleaner queues, better cost allocation, and unchanged demand pressure.
SB6 seems like useful governance and every state expecting meaningful data center buildout may want to consider something similar. But no version of it constrains demand.
The policy instrument that would do that would need to price regulatory friction at levels comparable to the CHR itself, thousands of dollars per megawatt-hour. No legislature is going to write that law.
The queue survived SB6 because SB6 was calibrated to the old electricity market. The market of the future has a different price tolerance. Until grid planning, market design, and forward curves internalize that difference, every filter will produce the same outcome: a cleaner list of buyers who can still afford to pay.
Hans Royal is the originator of the Compute Heat Rate™ (CHR) framework. All views are his own and do not represent those of any employer or affiliated organization.
Royal, Hans, The Compute Heat Rate: Quantifying AI-Driven Electricity Price Tolerance
and Its Implications for Wholesale Market Repricing (February 28, 2026).
Available at SSRN: http://dx.doi.org/10.2139/ssrn.6322318