ERCOT tells its board that AI training loads can fatigue grid generators and drafts a limit on how fast their power draw may swing
The Electric Reliability Council of Texas put a list of emerging Large Load risks in front of its own Board of Directors at the September 14th and 15th, 2026 meeting. The presentation names three risks and a mitigation for each, and the three sit at three different stages. Ride through requirements are already in force. The model quality rule cleared the board on September 15th and now waits on the Public Utility Commission of Texas. The third rule has not been written yet. The grid operator states that Large Loads tied to data center AI training processes can swing their power draw fast enough to perturb synchronous generator shafts. It says that swing can cause equipment fatigue or damage. Its draft rule would hold a Large Load's power variation to 10 MW per 5 seconds. The same page states that the dynamic models the grid operator uses to study these loads have proven inaccurate, and that Texas A&M University is building better ones. The presentation itself was for information only and the board was asked to take no action on it.
How to take part
The board meeting the presentation was made to is open to the public and it has now been held. The revision request the presentation names, PGRR144, keeps a public page on the grid operator's market rules site. That page lists every vote taken on it and what body takes it next, which is now the Public Utility Commission of Texas. The draft rule on power variation has no number and no filing date yet. Nothing on it can be commented on until it is submitted, so there is no window a reader can act in today.
Where
Statewide. This decision applies across Texas rather than to a named county.
Timeline
- passed
Technical Advisory Committee recommended PGRR144 for approval
- hearing
Board of Directors meeting at which the risk list was presented
- passed
Board of Directors recommended PGRR144 for approval and sent it to the Public Utility Commission of Texas
- Today
How this decision moved
One dated line per check, oldest first. A line that says nothing changed means somebody looked and it had not.
- 2026-09-15
Admitted on the grid operator's own presentation to its board. Three Large Load risks are named with a mitigation each, and the three mitigations sit at three different stages.
- 2026-09-18
The model quality rule moved and the other two did not. The market rules page for PGRR144 now carries a September 15th board vote recommending approval. It names the Public Utility Commission of Texas as the body that takes it next. The draft rule on how fast a large load may swing its power draw still carries no number and no filing date.
- 2026-09-21
The rule on dynamic model quality is still with the commission rather than in force. The market rules page shows it pending and names the board's recommendation to approve as the last action taken. The Public Utility Commission of Texas is named as the body that takes it next. The presentation the risk list comes from still reads as the record holds it. That goes down to the ramp limit the draft rule will recommend and the modeling work the grid operator is doing with Texas A&M.
- 2026-09-24
The board materials describing generator fatigue from AI training loads still read as the record took them.
- 2026-09-27
The board materials on generator fatigue from training loads still read as posted, and the draft limit on how fast that draw may swing is still pending.
The evidence
Every fact above rests on one of these. The words are the source's own.
Large Loads, particularly those associated with data center AI training processes, are capable of producing rapid fluctuations in active power consumption. These variations can induce perturbations on synchronous generator shafts and potentially result in equipment fatigue or damage.ERCOT, Item 10 Interconnection and Grid Analysis Update, Board of Directors Meeting September 14 to 15, 2026 Primary source, official · ercot.com
Consequently, ERCOT's draft NOGRR, Large Load MW Variation Requirements, will recommend a limit of 10 MW per 5 seconds for power variation in Large Loads.ERCOT, Item 10 Interconnection and Grid Analysis Update, Board of Directors Meeting September 14 to 15, 2026 Primary source, official · ercot.com
Since July 2025, ERCOT collaborated with industry experts and engaged stakeholders to investigate these risks.ERCOT, Item 10 Interconnection and Grid Analysis Update, Board of Directors Meeting September 14 to 15, 2026 Primary source, official · ercot.com
Dynamic models for Large Loads have proven to be inaccurate and thus are inadequate for studying their grid impacts.ERCOT, Item 10 Interconnection and Grid Analysis Update, Board of Directors Meeting September 14 to 15, 2026 Primary source, official · ercot.com
ERCOT is working with Texas A&M University on developing more accurate models.ERCOT, Item 10 Interconnection and Grid Analysis Update, Board of Directors Meeting September 14 to 15, 2026 Primary source, official · ercot.com
Texas A&M has released a new model for crypto mining facilities and is working on a data center model.ERCOT, Item 10 Interconnection and Grid Analysis Update, Board of Directors Meeting September 14 to 15, 2026 Primary source, official · ercot.com
The PUCT approved NOGRR 282/ NPRR 1308 in July, effective on August 1.ERCOT, Item 10 Interconnection and Grid Analysis Update, Board of Directors Meeting September 14 to 15, 2026 Primary source, official · ercot.com
TAC endorsed PGRR 144 in August and is up for Board consideration in September.ERCOT, Item 10 Interconnection and Grid Analysis Update, Board of Directors Meeting September 14 to 15, 2026 Primary source, official · ercot.com
ERCOT plans to submit NOGRR in September or October.ERCOT, Item 10 Interconnection and Grid Analysis Update, Board of Directors Meeting September 14 to 15, 2026 Primary source, official · ercot.com
No action is requested; for discussion only.ERCOT, Item 10 Interconnection and Grid Analysis Update, Board of Directors Meeting September 14 to 15, 2026 Primary source, official · ercot.com
This Planning Guide Revision Request (PGRR) clarifies requirements for submitting dynamic model data for Large Loads to ensure accurate representation of each Large Load in ERCOT's models.ERCOT Market Rules, PGRR144 Primary source, official · ercot.com
Status: Pending Date Posted: Feb 18, 2026 Sponsor: ERCOT Urgent: NoERCOT Market Rules, PGRR144 Primary source, official · ercot.com
08/26/2026 TAC Recommended for Approval Board for considerationERCOT Market Rules, PGRR144 Primary source, official · ercot.com
09/15/2026 BOARD Recommended for Approval PUCT for considerationERCOT Market Rules, PGRR144 Primary source, official · ercot.com
To recommend approval of PGRR144 as recommended by TAC in the 8/26/26 TAC ReportERCOT Market Rules, PGRR144 Primary source, official · ercot.com
Questions about this decision
Answered from the record itself. Every answer is assembled from stored fields, so an answer the record has no basis for is left out rather than guessed.
What is this decision?
The Electric Reliability Council of Texas put a list of emerging Large Load risks in front of its own Board of Directors at the September 14th and 15th, 2026 meeting. The presentation names three risks and a mitigation for each, and the three sit at three different stages. Ride through requirements are already in force. The model quality rule cleared the board on September 15th and now waits on the Public Utility Commission of Texas. The third rule has not been written yet. The grid operator states that Large Loads tied to data center AI training processes can swing their power draw fast enough to perturb synchronous generator shafts. It says that swing can cause equipment fatigue or damage. Its draft rule would hold a Large Load's power variation to 10 MW per 5 seconds. The same page states that the dynamic models the grid operator uses to study these loads have proven inaccurate, and that Texas A&M University is building better ones. The presentation itself was for information only and the board was asked to take no action on it.
Who decides it?
Electric Reliability Council of Texas decides. The record names the deciding body for every entry it carries.
Can the public take part?
The board meeting the presentation was made to is open to the public and it has now been held. The revision request the presentation names, PGRR144, keeps a public page on the grid operator's market rules site. That page lists every vote taken on it and what body takes it next, which is now the Public Utility Commission of Texas. The draft rule on power variation has no number and no filing date yet. Nothing on it can be commented on until it is submitted, so there is no window a reader can act in today. No dated public window is on the record. The deciding body is named and reachable.
Where in Texas does it apply?
It is a statewide body's action and it governs the ERCOT region rather than every county.
Has it been decided?
It is pending. No final decision is on the record.
What happens next?
No future date is on the record. The last dated step on it was passed on September 15th.
When did it start?
The earliest date on its record is August 26th, 2026.
What kind of decision is it?
It is filed under power and the grid.
What sources back it?
Two sources back it. Two of them are primary.
Is it on the ERCOT grid?
Yes. It sits inside the ERCOT interconnection.
When was it last checked?
Every fact on it was last verified against its source on September 27th, 2026.
Cite this
Texas AI Docket, ERCOT tells its board that AI training loads can fatigue grid generators and drafts a limit on how fast their power draw may swing. Tracked since August 26th, 2026. Last verified September 27th, 2026. https://texasaidocket.com/item/tx-2026-0159/. Reuse permitted under CC BY 4.0 with attribution. The same entry is in the docket JSON as item tx-2026-0159.
Beat
Filed under Power and the grid, with every other decision on that beat.