Resource Guide
ERCOT's new rule on large load ride-through: NOGRR 282
What hyperscalers and data center developers need to know
Resource Guide
What hyperscalers and data center developers need to know
NOGRR 282 is a new Electric Reliability Council of Texas (ERCOT) rule that has been approved by the Public Utility Commission of Texas (PUCT). It requires Large Computational Loads (LCLs) to ride through voltage and frequency disturbances instead of shutting down. This applies to AI data centers and cryptocurrency mining facilities that use 75 megawatts or more of power, where at least half of that load runs through power electronics. The ERCOT Board recommended approval, and the PUCT approved the rule. These rules are the first of their kind in North America, and we expect other grid operators to quickly follow, particularly since NERC's Level 3 Computational Load Alert was published urging others to act.
Several argued that ERCOT does not have the legal authority to place these requirements directly on retail customers and believed any requirements should instead run through utility delivery tariffs or through formal rulemaking by the Commission. Texas Industrial Energy Consumers' (TIEC) motion to intervene specifically on NOGRR 282 positions them to make further legal challenges to the rule. Given Governor Abbott's directive that regulators ensure data centers do not harm ordinary ratepayers, we expect the rule's technical requirements to remain in place, even if it requires a different implementation path.
New LCLs of 75 megawatts or more need to prove they can ride through disturbances before they can connect to the grid. This builds on ERCOT's interim rules, which already treat ride-through as a requirement for approval.
NOGRR 282 lays out specific steps if a facility fails to ride through a disturbance. Within 90 days, the facility must explain what went wrong. Within another 90 days, it must submit a plan to fix it. Within another 180 days, it must carry out that plan. If ERCOT determines the facility poses an immediate reliability risk, it can order the facility to disconnect and remain disconnected until it proves compliance.
Although more limited than stakeholders requested, there are carveouts for projects that have completed interconnection studies or were part of the interim Large Load Interconnection process.
ERCOT has broad discretion in this process, so having a proven compliance plan significantly derisks a project from future issues. If there are design changes at the site or operational issues, ERCOT has broad authority for restudy.
For LCLs planning to operate a data center on the ERCOT grid:
Do both our computing load and our cooling equipment meet the approved ride-through requirements on their own?
Does our project qualify for the exemptions to compliance? If we make design changes, will we still be eligible?
Can our current power protection equipment handle a deep voltage drop, the hardest part of the new curve? Are we able to prove that to ERCOT, given their broad discretion?
Now that NOGRR 282 has passed largely as drafted, what would it cost, and how long would it take, to upgrade our power protection equipment after the fact, instead of building it right the first time?
ON.energy did not wait for NOGRR 282 to be finalized before testing against it. In January 2026, ON.energy installed its AI UPS™, a fully inline medium-voltage power system, at the U.S. Department of Energy's National Laboratory of the Rockies (NLR). Independent testing there gathered full-scale hardware test results of AI UPS™ performance under a wide variety of voltage ride-through scenarios.
| Test Condition | Relevant ERCOT Requirement | AI UPS™ Result |
|---|---|---|
| Zero Voltage Ride Through (ZVRT) | Must stay connected for 150 ms | Meets and exceeds |
| Overvoltage (120%) | Must ride through for 1 second | Meets and exceeds |
| AI Workload Transients (±70% swing) | Ramp rate compliance required (20%/min proposed) | Meets and exceeds |
Unlike other options on the market, AI UPS™ is fully inline, meaning power runs through it at all times rather than switching over only when a disturbance occurs. Its architecture allows for a faster and more reliable response compared to other designs that react after a voltage sag. This design enables AI UPS™ to handle the hardest part of the new rule, a deep voltage drop, where other designs fall short. ON.energy's design is protected under U.S. Patent 12,614,920, with a priority date of September 2, 2022.
“NOGRR 282 is achievable.
We deployed at NLR specifically to run these tests under real-world conditions that match what AI data centers face at interconnection. The validated results are definitive: AI UPS™ rides through a complete zero-voltage event while keeping load voltage stable, and handles ±70% GPU transients without registering on the grid.”

AI UPS™ does more than meet the rule. The same system can also earn revenue through peak shaving, demand response, energy arbitrage, and potential participation in ERCOT's ancillary services markets.
Full test data, including charts, ride-through curves, and a cost comparison, is available in the AI UPS White Paper.
Contact ON.energy to talk through what this rule means for your facility.