{"slug": "on-energy-tests-medium-voltage-ai-ups-system-designed-to-prevent-grid-data", "title": "ON.energy tests medium-voltage AI UPS system designed to prevent grid-destabilizing data center crashes", "summary": "Miami-based ON.energy has developed and independently validated a medium-voltage AI UPS system at the US Department of Energy's National Laboratory of the Rockies, where it met ERCOT large-load ride-through standards after a July 22, 2026 transmission fault in Ashburn, Virginia removed more than 3 gigawatts of load from the PJM grid in seconds. The system achieved zero-voltage ride-through for over one hour, sustained 120% overvoltage for 12 seconds, and limits ramp rates to 20% per minute or less during load swings of up to 70%, exceeding ERCOT's NOGRR 282 requirements. ON.energy reports over 2.5 GW in active deployment and roughly 3 GW more selected, with Crusoe planning to implement 5 GW across several campuses starting in 2026.", "body_md": "# ON.energy tests medium-voltage AI UPS system designed to prevent grid-destabilizing data center crashes\n\nA single transmission fault in Virginia knocked 3 GW off the grid in seconds, exposing a blind spot in how AI data centers interact with power infrastructure\n\nWhen a transmission line fault hit Ashburn, Virginia on July 22, 2026, more than 3 gigawatts of load vanished from the grid in seconds. That’s roughly 3% of the entire PJM grid’s demand, gone in a blink because hundreds of data centers all responded to the disturbance the same way at the same time.\n\nMiami-based ON.energy thinks it has a fix. The company has developed and tested a medium-voltage AI UPS system that keeps data center workloads running through grid disturbances instead of letting them crash offline and send shockwaves back into the power network. The system was independently validated at the US Department of Energy’s National Laboratory of the Rockies, where it met ERCOT standards for large-load ride-through.\n\n## Why data centers are the grid’s new nightmare\n\nTwo years before the July 2026 event, a single failed surge arrester knocked roughly 60 Virginia facilities and 1,500 megawatts offline simultaneously. Traditional low-voltage UPS systems, the kind already installed in most data centers, increasingly fail to meet emerging large-load ride-through standards. They were designed for an era when data centers were smaller and their collective behavior during faults didn’t threaten regional grid stability.\n\n## What ON.energy’s system actually does\n\nON.energy’s AI UPS functions as both a campus-wide uninterruptible power supply and a grid-interactive battery energy storage system. During testing at the National Laboratory of the Rockies, the system achieved zero-voltage ride-through for over one hour and sustained 120% overvoltage for 12 seconds.\n\nThe system also limits ramp rates to 20% per minute or less during load swings of up to 70% in either direction, exceeding what ERCOT requires under its NOGRR 282 standards.\n\nON.energy holds US Patent 12,614,920 for its UPS architecture, with a priority date stretching back to 2022. The company has also secured validation from UL and independent system operators.\n\nOn the deployment side, the company reports over 2.5 GW in active deployment with an additional 3 GW or so selected for its system. Its most prominent partnership is with Crusoe, which plans to implement 5 GW of the AI UPS across several campuses starting in 2026.\n\n## The grid compliance bottleneck\n\nERCOT’s NOGRR 282 standards represent one of the first formal attempts to address synchronized load disconnection risk, requiring large loads to ride through disturbances rather than tripping offline and compounding the problem.\n\nON.energy has secured multiple supply-chain agreements for essential components, a move that suggests it’s planning for multi-gigawatt scale rather than one-off installations.\n\n**Disclosure:** This article was edited by Editorial Team. For more information on how we create and review content, see our\n\n[Editorial Policy](https://cryptobriefing.com/editorial-policy/).", "url": "https://wpnews.pro/news/on-energy-tests-medium-voltage-ai-ups-system-designed-to-prevent-grid-data", "canonical_source": "https://cryptobriefing.com/on-energy-ai-ups-grid-compliance/", "published_at": "2026-09-10 11:26:55+00:00", "updated_at": "2026-09-10 12:00:06.954656+00:00", "lang": "en", "topics": ["ai-infrastructure", "ai-chips"], "entities": ["ON.energy", "National Laboratory of the Rockies", "ERCOT", "PJM", "Crusoe", "US Department of Energy", "UL", "Ashburn, Virginia"], "alternates": {"html": "https://wpnews.pro/news/on-energy-tests-medium-voltage-ai-ups-system-designed-to-prevent-grid-data", "markdown": "https://wpnews.pro/news/on-energy-tests-medium-voltage-ai-ups-system-designed-to-prevent-grid-data.md", "text": "https://wpnews.pro/news/on-energy-tests-medium-voltage-ai-ups-system-designed-to-prevent-grid-data.txt", "jsonld": "https://wpnews.pro/news/on-energy-tests-medium-voltage-ai-ups-system-designed-to-prevent-grid-data.jsonld"}}