Complex Systems Don’t Fail Without Warning. There Is a Better Way to Watch.
John Carmean
The Texas power grid didn’t collapse on February 15, 2021. It started collapsing weeks earlier. Generation reserves were thinning. Demand forecasts were diverging from capacity models. Equipment performance metrics were drifting outside historical norms.
People were watching. Grid operators monitor their systems around the clock. Meteorologists were tracking the weather. Market analysts were reading the curves. Every one of those professionals was doing their job with the tools they had. The tools worked exactly as designed.
The problem is that the tools were designed to watch one domain at a time. Weather models watch weather. Grid models watch load. Market models watch price. None of them were built to synthesize across all of those domains simultaneously and ask a different question: is this system, as a whole, drifting toward a state it cannot recover from?
When the storm hit, 4.5 million people lost power. 246 people died. The economic damage exceeded $300 billion. What followed has become the largest utility litigation event in U.S. history, and five years later, it is still unresolved.
31,600 lawsuits were consolidated into a single multidistrict litigation in Harris County under Judge Sylvia Matthews. 418 attorneys across the state are actively working the cases. 120 trial lawyers represent approximately 30,000 individuals and small business owners pursuing claims for wrongful death, personal injury, and property damage. The defendants include the largest transmission and distribution utilities in Texas: CenterPoint Energy, Oncor Electric Delivery, and American Electric Power.
The legal landscape has narrowed but not disappeared. In 2023, the Texas Supreme Court granted ERCOT sovereign immunity, removing the grid operator from litigation. That same year, claims against natural gas companies were dismissed. In December 2023, the First Court of Appeals in Houston ruled that lawsuits against wholesale power generators like Luminant, NRG, and Sempra had no basis in law or fact. Generators are out. ERCOT is out. Gas companies are out.
The transmission and distribution utilities are not out.
In April 2024, the Texas Fourteenth Court of Appeals unanimously ruled that allegations of gross negligence and intentional misconduct against the TDUs can proceed to trial. The court held that these companies are not shielded from claims that they made deliberate operational decisions during the storm that worsened outcomes for their own customers. In August 2025, the Texas Supreme Court dismissed the bulk of remaining claims, but the underlying question has not been settled. As of February 2026, there have been zero jury trials. The defendants are planning further appeals that could add another two years before any case reaches a courtroom.
The legal standard being tested is gross negligence: did the utility know about the risk and fail to act? That question will define liability for every utility operating in every territory where extreme weather, aging infrastructure, and accelerating demand create the conditions for the next major event. And those conditions are not unique to ERCOT. They are structural. They exist in PJM, MISO, SPP, and every other major interconnection.
The instinct after a crisis is to build more redundancy. Add capacity. Harden equipment. All necessary. But none of it addresses the deeper gap: complex systems generate measurable warning signals across multiple domains long before catastrophic failure, and the existing monitoring tools were never designed to read them together.
The data is not the bottleneck. Federal agencies publish grid operations data, generation metrics, weather patterns, and demand forecasts in near real time. The bottleneck is synthesis. It is knowing which combination of signals, across which domains, indicates that a system is drifting toward failure rather than experiencing normal fluctuation.
That distinction is solvable. However, it requires a different kind of instrument. One that reads across domains rather than within them. One that uses the same methodology regardless of what system it is monitoring, so the results are not biased by what the designer expected to find.
The organizations that will be positioned for the next major event are not the ones with the most infrastructure. They are the ones with the most sophisticated way of watching. Documented foresight is not just an operational advantage. In the courtroom that 30,000 plaintiffs and 418 attorneys are fighting through right now, it could be the difference between negligence and due diligence.
The warning signals have always been there. The instrument to read them now exists.