Winter Storm Uri: When Physical Disruption Reached the Contract

Historical cases become too easy when the ending is allowed to rewrite the beginning. This replay uses two clocks.

The first stops on 13 February 2021. One day earlier, Texas had issued a statewide disaster declaration because prolonged freezing temperatures, heavy snow, and freezing rain posed an imminent threat across all 254 counties. That was enough to raise the readiness posture. It was not enough to know which generating units would fail, how the market would price scarcity, or what one project would owe.

The asset and contract records create a second distinction. Shannon Wind's 2026 court filing establishes that the project and its fixed-quantity hedge existed before the storm. Those terms were available to the relevant owner, counterparty, and potentially its financing parties. They were not necessarily public information an outside analyst could retrieve on the cutoff date.

So the decision-time question is conditional and practical:

If the asset, operating, and contract records were assembled before the outcome, what connected exposures could a decision-maker have tested?

The later FERC, NERC, and regional-entity investigation shows why a single-cause story is inadequate. ERCOT ordered 20,000 MW of rolling blackouts to prevent grid collapse. Freezing issues accounted for 44.2% and fuel issues for 31.4% of unplanned generating-unit outages, derates, and failures to start—75.6% combined. Of the fuel-related outages, 87% were tied to natural gas.

Those percentages describe the system record, not Shannon Wind. Their value is structural: equipment, fuel production and processing, electric dependencies, and grid operations failed together across the same event.

On 15 February, after the replay cutoff, the Public Utility Commission of Texas order directed ERCOT to make energy prices reflect the $9,000/MWh system-wide offer cap during firm load shed. That was an observed market and policy state, not something this replay inserts into the pre-event evidence set.

The ERCOT Independent Market Monitor's 2021 State of the Market report makes the price environment visible. February zonal averages were roughly $1,483–$1,488/MWh day-ahead and $1,516–$1,545/MWh real-time across Houston, North, South, and West. These are monthly zonal averages, not a Shannon Wind settlement-node price.

The annual comparison shows how much the event changed the year's market average. With Uri included, average real-time prices were $158–$162/MWh across the four zones. Excluding Uri, they were $35–$40/MWh.

These charts establish the market regime. They do not reconstruct the Shannon invoice, convert the difference into an avoided-loss estimate, or imply that every interval cleared at the same value.

The final FERC/NERC record attributes 44.2% of unplanned outages, derates, and failures to start to freezing and 31.4% to fuel. The remaining 24.4% falls across other causes. Within the fuel category, 87% was tied to natural gas.

This is a regional system mix. It explains why connected asset, fuel, and grid dependencies matter; it is not a project-level attribution for Shannon Wind.

The named project makes the financial turn concrete. In a declaration filed with the U.S. Bankruptcy Court for the Southern District of Texas on 26 January 2026, Shannon Wind's chief restructuring officer described:

• a 204.1 MW wind farm in Clay County, Texas, with 119 GE 1.7-103 turbines;
• 2015 hedge agreements with Citigroup Energy that generally required fixed quantities at $26.20/MWh;
• blade icing that impeded production during Uri, leaving the project producing little energy;
• a requirement to buy electricity at market prices to meet the delivery obligation; and
• a March 2021 power invoice for $39,486,641.34.

The important difference is the value basis. This was not only foregone revenue on electricity the turbines did not produce. A fixed-quantity obligation made the delivery shortfall and the replacement-market price matter together.

The filing is a sworn debtor account that grounds the project example. This draft does not treat every statement as an adjudicated fact, say the invoice was paid, or claim the hedge alone caused every later financial outcome.

The left track contains only what belongs in the pre-outcome decision context. The right track contains the later record used to test the method. That separation matters: a replay is useful only if it shows what a disciplined process could have asked without pretending the ending was already known.

A stronger infrastructure-risk process would not begin and end with “Will the turbines fail?” It would join the asset, component, operating, market, contract, and financing records before deciding which exposure deserves action.

Current InfraSure surfaces can assemble supported assets in Explore and Evaluate, compare visible risk and evidence gaps, and move a selected asset into Manage. The bounded replay method adds the deeper engineering-and-finance test around private operating, contract, policy, and financing records. Broader reusable response comparison and formal portfolio optimization remain in build.

The value is not that a framework makes the storm disappear. It is that the component, contract, dependency, and liquidity path becomes visible early enough to change what gets inspected, negotiated, financed, insured, or monitored.

Applied before the outcome—with the relevant private records—it could have:

• separated ordinary lost-revenue exposure from fixed-delivery replacement-power exposure;
• prioritized the exact hedge volume, settlement interval, force-majeure language, collateral, liquidity, and covenant path for review;
• tested a joint scenario in which project output falls while replacement-market prices rise;
• exposed dependency questions around cold-weather-critical components, natural-gas supply, and grid-wide correlation; and
• informed readiness, contract, insurance-gap, liquidity, and lender diligence conversations.

See how the portfolio-diligence Workflow moves from screen to deeper testing.

The language is conditional for a reason. The replay validates the usefulness of the questions. It does not validate an exact forecast or a prevented-loss claim.

Use the same consequence chain, then apply each actor's own records and terms.