A retirement date is now a forecast, and storage models pay for it
Victoria Shinder2 min readTwo items from this week describe the same problem from opposite ends.
In the United States, the Department of Energy issued a fourth emergency order keeping Unit 1 of Colorado's Craig Station available. It was due to retire at the end of 2025; each order runs for about a quarter, and the latest takes it to 25 December 2026. The department says more than 17GW of coal capacity has been kept from retiring since 2025.
In Australia, Energy-Storage.News recalled remarks by Thomas Schmitz, general manager of energy markets at Aquila Clean Energy APAC, that battery revenues in the National Electricity Market had narrowed as arbitrage spreads compressed, and that developers who had built financial models around coal closures were absorbing the consequences of those closures being delayed. The report came with the news that ESR, a logistics and data-centre owner, is buying Aquila's Asia-Pacific platform.
These are different markets with different causes. Craig is kept available by federal order on reliability grounds; delays in the NEM have come through owners' and governments' own decisions. The common thread is not the reason for the delay but what it does to everyone else's planning.
Why the date matters so much
A retirement date is an input to almost every other investment on the system. A battery earns from the spread between low and high prices, and the spread depends on how much dispatchable capacity is left to set the high price. Replacement generation is sized to fill a gap that opens on a given day. Network upgrades and capacity procurements are scheduled around it. When the date slips, none of those projects is wrong in principle - but each one earns later, or less, than its model said.
The Craig case adds a second layer: the date does not slip once. A time-limited order that is renewed each quarter means the unit is always officially retiring and always available for a few more months. That is harder to plan around than a clear decision either way, because a planner cannot choose one assumption and hold it.
What follows
For investors, the lesson Schmitz draws - that a battery's revenue resembles a peaking plant's more than a stable infrastructure asset's - means retirement timing belongs in the risk analysis as a variable, not as a fixed line in the base case. For regulators and system operators, the lesson runs the other way: if they want replacement capacity to be financed, the most useful thing they can provide is a date that holds.