At the beginning of October, NASA reported that it had assembled the center segments of both solid rocket boosters for Artemis III — and that another set is being prepared for future missions, its parts simply sitting in a warehouse while the rocket waits its turn. A piece of hardware we are used to thinking of as the pinnacle of engineering has a shelf life, and it ticks away the entire time it sits there.
Then I came across an oddity that became the real hook. Normally, shelf life is about when you throw something out. Rocket parts behave the other way around: for the most critical systems, actual service life turns out to be many times longer than the stated one; it gets extended for years and decades — and that extension, not wear and tear, costs the most. There is physically nothing to replace a component with, because production has shut down, while retiring it costs more than keeping it going.
So the question I set out to check sounds accountant-level dull: how much does one year of keeping an operational missile on alert cost? The answer turned out to be about bills for evidence: about who, and at what price, is willing to certify that an old thing is still good for another year.
Let’s begin with that very piece of hardware from the news. SLS is the rocket NASA is building to send people back to the Moon; it is lifted into the air by two side-mounted solid rocket boosters, and each is not a single piece but a stack of segments fastened together on site, in the assembly building. The joints between the segments have their own separate deadline: they are certified for one year from the start of stacking. The clock starts on the day the first segment is placed on the bottom one. Neither the pouring of the propellant nor the day the hardware leaves the factory gates count toward that period.
This constraint was not invented for the new launch vehicle — it was inherited from the shuttle. It was not really about the fuel: at the joint there are seals and thermal barriers, and their interfaces change over time under gravitational loads and temperature. Officially, it is the certification paperwork that expires, but beneath that paperwork sits precisely the part that failed on Challenger: a ring that may not hold back hot gas.
And this is where the interesting part begins. When Artemis I did not roll out on schedule, the limit did not “expire” — it was extended. The original twelve months were stretched to twenty-three: they gathered data accumulated during stacking itself, added analysis and testing. One component expired on December 9, 2022, another on the 14th, another environmental-exposure rating on the 15th. Had the rocket not flown by those dates, more calculations would have followed. No one intended to replace the seals — the plan was to prove that they were holding just fine as they were.
While SLS counts months, ground-based missiles count decades. Minuteman III is an American silo-based intercontinental missile: it stands around the clock in a concrete silo in the middle of the prairie, and launches from there. It entered service in the early seventies; its design life was ten years. It remains on alert to this day. A report from the U.S. auditor released in April states matter-of-factly: the system has been in use for more than fifty years, its service life has been extended many times, and the Air Force is now considering keeping it until 2050 — because its replacement is bogged down.
The extension has numbers attached. The program called the Propulsion Replacement Program aimed to remanufacture and replace all three solid-propellant stages before “age-out” set in — that point of resource exhaustion beyond which a stage can no longer be justified by anything. The 2000 estimate: around $331 million for development and $1.8 billion for procurement. The first replacement booster was delivered in April 2001; the last, the one hundred and fiftieth, was lowered into a silo in August 2009.
One detail matters here, because it explains the entire logic of such programs. The overhaul does not involve dismantling a stage and replacing the part that has gone bad. The stage is removed whole and a new one installed in its place, because the old one cannot be repaired — I will come back to that separately. For now, remember the order of magnitude: extending a missile’s life by twenty years cost around two billion, and that was only the stages — without silos, cables, and command centers.
No one knew the right timing for these replacements in advance — they worked it out as they went. In an interview in 2024, the program manager admitted it outright: they had never expected the system to live long enough that they would need to predict exactly when particular components would fail. The engineers had entered a mode military hardware was never originally designed for. They now had to predict what would fail, and when, in an old machine.
The work was deliberately not called a service-life extension. A senior general explained in July 2024 why the term did not fit: a formal extension program is a process meant for new permanent modifications, and for Minuteman it “doesn’t make sense.” Instead, there is “sustainment”: keeping it alive until the replacement arrives. The difference lies in the legal framework: a full-fledged extension comes with a fixed commitment and a budget line; sustainment comes with the promise that “we’ll deal with it if something comes up.”
The general bluntly named the main risk this regime operates under: “we can only plan for what we know” — and then, he added, there is always the chance of unknown unknowns. He also mentioned separately that some parts are no longer manufactured, so procuring them requires long lead times. A system that is fifty years old exists in a world where the workbenches built for it are gone.
There is another side to this story, where the decision was made not in favor of aging. Sentinel is the missile meant to replace Minuteman. It was supposed to go into the same 450 silos, after refurbishing them: that was thought to be faster and cheaper. Then it emerged that the opposite might be true. In May 2025, the Air Force reported that Sentinel would most likely require new silos, rather than the repair of old ones.
The general’s wording at a spring briefing sounded almost comic: part of the original requirement had been to reuse “the holes, the missile holes at the launch facilities,” and that seemed more economical — but, “surprisingly, as we look at it now, that may not be the answer.” It turned out the old silos were no less obsolete than the missiles themselves: new cable networks, command centers, and buildings were needed. The “command and launch” segment accounted for around 81% of the budget overrun and drove the program’s price to $141 billion.
This is worth pausing over, because the answer to the original question is hidden in that figure. The most expensive thing turned out to be the place where the missile sits. The motor and the life extension of the old stages are far more modest line items. Infrastructure built for the previous generation and now unfit for reuse — that is where all the bills converge. It, too, was once built for a specified service life, and no one planned to keep it around for seventy years.
Now for why a stage gets replaced whole rather than refreshed. Over its life cycle, a solid-propellant motor endures thermal stress as the warehouse goes from warm to cold and back again, plus vibration and shocks from transport and handling. The accumulation of these effects causes irreversible changes in chemical composition and mechanical properties, cracking. In the worst case, the result is unstable combustion and, at the limit, motor rupture.
The motor is designed around very narrow limits for operating parameters. There is no “we’ll tweak this bit a little”: if the propellant grain has cracked or lost adhesion to the casing, it cannot be fixed in place. That is why life-extension programs remove stages and install new ones — repairing the energetic material is impossible in principle.
How do they measure how much life a motor has left? There is a distinction worth keeping in mind. Destructive methods — live firings, chemical analysis, mechanical samples — are considered the standard because they provide a direct number. Nondestructive methods, from X-rays to ultrasound, are physically capable of doing the same, but are rarely used in the life cycle of munitions: there is no unified standard for them. They are used after destructive tests to “transfer” the result obtained from one dismantled stage to the entire batch. In other words, when it comes to the question “is this particular missile still good?” the industry has no exact answer, only a statistical estimate for the batch.
One more detail, my favorite of all: participants in discussions around those same SLS boosters noted that the propellant’s “good until” clock starts from the beginning of casting — meaning the propellant grain has its own timer, the seals have theirs, and neither is tied to either rocket assembly or the launch date.
Putting it all together, I ran into something that the engineers behind these programs seem to know about their work but rarely say aloud. A system’s service life is determined not by its most important component, but by the shortest-lived one that cannot be replaced. A table for tactical missiles puts these side by side: pyrotechnics — twenty years, propellant — twenty to thirty-five, electronics — twenty-five to thirty, and rubber and plastics — ten to twenty. Metal, meanwhile, lasts more than fifty, so the airframe itself, around which the entire missile is built, would not be a problem at all.
In the end, a missile’s expiration date is set by a piece of rubber. Then the regime I have been circling around throughout this entire piece kicks in: it gets extended not because everything is fine, but because the alternative costs more. Dismantling and replacing a stage — two billion and years of work. Building silos anew — 141 billion and a blown schedule. Living with an empty hole in the ground is not considered in any scenario.
So in this field, “expired” does not mean “the thing has broken.” It means the paperwork certifying that the thing is intact has run out — and now people must be paid for new paperwork. Apparently, no one can calculate the honest price of a year in storage, whether for a silo-based missile or a booster in a warehouse: the documents contain prices for replacements and new silos, but no figure for what a year of alert duty costs for something already in place. Not because they are hiding it, but because no one has ever set out to derive it. A dull detail, perhaps, but it is precisely why the decision every time comes down to “extend it and do the math later.”