The Durability of the B-52 Will Outlive Us All
The Short Life of the F-22
What Comes Apart, Lasts
Inventor's Mind — feature essay.
For three budget cycles the Air Force tried to throw away some of its stealth fighters.
Not old ones, by any normal measure. The aircraft in question — the earliest F-22s, the Block 20 jets — have been flying since the early 2000s. They are barely two decades old, and they are still, by wide agreement, the stealthiest air-to-air fighters ever built. And the Air Force's repeated, considered judgment was that it would rather retire them than upgrade them, because bringing them up to current combat standard was priced at something like one and a half to two billion dollars over eight years, and the service decided that money was better spent almost anywhere else. Congress blocked the retirement, twice. The reason the upgrade costs so much is the reason this essay exists: the older jets, in the words of the trade press, lack modern architecture, making upgrades complex and costly.
Hold that sentence against the other airplane in this series. In the same era, the Air Force is spending billions to put new engines on a bomber designed in the early 1950s, so it can fly toward its hundredth birthday. We are scrapping — or nearly scrapping — twenty-year-old stealth fighters because we can't affordably get into them to change anything, while we cheerfully rebuild a seventy-year-old bomber for the fourth or fifth time.
The instinct is that the Raptor must be worn out. It is the opposite of worn out. The problem isn't the metal. The problem is that the B-52 comes apart and the F-22 does not. And what comes apart, lasts.
Durability is a property of seams, not strength
We think of durability as toughness — thick metal, strong structure, the ability to take a beating. That is almost entirely beside the point for anything that has to last decades, because nothing lasts decades by being beaten on. Things last decades by being changed — re-equipped, re-powered, re-missioned as the world moves — and the thing that determines whether you can change them is not their strength. It's their seams.
Picture any long-lived thing as a stack of layers that change at different speeds. An architect named Stewart Brand described buildings this way, and the idea travels: the site lasts forever, the structure lasts a century, the skin a few decades, the wiring and plumbing fifteen years or so, the floor plan five, the furniture a season. A building endures not when its structure is strong but when its fast layers can change without disturbing its slow ones — when you can rewire it without touching the frame, move the walls without moving the foundation. The layers shear past one another. The seams are where the life is.
The B-52 is all seams. Its engines, its radar, its avionics, its weapons, its wiring — the fast layers — can be pulled and replaced without touching the airframe, the slow layer underneath. Seventy years of upgrades, and the structure doesn't care what's bolted to it this decade, because nothing critical is welded to anything it shouldn't be. The bomber's secret was never that it was tough. It was that it was loose — loose enough to keep accepting a future its designers couldn't imagine, one bolted-on layer at a time.
The Raptor is shear-locked
The F-22 is the opposite, and the opposite is the source of both its brilliance and its short life.
On the Raptor, the layers do not shear. Stealth coatings, thermal management, electrical power, avionics, and airframe are co-optimized into a single integrated whole, each tuned against the others to wring out performance no looser design could reach. Which means you cannot change the fast layer — the avionics, the five-year layer — without disturbing the slow one. There is no spare power to feed a new sensor, no cooling headroom to carry its heat, no internal volume to put it in, because every watt and every cubic inch was spent buying performance on the day the jet was delivered. The Block 20 jets aren't expensive to upgrade because they're old. They're expensive to upgrade because they're full — there's no room to put the future, and no seam to put it through. The aircraft is welded shut around a moment of brilliance.
That is the whole story of why a twenty-year-old stealth fighter gets considered for the scrapyard while a seventy-year-old bomber gets new engines. Not age. Coupling. And it points at the variable that actually governs how long a thing lives, the one we almost never name: the question is not whether a design is strong, but whether its fast layers can move without dragging its slow ones along. Coupling, not margin, is the real variable in longevity. A thing dies young not because it wore out but because you can't get inside it to change anything.
But coupling is not a sin
Here is where the easy version of this essay would go wrong, and go wrong badly, so I want to stop it before it starts.
The easy version says: see, integral design is the mistake, the F-22's engineers over-coupled, modularity is virtue, build everything with clean seams. That is false, and any engineer who has worked at the frontier knows it's false, because some capabilities exist only when the layers are coupled. Stealth is not a bolt-on. You cannot modularize it any more than you can modularize the aerodynamics of a wing — it is an emergent property of the entire shape, the materials, the thermal and electromagnetic management all acting as one inseparable system. The moment you put a clean, serviceable seam into that system, you have put a discontinuity into the thing that had to be continuous, and the stealth degrades.
Modularity has a price, always. Every clean interface, every serviceable seam, every layer you keep loose enough to change independently costs you weight, volume, and peak performance. Most of the time that price is well worth paying. But at the bleeding edge of the possible — where you are trying to do something no one has done, on technology that is still immature — that price exceeds the budget. There is no performance to spare for the luxury of seams. So you couple everything, and you reach a capability you could not have reached any other way. The F-22's shear-lock is not a defect. It is the price of an emergent capability that is non-modular by physics. Demand that the Raptor shear like a Stratofortress and you have forbidden it from being a Raptor at all.
The law: architecture has to match the clock
So coupling is neither virtue nor vice. It's a bet, and like every bet in this series it's a bet about time.
Integral, coupled architecture is the right answer when you are buying peak performance at the frontier, on immature and fast-moving technology, for a thing you do not intend to keep very long. Modular, loose-fit architecture is the right answer when the technology has matured, when adaptability and cost matter more than peak performance, and when you intend to keep the thing for decades. The B-52's mission is mature — delivering mass from altitude, a problem whose physics have not moved in seventy years — so loose-fit is correct, and the seams that make it modifiable are exactly right. The F-22's mission was frontier — stealth air dominance at the absolute edge of the achievable — so integral was correct, and the coupling that makes it unmodifiable was exactly right for what it was buying.
Neither airplane was badly architected. They were architected for different points on the same curve — one for a mature mission and a long life, one for a frontier mission and, properly, a short one. The architectures are not in competition. They are answers to different questions about time.
The sin is the mismatch
Which means the failure, when it comes, is never the coupling. It is keeping a coupled, frontier-optimized thing past the point its architecture was built for — freezing a brilliant point-solution and then demanding it serve like a loose-fit keeper.
That is what happened to the Raptor, and notice where it happened: not in the design room, but in the budget office. The architecture did its job. The jet hit a capability nothing else has matched. The failure was a lifespan decision layered on top of a correct architecture decision — the buy got truncated to 187, the production line closed behind it, and a frontier-integral object that should have been iterated and replaced was instead frozen and kept, and kept, and kept. Then, twenty years on, we act surprised that we can't affordably upgrade the thing we deliberately built to be un-upgradeable in exchange for performance we deliberately chose to maximize. The architecture was right. The lifespan was wrong. And those two decisions were made by different people, in different rooms, who never had to reconcile their answers — which is the same two-office disease that strands every program's true cost, wearing an architecture costume this time.
Evolvability doesn't vanish — it moves
There is a way to build coupled and frontier and still be wise, and it resolves the whole apparent paradox.
A loose-fit thing like the B-52 evolves by changing the unit — you keep the airframe and swap its contents for seventy years. The buffer lives in the unit, and the unit shears. But a coupled frontier thing can evolve a completely different way: by changing the line instead of the unit. You build the thing cheap and integral and disposable, you fly it for eighteen months, and then you throw it away and build the next, better one — the buffer lives in the production line, and the unit is meant to die. The attritable drone, the iterate-and-replace platform, the swarm: these are as coupled and edge-optimized as any Raptor, and they are wildly evolvable, because the evolution happens at the line, not in the airframe.
So edge-coupling does not kill evolvability. It relocates it — from the unit to the line. And the F-22's real tragedy comes into focus: it was coupled like a disposable but kept like a keeper. Frozen at the line and held for decades — the worst of both architectures, with no shear in the unit and no iteration at the line. It got the short-life architecture and the long-life expectation, and the gap between them is exactly the bill we're now arguing about in Congress.
The decision, before you draw anything
All of which reduces to two questions you have to answer before you choose an architecture, not after.
Is the mission mature or frontier? And do you intend to keep this thing or replace it? Four answers, and only one is a trap. Mature and keeper: build it loose, build it to shear, build it like a B-52, a bridge, a building. Frontier and disposable: couple it to the edge, iterate at the line, and plan its death from the start — the drone, the munition, the thing you mean to replace. Mature and disposable: fine, make it cheap and simple. And frontier and keeper — the integral, edge-optimized thing you nonetheless intend to hold for decades — is the one incoherent cell, the one where the expensive failures live, because you have committed to keep for thirty years a thing whose entire justification was peak performance against a need you've already admitted is moving. That cell has a name now. It's called the F-22.
Lessons learned
Durability is a property of seams, not strength. Things last by being changed, not by resisting damage. Build the seams — the places the fast layers can move without disturbing the slow ones — not just the structure.
Coupling, not margin, is the real variable in longevity. A thing dies young not because it wore out but because you can't get inside it to change anything. Ask of any design: can the five-year layer move without touching the thirty-year layer?
Coupling is not a sin — it's the price of frontier capability. Some things, like stealth or a wing's aerodynamics, are emergent and cannot be modularized. Modularity taxes peak performance, and at the frontier the tax exceeds the budget. Sometimes you have to weld it shut to reach the capability at all.
Match the architecture to the mission's maturity and the thing's intended life. Integral for frontier-and-short. Loose-fit for mature-and-long. The only incoherent combination is frontier-and-keep-for-decades.
Evolvability relocates; it doesn't disappear. Loose-fit things evolve the unit and keep it. Coupled things evolve the line and discard the unit. Decide where your evolution lives — in the airframe or in the factory — and architect for that, deliberately.
The architecture can be right and the lifespan still wrong. The two decisions live in different rooms. The Raptor was coupled correctly and kept incorrectly, and the second decision was made by people who never had to sit with the first. Force the architecture decision and the lifespan decision into the same room, under one owner.
Close
We are retiring the Raptor young — or fighting over the bill to keep it — not because its designers failed but because they succeeded at precisely the wrong problem to hold onto for thirty years. They built a flawless point-solution, welded shut around a moment of frontier brilliance, exactly as that mission required. And then someone in a different room decided to keep it for decades, and stranded it there, full to the rivets, with no room to grow and no seam to grow through.
The B-52 will outlive it because the B-52 was built to come apart. What comes apart can be rebuilt, and what can be rebuilt does not die — it just becomes, quietly, a different airplane wearing the same shape, decade after decade. The Stratofortress was never the tougher machine. It was the looser one. The Raptor is held together by the brilliance of its coupling; the bomber is held together by the patience of its seams — and across seventy years, it turns out, the seams are where the life is.
Sources: F-22 Block 20 retirement and upgrade history (≈32 aircraft; ≈$1.8B/8-year upgrade estimate; repeated congressional blocks; 2025 reversal toward combat-coding as an F-47 bridge) per Air & Space Forces Magazine, 19FortyFive, National Security Journal, and Defense News reporting, 2023–2025. F-22 program figures (187 built; truncated buy) per CRS and GAO. B-52 re-engine per Defense One, May 2026. Shearing-layers framework after Stewart Brand, "How Buildings Learn."