The 30-Year Chip

Why Your Airplane Is Flying on Components Older Than Its Pilots

I once watched a program manager cry.

Not from sadness. From rage. Texas Instruments had just sent out a Last Time Buy notice on a twenty-five-cent microcontroller. An 8-bit MCU designed in 1994. The kind of chip your microwave has three of. Except this one was qualified for a flight control computer. Qualified as in tested under DO-254 rigour, traced back through every lot, flown for eighteen years without a single reported failure across a fleet of four hundred aircraft.

And it was about to be gone. Forever.

The panic wasn’t about finding a new chip. That’s easy. The panic was about what it costs to prove a new chip is safe enough to sit in a cockpit. That number, for that particular MCU in that particular application, was seven million dollars and thirty-two months of re-qualification work. Nobody had that money. Nobody had that time.

So we hoarded. Four thousand eight hundred chips. Vacuum-sealed. Nitrogen-flooded. Locked in a climate-controlled vault that, I am told, has better physical security than most banks in Quebec.

Welcome to obsolescence management. The slowest, strangest, most quietly terrifying corner of the aviation industry. The place where Silicon Valley’s favourite slogan — “move fast and break things” — meets the one industry on Earth that cannot, under any circumstance, break anything.

The Decay Nobody Outside Aviation Understands

Civilian consumer electronics refresh their component pool every eighteen months. Aviation refreshes every thirty years. Sometimes forty. A Boeing 737–800 rolling out of Renton today contains ICs whose fabrication process was deprecated in the Clinton administration. That is not a scandal. That is the plan.

Every component in a safety-critical system lives inside a bubble of qualification documents. MIL-STD-883 test reports. Radiation tolerance curves. Thermal cycling data. Mean Time Between Failures calculations assembled from millions of field hours. To swap the component is to throw out the bubble.

And the bubble costs more than the aircraft itself.

A new part means new DO-254 hardware assurance documentation. A new DO-254 package means a fresh Designated Engineering Representative sign-off. That sign-off means thousands of pages of traceability — from die-level test data through integrated system behaviour — every one of them reviewed and stamped. At roughly four hundred dollars an hour, times eight thousand hours of engineering, times the six to eighteen months you pay flight crews and maintenance staff not to fly — you are looking at fifteen to forty million dollars of swap cost.

Per chip.

For a twenty-five-cent MCU, nobody in their right mind makes that call. Nobody makes it at all unless the supply chain truly dies.

So the part stays. For decades. And when the manufacturer finally turns off the lights, an acronym you have never heard of goes into full crisis mode.

DMSMS: The Acronym That Keeps You Alive

DMSMS. Diminishing Manufacturing Sources and Material Shortages. Every major aerospace prime — Lockheed, Boeing, Airbus, Bombardier, Embraer — has a DMSMS team. Nobody outside the industry knows they exist. Inside the industry, they are the reason your Q400 takes off on schedule.

Their job is to stare into the future and predict which semiconductor the civilian market is about to abandon. Not six months in advance. Two to five years in advance. By the time a distributor posts an official End-of-Life notice, a DMSMS team should already have executed a Last Time Buy, seeded two alternate qualification programs, and flagged the affected line replaceable units for long-horizon cost modelling.

They subscribe to GIDEP — Government-Industry Data Exchange Program — a quietly essential database where contractors warn each other about EOL notices, counterfeit lot reports, radiation-hardened shortages. GIDEP is where the difference between a fifty-dollar part and a four-thousand-dollar flight delay gets decided eight months before anyone at the airline notices.

DMSMS is the least-celebrated job in aviation. No TED talks. No LinkedIn thought leaders. No YouTube explainer with an earnest narrator holding a diagram. Just grey-haired engineers cross-referencing distributor stock levels against fabrication roadmaps at four in the afternoon while the office coffee machine quietly gives up.

I have met these people. They are not impressed by much. They have seen too many startups promise to “disrupt supply chain visibility” and then vanish after their Series A.

The Strange Economics of Not Dying

When Freescale announced End-of-Life on the MC68332 — Motorola’s great-grandfather of automotive and avionics processors, shipped in something like a hundred and twenty different aircraft platforms — every defense contractor in North America pounced.

Distributor prices quietly quadrupled over six weeks. Gray-market brokers appeared out of nowhere, offering suspiciously abundant stock. Forums filled with small outfits in Shenzhen promising “OEM quality, traceable lot, reasonable pricing, discreet shipping.”

Some of that stock was real. Pulled from obsolete industrial equipment, re-reeled, re-labelled, genuinely functional.

Some of it was real but remarked — older die with a newer date code laser-etched onto the package, inflating the grade from industrial to military.

Some of it came out of Shenzhen loading docks with original markings sanded off with a Dremel and replacement markings applied by a printer that somebody bought on AliExpress.

This is where AS6081 and AS6171 enter the story. SAE International’s standards for counterfeit electronic parts detection and test methods. Every serious MRO and defense prime now has X-ray machines, decapsulation setups, Scanning Acoustic Microscopy, electrical test benches specifically designed to catch fakes.

I have watched a senior reliability engineer in Saint-Laurent spend fourteen hours verifying a single lot of a hundred chips, one sample at a time. X-ray first. Compare the die bonding pattern to the reference library. Decap a random sample. Photograph the die under a microscope. Compare the die markings. Run electrical at temperature. Sign the lot acceptance or fail the whole batch and eat the cost.

The alternative was installing untrusted silicon in a flight management computer and trusting a broker three countries away.

He chose the fourteen hours.

The Report Nobody Talks About

The counterfeit problem is not a hypothetical.

In 2012 the U.S. Senate Armed Services Committee published a report identifying more than one thousand eight hundred cases of suspect counterfeit electronic parts in U.S. military systems. Including parts in the Boeing P-8A Poseidon maritime patrol aircraft. Including parts in the SH-60B helicopter forward-looking infrared system. Including memory chips in the C-130J Super Hercules display units.

Some of those parts were traced back through a distribution chain that started in U.S. government surplus auctions, travelled to a broker in Florida, crossed to a facility in China that rewashed the silicon in the river — literally, in the river — to strip the original markings, dried them, re-marked them, shipped them back to a broker in California, and ended up in cockpits.

Read that sentence again. The parts keeping your aircraft alive had, at one point in their travels, been dunked in a Chinese river.

Nobody writes Medium posts about this. Nobody runs a conference keynote on it. The people responsible for catching it are a grey-haired metallurgist in Bakersfield, a reliability engineer in Saint-Laurent, a counterfeit detection specialist at the Defense Logistics Agency in Columbus, Ohio. All of them tired. All of them underfunded. All of them correct.

Why Aviation Cannot Just Modernise

At this point every programmer reading this wants to ask the obvious question. Why don’t you just redesign the board? Move to a modern, supported, traceable part? Be smart. Be modern. Move fast.

Try it. I dare you.

You will burn eight to eighteen months on re-qualification alone. You will write a new DO-254 assurance package. You will run fresh DO-160 environmental testing — Section 4 temperature cycling, Section 8 vibration, Section 22 lightning strike, all thirty sections — for a unit that is functionally identical to the one that passed ten years ago. You will produce a Technical Standard Order supplement. You will update the Aircraft Flight Manual. You will brief the airlines, who will scream about retrofit windows. You will coordinate with a DER who has a three-month backlog. You will pay a certification audit that costs roughly the price of a small house in Laval.

All of this for a chip that does a job the old chip was already doing perfectly for seventeen years across half a billion flight hours.

You start to understand why engineers hoard.

Where AI Actually Starts to Help

This, finally, is where artificial intelligence stops being a pitch deck and starts being useful in aviation. Not in the cockpit. Not as some Silicon Valley fever dream of autonomous taxiing and AI co-pilots. Somewhere much more boring. Somewhere much more real.

Machine learning models trained on twenty years of EOL announcements, distributor stock movements, commodity pricing signals, semiconductor industry roadmaps, and lead-time drift data can predict obsolescence risk on a specific part number with genuinely better accuracy than human analysts. Not because they are smarter. Because they are tireless.

I have seen a model correctly flag a Microchip SRAM variant as at-risk eight months before the official EOL notice landed. The signal wasn’t dramatic — a distributor quietly stopped accepting bulk orders over five thousand units, lead times crept from twelve weeks to twenty-six, one small vendor posted unusual “excess stock” on a B2B marketplace. No human DMSMS analyst had time to watch all three signals at once across the six thousand part numbers in a typical aircraft bill of materials. The model did. The model didn’t sleep.

That is the shape of AI’s honest contribution to aviation. Narrow. Tireless. Pattern-matching on boring industrial signals that no one else is paid to watch. Saving a certification package from dying eight months earlier than necessary. Letting a DMSMS team execute a Last Time Buy before the gray market sets the prices.

It is not sexy. It will not be in any conference keynote. Nobody will write a Forbes article titled “How AI is revolutionising aircraft certification.”

It will just save seven million dollars the next time a twenty-five-cent chip tries to disappear.

The Closing Joke Nobody Laughs At

In Montreal there is an engineer — I will not name him, he will laugh at me if he reads this — who has a framed photograph above his desk of the last reel of MC68332 microcontrollers his division ever purchased. Two thousand four hundred parts. Vacuum-sealed. Nitrogen-flooded. Stored in a climate-controlled vault in Saint-Laurent with temperature sensors, humidity sensors, and an access log that is reviewed every quarter.

He calls them “the retirement chips.” Because when the last one goes into a production unit, a certification program ends, and he gets to retire.

He plans to be there. Personally. Holding the last chip. Like a funeral.

I have thought about that a lot. About the quiet dignity of an engineer who has spent thirty years babysitting a reel of semiconductors so that two-hundred-seat aircraft can keep flying without anybody having to think about it. About how unsung that work is. About how nobody on a packed flight from Montreal to Toronto has any idea that their safe arrival depends, somewhere in the chain, on a grey-haired man checking a humidity log.

Would you fly on a plane whose most critical processor is a thirty-year-old chip stored in a nitrogen vault, tracked by a Montreal engineer’s spreadsheet, and guarded by a grey-market detection lab in California?

Yes. Because the alternative is worse.

The alternative is a shiny new chip that nobody qualified. And nobody — no engineer, no DER, no airline safety officer — wants to be in the air when that particular decision gets made lightly.

Written by Andrii Klymenko — Researcher & Writer. Exploring the human side of science. One micro-shift at a time.

☕ Want to support my work or explore more? → andriiklymenko.carrd.co Or tip $1 instantly via PayPal: paypal.me/AndriiKlymenko400/1