On 4 November 2025 UPS flight 2976, an MD-11F freighter, began its takeoff roll on Runway 17R at Louisville, Kentucky. During rotation the left engine and the pylon holding it came away from the wing together, already on fire, and passed over the top of the fuselage. The aeroplane never got higher than about thirty feet. Nineteen seconds after the first bad parameter the recording stops. Fifteen people were killed.
There is a camera at the end of Runway 17R in Louisville, Kentucky. It belongs to the airport, and it points down the runway all day at nothing. On the fourth of November 2025, at about fourteen minutes past five in the evening, it recorded this. A cargo aeroplane rotating — and then the left engine leaving the wing.
Not shedding a part. Leaving. The engine and the pylon that held it come away together. There is already fire on the engine as it goes, and it passes over the top of the fuselage before it hits the ground.
And a second fire starts, up at the wing, where the pylon used to be attached. That one does not go out. It burns for the rest of the flight. The flight lasts about nineteen more seconds.
The aeroplane never gets higher than about thirty feet above the ground. Thirty feet. A three-storey building. The controller in the tower said it never came up to the height of the tower.
That recording is not ours to show you. The camera belongs to the airline, and reaching the public inside a federal report is not the same as being free to use. So everything you see of those twenty seconds in this film is a reconstruction — built from the aeroplane's own recorders and the investigators' own data, and from nothing else. The track, the height, the speed, the second the fire signal changed.
Where you are looking at something real, this film will tell you so. And here is the first of those, because the recording is not the only thing that survived. The part that broke was recovered, photographed on a bench, and taken apart in a federal laboratory — and those photographs are public. You are going to see the actual object.
It is smaller than a fist, and it was inside the wing. A bearing. A ball, in a ring of steel, in the joint where the pylon attaches to the wing. It had a groove machined into the inside of that ring.
And the crack in it started at the edge of that groove and ran all the way round. Boeing knew about that groove. In February of 2011 it told operators that this bearing had failed this way before — four times, on three aeroplanes — and that a redesigned bearing existed with the groove taken out. Fitting the redesigned one was recommended.
Fitting the old one was not prohibited. TWENTY-ONE THOUSAND AND FORTY-THREE CYCLES THE AEROPLANE
The M-D-eleven is a three-engine widebody. One engine under each wing, one at the base of the tail. It was designed by McDonnell Douglas, which later merged with Boeing, and it was developed from the DC-10. Remember that last part.
It comes back, and it comes back because the NTSB put it there. As an airliner the type had a short life. What happened to the survivors is what happens to most big three-engined aeroplanes: they became freighters. The F in M-D-eleven-F is freight.
Freight is not a gentle retirement. A passenger widebody on long routes flies one or two sectors a day. A freighter on an overnight network flies short legs, several a night, and every leg is a takeoff and a landing — a wing loaded and unloaded, a structure worked through its full range and back again. Cycles are what fatigue counts, and freight work is high-cycle work.
The accident aeroplane was registered November-two-five-nine-Uniform-Papa. At the time of the crash it had about ninety-two thousand nine hundred and ninety-two hours, and twenty-one thousand and forty-three cycles. Twenty-one thousand and forty-three. Hold on to that number.
It is going to matter, and not for the reason you would expect. It is not too high. Nothing about it was out of limits. That is the point.
AND THE BEARING INSIDE THE JOINT WHAT HOLDS AN ENGINE ON
An engine does not bolt to a wing. It hangs from a pylon, a fabricated strut that stands off the wing and carries the engine ahead of it and below it. The pylon has to do two contradictory things at once. It has to hold several tons of engine rigidly enough that the engine keeps pointing where the aeroplane is going.
And it has to let the wing bend — because wings bend, by feet, on every flight, and a mount stiff enough to resist that would tear the wing instead of protecting it. So on the M-D-eleven each pylon attaches to its wing in three places. A forward mount, which contains two spherical bearings, one above the other. A thrust link, immediately behind the lower one, which carries the engine's thrust.
And an aft mount. The aft mount is where this film lives. It is an assembly of two separate fittings bolted together, and it has two lugs — a forward lug and an aft lug. Between them sits one spherical bearing: a steel ball, and a ring around it called the race.
A clevis on the underside of the wing connects to that joint. Now, two facts about looking at it. The first: in a normal installation, the bearing itself is hidden. The NTSB's own photograph caption says so — the aft mount spherical bearing is obscured by the wing clevis.
The second, and this is the one that matters: the failure was still supposed to be visible. Because when this bearing fails, the race — normally one piece, normally sitting flush with the outside faces of the lugs — cracks all the way around, splits into a front half and a back half, and the two halves slide apart. Forward and aft. Past the faces of the lugs.
Which means you can see it. Not the crack. The consequence of the crack, sticking out where it should be flush. Boeing wrote that down.
The maintenance manual was changed to say: check that the bearing is not protruding forward or aft beyond the surfaces of the lugs. There was a check. It was a look. And it found nothing on this aeroplane, because the last time anybody performed it was four years before the accident.
TUESDAY, 4 NOVEMBER 2025 RUNWAY 17R
The fourth of November was a Tuesday, and it was a clear day: seventeen degrees, six knots of wind, ten miles of visibility. Flight 2976 was Louisville to Honolulu — eight and a half hours, which means the aeroplane was heavy, and most of that weight was fuel. Three people were aboard. The captain, pilot monitoring, with about eight thousand six hundred hours, nearly five thousand of them on this type.
The first officer, flying, with about nine thousand two hundred hours. And a relief officer with fifteen thousand two hundred and fifty hours, of which eight thousand seven hundred and seventy-five were on M-D-elevens. Between them, more than thirty-three thousand hours. Takeoff clearance about eleven minutes past five.
The taxi and the takeoff roll were uneventful. The flight data recorder shows the aeroplane and all three engines performing normally until about thirteen minutes and eleven seconds past five. Then the parameters from the number one engine stop being reliable — because the number one engine is no longer on the aeroplane. The fire warning for that engine goes from no fire to fire.
The throttles for engines two and three go forward: somebody in that cockpit pushed up the power on the engines they still had. Groundspeed about one hundred and eighty-four knots. The aeroplane climbs, and stops climbing. A witness in the tower said the takeoff speed looked normal for the type but the climb rate did not, because it never came up to the height of the tower, about two hundred feet.
Another witness said it stopped climbing, began to lose altitude, and rolled slightly to the left. The recorded data ends at thirteen minutes and thirty seconds past five. Nineteen seconds from the first bad parameter to the end of the recording. At thirty feet.
I am not going to tell you what the crew tried, beyond the throttles, because that is what the published record contains. What I will say is that the record shows them adding power on the engines that were left, which is the correct action, in a situation where the correct action was not going to be enough. THREE THOUSAND FEET OF IT THE GROUND IT CAME DOWN ON
Beyond the end of Runway 17R there is a blast fence. The aeroplane cleared it. Past that, at the southern edge of the airport, there is a stretch of working ground of the kind that exists at the edge of every freight airport in the world. A UPS Supply Chain Solutions warehouse.
A storage yard. A petroleum recycling facility. The left main landing gear hit the roof of the warehouse first. Then the aeroplane went into the storage yard and two more buildings, one of them the petroleum recycler.
It was mostly consumed by fire. The wreckage ran on for about three thousand feet to the south-southeast. Three crew and eleven people on the ground were killed. Two more on the ground were seriously injured, and twenty-one were hurt less badly.
One of the two who were seriously injured died fifty-one days later. Fifteen people. The rest of this film is about a groove machined into a steel ring, and a letter written in 2011, and an inspection that fell due in 2027. That is the right subject and it is where the answer is.
But the reason any of it is worth an evening of your attention is this stretch of ground, and it is worth saying so once, plainly, before the paperwork starts. NOTHING HERE WAS AGAINST THE RULES CRACKS ARE ALLOWED
To follow the rest you need one idea from engineering, and it is not the idea most people have. Most people assume aircraft structure is built not to crack, and that a crack means something has gone wrong. That is not how large aircraft are certified. Metal that is loaded and unloaded, over and over, eventually cracks.
Not because it was badly made — because that is what metal does. Load it, unload it, load it again a few thousand times, and a crack starts somewhere small and grows a little on each cycle. You cannot design that away. You can only design around it.
So transport structure is certified on a damage tolerance basis. The assumption written into the certificate is that cracks will happen. Safety does not come from preventing them. It comes from finding them — inspecting often enough that a crack is discovered while it is still small, long before it reaches a size that matters.
Which means the inspection is not paperwork wrapped around the safety of the structure. The inspection is the safety of the structure. Set the interval correctly and a cracking part is a maintenance item: found, replaced, forgotten. Set it too long and the same part, cracking at exactly the same rate, is an accident.
Nothing about the metal changed. Only the number of flights between the moments when somebody looks. Hold that against what you now know. A bearing hidden behind a wing clevis.
A failure that announces itself by sticking out past the lug faces, if somebody is there to see it. And an aeroplane doing high-cycle freight work. Everything that follows is about when somebody was scheduled to look. TWO LETTERS FROM BOEING MAY 2008, AND FEBRUARY 2011
There are two letters, and there are three years between them. Boeing issued the first in May of 2008, and revision A in February of 2011. A service letter is not an order. An airworthiness directive is the regulator's instrument and operators must obey it.
A service letter is the manufacturer telling operators what it thinks they should know. This one told them that the spherical bearing in the pylon aft mount — part number S-zero-zero-three-nine-nine-dash-one — had failed before. Four reported failures of the bearing race, on three different aeroplanes. And every one of them had started in the same place: at the design recess groove on the interior surface of the race.
A design recess groove is exactly what it sounds like. A groove, machined into the inside of the ring, on purpose, as part of the design. It is also, in metal that is loaded and unloaded, a place where stress concentrates and a crack can start. Four failures.
Three aeroplanes. One initiation site, every time. The letter explained how to catch it: the race is normally flush with the outer surfaces of the lugs, and when it fractures the two halves migrate outward, so the failure can be identified visually by the pieces protruding. The maintenance manual was updated to say so.
The letter said the inspection would be included in the general visual and detailed visual inspection of the pylon aft mount, normally at a repetitive sixty-month interval. And the letter said this, which the NTSB quotes: A review of the spherical bearing failure by Boeing determined it would not result in a safety of flight condition. S00399-523 THE PART THAT ALREADY EXISTED
There is one more thing in that letter, and it is the reason this film exists. Boeing had already redesigned the bearing. The letter describes a new configuration assembly, part number S-zero-zero-three-nine-nine-dash-five-two-three, which eliminated the design recess groove on the bearing race. The groove that every one of the four failures had started at.
Taken out. The letter recommended installing the new part in the event an installed bearing was found to be unserviceable. So: if your bearing is bad, put in the good one. And then the sentence the NTSB reproduces, which is the whole of it: However, the installation of spherical bearing assembly part number S-zero-zero-three-nine-nine-dash-one, to replace an unserviceable bearing, was not prohibited by the Service Letter.
The old part. The one with the groove in it. Still legal to fit. In 2011, and after.
Not an oversight and not a scandal — it follows directly from the finding in the same document. If this failure does not result in a safety of flight condition, then there is no basis on which to forbid the old part. The two sentences are consistent with each other. They are consistent with each other, and one of them was wrong.
THIS WAS NOT THE FIRST ONE THE LONGER RECORD
When the NTSB opened its hearing docket, it published a single page that is the whole of this chapter. It is titled: History of Reported Outer Race Fractures. It is a timeline with a legend, and the legend has three marks. Reported to Boeing.
Reported to the F A A's service difficulty reporting system. And not reported. Ten events. July 2002, a fractured race on a right pylon, Operator 1.
Boeing only. August 2007, right, FedEx — and this one goes to the F A A and not to Boeing. September 2007, right, Operator 2. Boeing only.
July 2008, a separated staking ring, right, UPS. Boeing only. October 2008, left, FedEx. Both.
February 2009, Operator 3 — and this one is fractured races on *both* wings of the same aeroplane. Boeing only. March 2017, right, FedEx. Both.
June 2020, left, FedEx. Boeing only. September 2020, right, FedEx. F A A only.
And December 2022, right, FedEx. That last one has the third mark on it. Not reported. Count the F A A marks and you get four.
Ten events over twenty years, four of which reached the regulator, and one that reached nobody at all. At the hearing the chair of the board went through them out loud — one, two, three, four, five, six, seven, eight, nine, ten — and put the question that the count raises. Now hold that against what the letters said. Four failures were in Boeing's hands by February 2011, and are described in Boeing's own letter.
And the NTSB has said, in writing, what it is now examining: how the letter's inspection was incorporated into Boeing's maintenance planning documents and manuals, how UPS incorporated it into its maintenance programme, and the correspondence history between Boeing and the FAA leading up to the issuance of the letter and afterwards. That is a list of the places where a known failure can stop being known. Not one of them involves anybody deciding to accept a risk. Every one of them is a question about transmission — whether the thing that was written down in one place arrived, intact, in the next.
TWO DATES IN THE LOGBOOK OCTOBER 2021, AND OCTOBER 2025
Now the aeroplane itself, and this is where the record is exact. The general visual inspection and the detailed visual inspection of the left pylon aft mount — the inspection the 2011 letter folded this check into, the one that would have looked for a bearing protruding past the lug faces — was required by UPS's maintenance programme at a seventy-two-month interval. Six years. It was last accomplished on the twenty-eighth of October 2021.
Which means that on the fourth of November 2025 it was four years old, and the next one was not due until late 2027. Boeing's letter had said this inspection normally happened at a sixty-month interval. UPS's approved programme ran it at seventy-two. I am not going to tell you that is improper, because an operator's maintenance programme is approved by the regulator and there may be a perfectly good reason in a document I have not read.
I am telling you the two numbers are different, and that the NTSB has said it is reviewing exactly this. Then the deeper inspections. A special detailed inspection of the left pylon aft mount lugs — the one that takes the joint apart — would have been due at twenty-nine thousand two hundred cycles. A special detailed inspection of the left wing clevis support would have been due at twenty-eight thousand.
The aeroplane had twenty-one thousand and forty-three. Neither task had been done, and neither was due. Seven thousand cycles short of one, and eight thousand short of the other. And then the detail that is hard to sit with.
There is a lubrication task for the pylon thrust links and the pylon spherical bearings. It runs every twenty-four months, or four thousand eight hundred hours. It was last accomplished on the eighteenth of October 2025. Seventeen days before the accident, somebody was working on those bearings.
THEN THE LABORATORY WHAT THE METAL SHOWED
The left pylon aft mount's forward and aft lugs were both found fractured. The left wing clevis, with the bearing and its hardware still attached, was found with a piece of the left wing at the accident site. The race had fractured all the way round, exposing the ball. Then the laboratory.
On the aft lug, on both the inboard and outboard fracture surfaces, fatigue. On the forward lug's inboard surface, fatigue along the bore. On the forward lug's outboard surface, no fatigue at all — pure overstress. That last one tells you the order of events.
One side had been cracking for a long time. The other side broke all at once, when it was the only thing left holding. And the bearing race itself. Fatigue cracking originating around the entire circumference, at the edge of the design recess groove — the same feature named in the 2011 letter — running through the thickness of the race toward the outside, and covering about seventy-five per cent of the fracture surface.
The remaining quarter was overstress. Seventy-five per cent. Three quarters of that ring had already failed before the day of the flight. The last quarter went on takeoff, because takeoff is when you ask a wing for everything.
And there was still grease in it. AN ENGINE HAD COME OFF BEFORE NINETEEN SEVENTY-NINE
The NTSB's preliminary report has a section near the end headed Similar Events. There is one entry in it. On the twenty-fifth of May 1979, American Airlines flight 191, a DC-10, was taking off from runway 32R at Chicago O'Hare. During rotation, the left engine and pylon assembly — and about three feet of the leading edge of the left wing — separated and fell to the runway.
It climbed to about three hundred and twenty-five feet. Then it rolled to the left, and kept rolling, until the wings were past vertical. Two hundred and seventy-one people on board were killed. Two more on the ground.
Now the discipline. Those are not the same accident. In 1979 the pylon had been damaged during maintenance — the engine and pylon were removed and reinstalled as a single unit on a forklift, which loaded the forward bulkhead in a way it was never designed for, and cracked it. A different structure, a different mechanism.
Not this bearing. What they share is a lineage, and the comparison is not mine. It is in the NTSB's own preliminary report, in a section the NTSB titled Similar Events. And the regulator said the same thing in the only language a regulator has.
On the eighth of November 2025 the FAA issued an emergency airworthiness directive grounding the M-D-eleven. On the fourteenth it issued another one, superseding the first — and this one covered the DC-10 as well, on the stated basis of its similar design to the M-D-eleven. Forty-six years, and the agency went back and looked at the ancestor. TWO DAYS OF TESTIMONY THE MISUNDERSTANDING
A hearing is not a trial and it does not produce a verdict. It exists to put the record on the table, under oath, in public, before the board decides anything. Two sentences from those two days belong here in the speakers' own voices, and nothing else does. The first is from the FAA, asked about the history of this bearing.
Dr Violette, Federal Aviation Administration NTSB investigative hearing into UPS flight 2976, day 1, 19 May 2026 There was a misunderstanding, initially, twenty years ago, about the severity of the event that might result from failure of this bearing. That's why it was not designated a principal structural element. That's why it was considered not a safety issue. A principal structural element is a designation, and the designation decides how a part is treated — how it is analysed for fatigue, and how often somebody has to go and look at it.
The severity was misunderstood, so the part did not get the designation, so it did not get the treatment. The second is from the operator's side, about the wording of the letter. Mr Springer, United Parcel Service NTSB investigative hearing into UPS flight 2976, day 1, 19 May 2026 So all that information - you know, you read the service letters, it's the same way, just like this library I referenced. It talks about, hey, it makes it sound almost benign.
We have a bearing fracture and it slides out, it migrates a little bit. It's not a big deal. The rest of the damage that was being caused was a big deal. And I think if we would have known that at UPS, I think we would have asked a lot of different questions over the years.
He was describing what he had learned since. That it was not only a bearing issue. It was the collateral damage that was happening to the lugs — deformation, gapping, damage to the lug bores. Almost benign.
Hold that against chapter ten. On the accident aeroplane it was not only the bearing that had been cracking. Both lugs were fractured, and the laboratory found fatigue in them as well. Neither of those is a confession and neither should be read as one.
They are the same event described from opposite ends of the same chain: a hazard that was correctly identified, correctly written down, correctly circulated — and filed under the wrong heading. After which everything downstream behaved exactly as it should have, given the heading. That is the ordinary way this happens. Not a decision to accept a known risk.
A decision about which category a risk belongs in, made once, early, by people who were not wrong on the evidence in front of them — and then never revisited, because nothing that arrived afterwards was ever gathered into a reason to revisit it. THE FLEET STOPPED FLYING WHAT CHANGED IN FOUR DAYS
UPS grounded its M-D-eleven fleet on the seventh of November, three days after the crash, at Boeing's recommendation. On the eighth, the FAA issued Emergency Airworthiness Directive twenty-twenty-five dash twenty-three dash fifty-one, prohibiting further flight of M-D-elevens until they were inspected and any corrective action approved by the FAA had been carried out. On the fourteenth, Emergency Directive twenty-twenty-five dash twenty-three dash fifty-three superseded it and added the DC-10. Ten days.
From a bearing that would not result in a safety of flight condition, to two aircraft types prohibited from flying. Nothing about the bearing changed in those ten days. What changed was that one of them was recovered from beside a runway in Louisville with three quarters of its fracture surface already worn smooth. WHAT THE GROOVE WAS FOR
The National Transportation Safety Board has not published a final report on this accident. There is no probable cause. Everything here that touches on why is the preliminary findings, the January investigative update, or sworn testimony at a hearing — and some of it may look different when the board finishes. Anyone telling you the cause is settled is ahead of the people whose job it is to settle it.
What is not going to change is what was in that letter. In February of 2011 Boeing told M-D-eleven operators that this bearing had cracked four times on three aeroplanes, always starting at the same machined groove; that a redesigned bearing existed with the groove removed; that the redesigned one should be fitted if the old one was found unserviceable; that fitting the old one was still permitted; and that a review had determined the failure would not result in a safety of flight condition. Fourteen years later, an aeroplane with twenty-one thousand and forty-three cycles on it — an aeroplane whose deep inspections were seven thousand cycles away, whose last visual inspection of that mount was four years old, and whose pylon bearings had been lubricated seventeen days earlier — rotated off Runway 17R with a crack already round three quarters of that groove. It got about thirty feet.
There is a recording of that, and we have not shown it to you, because it belongs to the airline and a federal report reproducing something does not make it free. What you have watched of those twenty seconds was rebuilt out of the aeroplane's own recorders. What we can show you is the ring. Investigators photographed it on a bench, in pieces, and marked on the fracture surface which part of it had been failing for years and which part broke on the day.
Three quarters, and one quarter. That photograph is public, and it is the closest thing this story has to an eyewitness. What no camera recorded is the fourteen years.
Wonder Minute