UPS Airlines Flight 2976

A cargo departure from Louisville

UPS Flight 2976 was scheduled to carry cargo from Louisville Muhammad Ali International Airport to Daniel K. Inouye International Airport in Honolulu on 4 November 2025. The aircraft was N259UP, an MD-11 F with three GE CF-6 engines. Its crew consisted of a captain, first officer and relief officer. There were no passengers.

The first officer was flying and the captain was monitoring. All three held airline transport pilot certificates and MD-11 type ratings. The preliminary report recorded about 8,613 total flight hours for the captain, 9,200 for the first officer and 15,250 for the relief officer. Their MD-11 experience varied, but this was a crew with substantial flying experience across its three members.

The aircraft had accumulated approximately 92,992 flight hours and 21,043 cycles. These figures describe its service history; they do not, by themselves, establish that it was unsafe or explain the accident. The NTSB's investigation focused on the condition and history of specific structural components and on how the aircraft performed after their failure.

Takeoff and separation

The tower cleared Flight 2976 to take off from runway 17R at about 17:11 eastern standard time. The crew acknowledged. The taxi and early takeoff roll appeared uneventful. Flight data began recording the accident flight at about 17:07 and placed the start of the takeoff roll at approximately 17:12.

Airport surveillance video then captured a major failure shortly after rotation, when the aircraft's nose rose for takeoff. The left engine and its pylon separated from the wing. Fire ignited on the detached engine as it passed above the fuselage and fell to the ground. A separate fire developed near the left pylon's wing attachment and remained visible as the airplane continued.

An engine pylon is the structure connecting a wing-mounted engine to the wing. Losing the engine together with that structure is different from an engine stopping while remaining securely attached. The observed separation left damage and fire at the attachment area. The precise consequences for thrust, airflow, control and the airplane's ability to climb are questions for the detailed performance analysis.

Too low to clear the buildings

The airplane initially became airborne but did not climb normally. Its flight recorder's radio-altitude data showed no climb above approximately 30 feet over the surface beneath it. The preliminary report also recorded a differing last ADS-B altitude indication. Those figures came from different systems and reference conditions, so they should not be silently treated as interchangeable measurements.

The MD-11 cleared a blast fence beyond runway 17R. Its left main landing gear then struck the roof of a UPS Supply Chain Solutions warehouse at the airport's southern edge. It continued into a storage yard and two more buildings, including a petroleum recycling facility. Most of the airplane was consumed by fire.

The wreckage spread approximately 3,000 feet south-southeast from the warehouse. The consequences extended into places where people were working or visiting on the ground. A tower witness described an abnormal climb, and another saw the airplane stop climbing, lose altitude and roll slightly left. These are observations that the investigators compare with recorded data and physical evidence.

The people affected

All three crew members died. Eleven people on the ground were initially recorded as fatally injured, and 23 other ground occupants were injured. The preliminary tables divided the latter group into two serious and 21 minor injuries. A victim who had been seriously injured died 51 days after the accident.

That later death is acknowledged directly in a footnote to the NTSB's investigative update. Counting it produces 15 deaths altogether: the three crew and twelve people on the ground. The update's main casualty text still gives 14 fatalities. This account states both the overall toll and the difference in the investigation's published classification, rather than hiding the additional death or changing the official table.

The airplane carried cargo rather than passengers, but that did not limit the human loss to its crew. The crash path passed through occupied ground facilities and produced a major fire. The total should therefore always distinguish deaths aboard the aircraft from those outside it. No one aboard survived.

Recovering the attachment components

The NTSB recovered the flight data and cockpit voice recorders and read them at its Washington laboratory. The voice recorder held about two hours and four minutes including the accident flight. The data recorder held around 63 hours across 24 flights and roughly 450 recorded parameters. Specialist groups examined the structure, engines, maintenance, operations, materials and performance.

Investigators found the left engine, pylon, mount components and engine fragments on or beside runway 17R. Portions of the left aft mount had fractured and separated. Other pieces, including the wing clevis, spherical bearing and attachment hardware, were recovered with part of the left wing at the crash site. The separated engine and pylon were therefore physical evidence as well as objects visible in the surveillance footage.

The left pylon aft mount consists of two fittings with lugs around a spherical bearing. The wing clevis connects through this assembly. In simpler terms, these parts form a joint that transfers loads between the engine-supporting structure and the wing. Examining them helps distinguish pre-existing cracking from damage caused when the assembly tore apart or struck the ground.

Fatigue cracking and the bearing race

The preliminary laboratory examination found fatigue cracking in parts of the aft-mount lugs as well as areas of overstress failure. Some fracture surfaces showed both kinds of damage; one showed overstress without fatigue. This distinction matters because fatigue can grow under repeated loading before a final failure, while overstress describes material breaking under a load beyond its capacity.

The subsequent investigative update described the spherical bearing race, the outer component surrounding the bearing's ball. It had split into forward and aft portions. Investigators found fatigue cracking around the circumference at the edge of an internal design recess groove. They described fatigue across approximately 75 percent of the fracture surface, with the remaining portion consistent with overstress.

These findings are strong evidence about the components' condition, but they are not yet an adopted probable cause. The NTSB must determine how the component failures progressed, how they relate to the mount and pylon separation, and what inspection, maintenance or design factors contributed. A description of fatigue does not establish when a detectable crack first existed or who should have detected it.

The earlier service letter and inspection history

The bearing design appeared consistent with a part number discussed in a Boeing service letter dated 7 February 2011. That letter described four previous bearing-race failures on three airplanes. The failures had begun at the same internal recess groove. In those cases, split race pieces moved beyond the outer surfaces of the mount lugs, making the problem visually identifiable.

The update reports that Boeing's earlier assessment said those bearing failures would not result in a safety-of-flight condition. The letter discussed inspections for race migration and a revised bearing design without the recess groove. It recommended the newer design when replacing an unserviceable bearing but did not prohibit installing the older part number. Investigators were reviewing how that information entered maintenance documents, the UPS program and correspondence with the FAA.

UPS records showed that general and detailed visual inspections of the aft mount, required at a 72-month interval, had last been completed in October 2021. A lubrication task had been performed in October 2025. More specialized inspections were due at higher cycle counts than the airplane had reached. The preliminary report therefore does not say those cycle-based inspections were overdue. Their thresholds and adequacy are separate issues for analysis.

Flight data and immediate safety action

The update described nominal engine and aircraft performance until about 17:13:11, when the No. 1 engine data became unreliable. A left-engine fire indication followed. Commands to the two remaining engines increased. Data from the center engine showed some fluctuations, while the right engine's rotor speeds increased and remained steady until shortly before the recording ended.

The airplane's recorded groundspeed was about 184 knots for much of those final seconds. The NTSB said its performance study would examine why altitude did not substantially increase after the left engine and pylon separated and how the loss affected handling. This does not support a definite claim that another engine failed or that a particular pilot action explains the failure to climb.

UPS grounded its MD-11 fleet on 7 November. The FAA issued an emergency airworthiness directive the next day prohibiting further flight until approved inspections and corrective actions. A superseding directive on 14 November included DC-10 aircraft because of their similar design. These actions were precautionary protections taken while the investigation continued, rather than a substitute for the eventual cause determination.

A hearing, not a final finding

The NTSB held a public investigative hearing on 19 and 20 May 2026. Its purpose was to gather facts and evidence. At the close, the chair explained that investigators would analyze the record and prepare a final report containing the probable cause and safety recommendations. Those later conclusions would be broader than the hearing's topics.

As of 5 October 2026, the official case remained ongoing. This account uses the preliminary report, the investigative update currently linked by the NTSB and official hearing information. It preserves the observed engine separation and material findings while leaving the unresolved failure sequence, maintenance implications and climb performance to the final analysis.