Four workers killed at TSMC site as new-fab steelwork stops across Taiwan

Four workers fell with the steel they were installing. By the following morning, all four were dead. The collapse at Taiwan Semiconductor Manufacturing Co’s Fab 22 construction site in Kaohsiung has left investigators examining a suspected broken steel connection—and asking why the protection reportedly in use did not keep the men from falling. ^{1 11}

The accident happened on October 20, 2026, during steel lifting and installation at the Phase 4 construction project in Nanzih District, part of the Southern Taiwan Science Park. Three men died that night. The fourth, surnamed Huang, died in hospital on the morning of October 21 after being treated in intensive care with unstable vital signs, according to the Southern Taiwan Science Park Bureau. ^{1 18}

TSMC’s response now extends beyond the area where they fell. On October 21, the company said it had voluntarily suspended work at the Kaohsiung construction site and steel-structure operations at its new fab construction sites across Taiwan. That is a substantial construction-safety response, not an announced shutdown of operating chip factories or every kind of construction nationwide. ²³

All four men worked for Shang Hao Machinery Co Ltd, identified in Chinese as 上昊機械有限公司. The bureau’s account places them on the third-floor exterior steel structure, carrying out lifting and installation; welding was also reported. Their employer is known, but the complete chain of contractors and subcontractors has not been established in the available reporting. ^{18 11 12}

The early emergency figures were not a different set of casualties. Focus Taiwan’s first account, published at 8:36 p.m. on October 20, described two workers without vital signs and two injured. Being found without vital signs is not the same as a formal declaration of death. The later reports followed the deteriorating condition of the same four men, all of whom were taken to hospital. ^{19 18 23}

The bureau’s preliminary account focuses on bolts in a connecting plate on the exterior steel structure at the third-floor level. Investigators suspect that those bolts broke, allowing the men and heavy steel components to fall. The distinction between that proposed sequence and a final cause matters: identifying a connection that apparently failed does not yet establish why it failed or who was responsible. ^{18 11}

In an October 21 report by Liberty Times, bureau director 鄭秀絨 described bolt fatigue or an insufficient number of bolts as preliminary possibilities. Neither has been demonstrated as the cause. The technical investigation remains unfinished, and the third-floor work location is not a measured fall distance. No exact fall height, beam weight or steel-component dimensions have been verified in the available accounts. ^{11 12}

Some descriptions also locate the failure in different equipment. China Times reported on October 21 that crane bolts broke; Liberty Times’ bureau-attributed account that day instead identified suspected failure of bolts in the exterior steel connection plate. Those are materially different descriptions, not interchangeable names for an established mechanism. The more specific bureau account gives the investigation a focus, but not a verdict. ^{12 11}

The same caution applies to construction pressure. China Times characterized the project as rushing construction. In Liberty Times’ October 21 report, the bureau director explicitly rejected a connection to the construction schedule. No final technical finding resolves that disagreement. Schedule pressure cannot responsibly be assigned—or excluded as a settled investigative conclusion—on those accounts alone. ^{12 11}

The director also said protective measures and safety ropes were in use. That statement makes their arrangement a central question rather than closing the safety inquiry. The reporting does not explain how the workers were secured or why the reported protection did not prevent fatal falls when the steel came down. Establishing that equipment was present is not the same as establishing that the work was safely protected. ¹¹

The stoppages developed in stages. TSMC’s initial October 20 statement referred to voluntarily halting the related construction work. The bureau’s specific local stoppage covered Phase 4 steel lifting and installation, with the accident scene to be preserved. The next day’s company announcement widened the precaution to new-fab steel-structure operations throughout Taiwan. These measures have different scopes and should not be collapsed into a claim that all TSMC activity has stopped. ^{30 18 31 23}

Labor inspectors have been sent to investigate the cause and determine responsibility, while TSMC says its safety personnel are assisting. The stopped Phase 4 steelwork can resume only after an improvement plan is submitted and approved. There is no verified restart date or authorization, final technical report, case-specific imposed fine or prosecutor disposition in the available reporting. A work suspension also does not establish that the remaining structure has been stabilized or cleared of continuing hazards. ^{1 11 23}

Fab 22 is being developed for 2-nanometer and more advanced chipmaking technologies, and Phase 4 construction had been scheduled for completion in 2027. Those plans explain the project’s place in TSMC’s expansion; they do not establish that the accident has changed its completion date, production or customer supply. The confirmed consequences are four deaths, stopped construction and a wider pause in steelwork. ^{1 23}

The test of the response is now more demanding than announcing a pause. Before workers return to the stopped steelwork, the investigation needs to explain the failed connection, the protection that did not save the men and the responsibilities behind both. An approved improvement plan must address the danger that killed them—not merely provide a date for construction to continue.