Exactly two years ago, on November 1, 2024, at 11:52, a concrete slab half a football field long and weighing about 190 tons came down in four seconds over the entrance to Novi Sad railway station. Sixteen people died.
Formally, this was a 1964 building, renovated by 2024 by a Chinese consortium under the Belt and Road Initiative. Formally, it is already the second anniversary, and the trial, despite every promise, has still not begun. But the real story is not about politics. It is about what an engineering failure looks like when no one had any right not to notice it. The debris yielded documents: more than 1,200 volumes of design documentation in which this canopy is not mentioned at all.
This is the story of how 190 tons of concrete hung from eighteen rods that no one remembered in a single document for ten years.
Let’s start with physics. The canopy is a rectangular slab, 48.38 by 7.05 meters, 45 centimeters thick, suspended 4.5 meters above the ground. Its front rests on massive reinforced-concrete columns; it is held by 18 identical tie rods running at a 70-degree angle to the cantilevered overhang of the famous “sawtooth” roof. The two outermost rods are vertical; the rest are inclined, about 8.65 meters long. The rod diameter is 150 mm, but it is not solid steel: inside a ribbed tube is a bundle of six 7-mm wires, prestressed and cement-grouted; around them are eight 12-mm bars; outside is the tube, and the whole thing was then encased in concrete. A multilayer composite.
It helps to remember one key figure immediately: the safety factor of those rods was 2.25. Witnesses from FTN—the Faculty of Technical Sciences in Novi Sad—explained at trial: the rods could have taken another 40–50 kilograms per square meter, an increase of about 10% in load. It seems like a margin.
But directly beneath that lies the second figure from the same testimony: corrosion 25–35 centimeters below the upper anchorage, in a zone invisible from the ground. The experts: “the collapse would have occurred even without any reconstruction—given the corrosion, and probably sooner.” The works did not create the problem. They simply failed to notice it, though they were obliged to: the condition of the rods “should always have been of particular importance,” and “inspection is not limited to a visual examination.” The reconstruction merely shortened the list of things that might collapse first.
The story of the documents is worth recounting in detail, because it explains where the whole chain came from.
The protests began on November 5. On November 22, students from the University of Belgrade were beaten at a highway blockade. Then came roughly sixty faculties at the country’s four largest universities under blockade, with demands addressed to the government, the prosecutor’s office, and the courts. And demand No. 1 was not “punish the guilty,” not “resignations.” Here it is: Publish the reconstruction documentation.
It was a brilliant engineering-legal trap: punishment is discretionary; comparing documents with reality is not. The documents either have been published or they have not; they either match what was built or they do not.
The government behaved strangely from the outset. For 40 straight days, the ministry rejected every request—from journalists, organizations, citizens—on the grounds of “protecting the criminal prosecution.” On the morning of December 11, 2024, Prime Minister Vučević called the students’ demand “strange and incomprehensible” and said that the documents had been “taken by the prosecutor’s office.” The next morning, following a “request” from President Vučić, the government decided to publish the documents.
They published 195 documents. Transparency Serbia reviewed them and found that the contract which should have been the first item in the folder was missing: the contract of 28.05.2018 between the Government of Serbia and the China Railway International / China Communications Construction Company consortium. The key contract was absent. Transparency Serbia openly called the official explanation—“it cannot be released without the consent of the Chinese side”—groundless: such agreements contain no such provision.
Still, even without that first contract, the picture emerging from the 195 documents was eloquent. An expert group from the University of Belgrade, assembled at the request of the blockaded faculties, issued a report in March 2025: the documentation was incomplete and formally defective; the April 2024 technical inspection on the basis of which the building was given a year of trial operation was hardly reliable—it concerned only the connection to the electrical grid, not the building; work on the station building was declared only 15 days before the collapse (16.10.2024); subcontractors were hired outside the tender process; the technical-supervision commission was appointed by the contractor itself rather than the client; in December 2022, the works were somehow accelerated, but the revised schedule was never published; “phantom companies with minimal capital, enabling the extraction and laundering of money.” And this was all from an official expert group assembled at the students’ request.
Formally, we have three competing expert reports, and none was conceived as a “defense theory.”
The first is official: FTN No. 01-2907/1 of 20.11.2024. Conclusion: the collapse was caused by corrosion of the steel wires in the upper anchorage zone of the tie rods. Testimony by FTN experts at trial (spring 2025 transcript): visible corrosion stains where the rods enter the roof can already be seen in photographs from 2014.
The second is dissident: engineer Slobodan Maldini’s theory (analysis in NIN, February 2025). The real support for the canopy was not the rods, but prestressed cantilever beams running along the facade. The rupture began in beam H/19—at the anchoring node of a prestressing tendon. Part B (which collapsed, with the rods) and Part C (which survived, without rods, the restaurant terrace) were different structural systems. The rods were “ripped out” because they were not the primary load-bearing elements.
The third is systemic: by professors Dobrić and Fric (Faculty of Civil Engineering, University of Belgrade, April 2025). Let’s forget “who is guilty” and look at the chain of decisions. One permit for all 108 km of the Belgrade–Subotica line (October 2021): 25 track segments, 65 crossings, 17 stations, 70 archaeological sites—in one document. The canopy is not mentioned in it—not in the design brief, not in the location conditions. The application for work on the station building was filed on October 16, 2024, 16 days before the collapse. The “as-built survey design” was produced in August 2024—after the station’s July opening, without a technical inspection. Weather cannot explain it: Novi Sad was +20.5 that day, with light winds and no precipitation.
All three reports converge on precisely one point: corrosion of the rods in the upper anchorage zone. The difference lies in what exactly each report explains.
And now for the most engineering-heavy part, where structural logic meets materials science. Prestressed galvanized wire, after cold drawing, carries residual stresses—and is therefore susceptible to hydrogen embrittlement and stress-corrosion cracking. This is not a theoretical vulnerability. It is an experimental curve:
Notice the asymmetry: strength declines by percentages; ductility is cut in half. The wire formally still “holds,” but it stops stretching—it snaps brittlely, without warning. That is exactly how the Morandi Bridge in Genoa collapsed in 2018 (43 dead), Nanfang’ao in Taiwan in 2019 (6 dead), Xiaonanmen in 2001. Always suddenly, always “the cause was a sudden brittle fracture.”
There are grounds for local criticism of the reconstruction—the roofing layers beneath the rod anchors were relaid during the rainy season, water ran down the valley between roof slopes directly toward the anchors, and leaks were recorded even after the repairs. But none of this has a major proven contribution to this specific failure: corrosion around the anchors is visible in photographs from 2014, long before the reconstruction. If FTN is to be believed, the collapse would have happened without it. And that is the most uncomfortable conclusion in the whole story: the system functioned normally, through normal degradation. The failure was not in the metal. It was in the procedure.
The full version of Dobrić and Fric’s systemic audit is effectively a manifesto of a decade of procedural acceleration. The essentials:
The inspection on November 2, 2024 shut down the construction site. Not the building. The building whose canopy had fallen, and on which there had been neither a technical inspection nor active work applications, was left untouched by the inspectors.
The chronology after the collapse is a textbook case of how a technical failure becomes political. I’ll recount it briefly.
November 2024: Minister Vesić resigns on the day of the collapse; 11 people detained on November 4–5; protests, 16-minute vigils of silence, faculty blockades. December 2024: on December 30, charges against 13 people. January 2025: Prime Minister Vučević steps down after people from his son’s circle beat a female student. March 2025: the 16th victim dies—19-year-old Vukašin Crnčević, after four months on a ventilator; March 15 sees the largest protests in Serbia’s history since the fall of Milošević. May 2025: a parliamentary commission of inquiry reports that the completion deadline was shortened “on instructions from the Presidency” to 31.12.2024.
Then begins a judicial labyrinth worth seeing in full. In two years, the case has not once reached a hearing.
Summer 2025: the Novi Sad Higher Court dismisses the indictment as it concerns top officials. September: the prosecutor’s office files a new indictment, against 13 people. November: the appellate court confirms the indictment—all must stand trial. And immediately afterward, in December 2025, the Higher Court’s pretrial panel terminates the case against six top figures—Minister Vesić, Infrastruktura železnice head Tanasković, Deputy Minister Dimoski, and three supervisory officials. The court’s wording: the prosecutor’s office filed the indictment “after only 38 days of investigation.” For the remaining seven—all implementers: the director during the reconstruction period, designers, construction supervisors—the indictment was confirmed.
The reasoning for termination was virtuoso: in the court’s view, the prosecutor had not established that the officials even had a “duty to comply with technical safety rules”; the law on public administration cited by the indictment did not concern technical matters. The prosecutor’s office appealed, and in June 2026 the Novi Sad appellate court overturned the termination—the indictment was once again in force, and the trial could finally have begun.
But the same canopy also has a second, Belgrade case—over elements of corruption in the reconstruction; Vesić and Dimoski have spent the entire time under house arrest in connection with it. On July 13, 2026, the Belgrade appellate court returns it for the third time—and again not on the merits, but over jurisdiction: whether the case was properly reassigned from the Belgrade Higher Prosecutor’s Office to the Organized Crime Prosecutor’s Office. On September 23, 2026, the defense for Vesić, Dimoski, and Tanasković files a “request for protection of legality” with the Supreme Court—a demand for a legality review; the preliminary hearing in the Novi Sad case is again not scheduled.
Against this backdrop, on September 27 Vučić resigns as president—to head the SNS list in snap elections. On October 25, the student list “Students Win” takes 141 seats out of 250; Vučić’s bloc takes 103; Vučić acknowledges defeat.
And the final touch, one only reality could invent. On October 22, 2026, the acting justice minister files a criminal complaint: someone forged a pardon decree—document No. KPRS 244 of 27.09.2026 bearing Vučić’s signature, which “releases” five defendants in the canopy case, including Vesić, Tanasković, Dimoski, and Šurlan. The forgery had been pushed through social media. Why it was plausible: on the day of his resignation, September 27, Vučić really did announce that he had pardoned 49 people “against the opinion of the services”—without naming them.
The symmetry: a state that spent two years refusing to publish 195 documents ends unable to prove the authenticity of its own pardon decrees.
Today at 11:52—at the exact minute of the collapse—FTN students will gather at the station. A route of three columns, converging at the Futoška Market. 16 minutes of silence—for the number of dead. This will no longer be only a memorial ceremony, but the second anniversary—and the first under a new government, a government of students being formed from 141 seats, half held by PhDs.
The canopy did not die of reconstruction—it died of procedure. A safety factor of 2.25 looked sufficient; corrosion at the upper anchorage had been visible since 2014; “inspection is not limited to visual examination”—and none of it was done by anyone for ten years. The difference between the “engineering” and “managerial” view of failure is that the engineer asks, “What broke?” The manager asks, “Who can we pin it on?” The documentation promised a double corrosion-protection barrier and a design service life—they did not exist.
The demand to “publish the documents” proved stronger than the demand to “jail the guilty.” Because documents can be verified, while guilt is a matter of judgment. The students intuitively found the only form of protest that could not be bought, outmaneuvered, or folded into a controlled opposition: the 195 documents turned out to be incomplete, and their incompleteness was documented. That is how the genre of the university blockade was converted into a parliament of PhDs.
The judicial system functioned as a passive filter: it filtered out the political figures, let the technical ones through. Over two years, the case against top officials fell apart three times—and each time not on substance (“they did not kill anyone”), but on procedure (jurisdiction, 38 days of investigation, “duty to comply with technical regulations not established”). At the same time, 7 implementers are headed for trial, yet no hearing has been scheduled—and political chronology is added to the “technical” one: a forged pardon decree that the government itself submits to the prosecutor’s office.
Corrosion of prestressed wires is a timer that ticks independently of political approval ratings. Of all the materials-science curves (−33% fatigue life at 5% corrosion, −51% ductility at 10%), one conclusion applies globally: the entire stock of prestressed structures from the 1960s–80s around the world is living on borrowed time. 190 tons of concrete suspended from 18 rods is not an exception, but a standard architecture of the era. The question is not “if,” but “who will inspect the rods, and when.”
Subjective opinion: what gets me most in this story is not the corruption or the elections, but the word “inspection”—and how quietly it degenerated into “visual examination.” The entire structure was designed to make failure impossible: rods with a 2.25 safety factor, a composite cross-section, a double barrier. But every layer of protection found its own quiet way to fail: the document—“canopy not mentioned”; the inspection—“close the construction site”; the expert assessment—“examined the wrong zone”; the court—“duty not established.” Not one failure looks like malicious intent; and that is precisely why the system as a whole became lethal. A brittle rupture is always sudden, always without warning. And when there are many warning layers, yet all of them fail simultaneously—that is not an accident. That is the constitution of the system. Today at 11:52 in Novi Sad, people will keep silent for 16 minutes—and that is the only inspection this system has passed as designed.