TRACE THE
FAILURE
MARITIME ENGINEERING

A bang on dive 80.
The next dive carried passengers.

Titan's carbon-fibre hull passed a pressure test in 2021, but nobody knew how many dives it could take. After a loud bang in July 2022, OceanGate kept diving. Less than a year later Titan imploded on its way to the Titanic, killing all five people aboard.

Follow the evidence
The white, rounded tail cone of the Titan submersible lying on the grey seafloor, open at one end, with cables and debris spread around it.
Titan's tail cone on the seafloor near the Titanic. Frame from remotely operated vehicle video released at the U.S. Coast Guard's September 2024 hearing, as aired by 6abc Philadelphia. Not AI-generated.
THE COMPANION FILMTitan Was Damaged. Why Did It Dive Again? Watch episode 9:54

Why Titan kept diving

On July 15, 2022, OceanGate’s Titan submersible was surfacing from a dive to the Titanic when the people inside heard a loud bang. TSB M23A0169, p. 39 Titan’s sensors recorded a burst of sound and a sudden shift in the gauges that measured how its hull flexed. NTSB MIR-25-36, PDF pp. 36 and 52 The company’s chief executive, Stockton Rush, explained the bang as the hull moving in its frame and decided to keep diving, according to the engineer who reviewed the data. The next dive carried passengers. Coast Guard hearing transcript, September 23, 2024, pp. 166 to 167 and 186

Less than a year later, on June 18, 2023, Titan imploded on its way down to the wreck, killing all five people aboard. NTSB, PDF pp. 9 and 15 to 17

The U.S. National Transportation Safety Board found that the layers of Titan’s carbon-fibre wall likely began to separate after dive 80, and that the hull “needed to be immediately removed from service.” OceanGate’s analysis of its own monitoring data was flawed, so the company never saw the damage. Underneath that, the company had never established how strong the hull really was or how many dives it could survive. NTSB findings 1, 5 and 6, PDF p. 73

Watch the episode

The film puts the investigators’ own pages on screen and plots the hull’s readings from before and after dive 80 against depth, the comparison investigators used to find the change.

What the 2021 pressure test proved

Titan carried up to five people inside a carbon-fibre cylinder closed by titanium domes. TSB, p. 15 NTSB, PDF p. 20 Between February 25 and March 4, 2021, OceanGate tested its final hull in a pressure chamber to the equivalent of a dive to 4,200 metres. The Titanic lies at about 3,840 metres. NTSB, PDF pp. 30 and 33

The hull survived, which showed it could take that pressure. It said nothing about how much more it could take before breaking, or how many dives it could make before it wore out. NTSB, PDF pp. 62 to 63

One hull, built once

OceanGate had seen a hull fail before. Titan’s first carbon-fibre hull cracked in 2019. Its first director of engineering told Rush the hull was “done” and refused to sign off on that summer’s Titanic expedition, and he was let go shortly afterward. The company stopped field testing, cancelled the expedition and later condemned the hull. NTSB, PDF p. 26

For the replacement, OceanGate changed the carbon-fibre material and the way it was made, and in 2020 it tested smaller models in a pressure chamber. The first imploded at the equivalent of 2,800 metres, short of the Titanic’s depth. On the second, OceanGate stopped the test at 2,500 metres because failure was imminent. NTSB, PDF p. 27

The company identified wrinkles in the carbon fibre as a likely problem and changed the construction again. The new wall was built in five layers, each about an inch thick and cured before a film of adhesive bonded it to the next. No scale model was ever built that way. The only hull made with that method was the full-size cylinder that later carried passengers. NTSB, PDF pp. 27 to 29 and 62

Neither that hull nor any model of it was tested until it broke, so OceanGate never learned its real breaking strength or how it would fail. Its design relied on theoretical models of an ideal hull. Contractors applied reduction factors to those numbers, but the real cylinder had flaws the models didn’t include. NTSB, PDF pp. 62 to 63

What every dive does to the hull

On every dive the water squeezes the hull harder as the sub goes deeper, then lets go as it comes back up. Engineers test what that repetition does to a hull: how it changes, what signs mean it’s weakening and when it has to be retired. The NTSB calls that a cycle life evaluation, and OceanGate never did one for the replacement hull. NTSB, PDF pp. 62 to 64 In the NTSB’s words, the company “was unaware of its actual strength and durability, which was likely much lower than their target.” NTSB finding 5, PDF p. 73

Titan reached the Titanic thirteen times in 2021 and 2022. Its dive numbers run much higher because they also count shallow test dives and dives with the first hull. NTSB, PDF pp. 14 and 33 to 36

Counting on the monitoring system

In place of a known retirement point, OceanGate counted on a monitoring system to show the hull weakening in time to stop using it. Titan had two kinds of sensors. Strain gauges measured small changes in the hull’s shape, and the company reviewed those readings after each dive. Acoustic sensors listened for the sounds of damage and were meant to warn the pilot during a dive. NTSB, PDF pp. 31 to 32 and 64

Rush had argued for monitoring over inspection years earlier, with Titan’s first hull. In January 2018, OceanGate’s director of marine operations reported “very visible signs of delamination and porosity” in pieces cut from the ends of that first cylinder. NTSB, PDF p. 47 His report said the hull needed nondestructive inspection before any crewed dives. Quality inspection report, p. 3 At a meeting with him on January 19, which Rush recorded, Rush said “we have acoustic monitoring which will tell me way before it fails,” and that no nondestructive testing existed for five-inch-thick carbon fibre of that size. Meeting transcript, p. 5 Four days later, OceanGate ended the director’s employment. NTSB, PDF p. 48

Members of the Marine Technology Society’s submarine committee also drafted a letter to Rush in 2018, warning that OceanGate’s “experimental” approach “could result in negative outcomes (from minor to catastrophic).” The society never formally sent it, but one signatory said Rush received it. That same year, OceanGate gave up its plan to have Titan classed, an outside review explained below. NTSB, PDF pp. 23 to 24 These 2018 warnings were about Titan’s first hull, the one condemned in 2019, not the hull that imploded.

Dive 80: the bang and the decision

The bang came as Titan surfaced from dive 80 on July 15, 2022. TSB, p. 39 Phil Brooks, then OceanGate’s engineering director, reviewed the sensor data. At the Coast Guard’s hearing on September 23, 2024, he testified that the strain data showed a shift in level, “fairly minor” and on two of the hull’s 16 strain channels, and that a sudden shift in level was “something to be concerned about, for sure.” He said he discussed it with Rush, who wanted to continue, and that the plan as he remembered it was “well, let’s dive again and see what happens.” Asked who decided to resume diving, he answered, “Definitely Stockton.” Hearing transcript, pp. 166 to 167

Brooks said Rush’s explanation, that the hull had shifted in its frame, sounded plausible to him, though he wasn’t a qualified mechanical engineer, and that the previous director of engineering, Dan Scoville, “kind of confirmed” it. He didn’t think anyone considered an uncrewed test dive. The next dive carried passengers. Hearing transcript, p. 186

The hull got only a limited look. Brooks said the insert lining the inside was never taken out, so the inspection meant shining a flashlight down a narrow gap between the insert and the hull. Hearing transcript, p. 167 The outer coating also blocked a visual check. Canada’s investigators note there was no easy way to inspect the cylinder for damage, because that would have meant disassembling it. NTSB, PDF p. 64 TSB, p. 108 That left the sensor data as the main evidence.

Plotting the strain against depth

OceanGate plotted each dive’s strain readings over time, beside a line for depth. The NTSB says that made one dive hard to compare with another. Its investigators plotted strain against depth instead, because depth tells you how hard the water is pushing on the hull. At the same depth, a healthy hull should respond the same way every dive. NTSB, PDF pp. 51 to 53 and 64

Plotted that way, the later dives looked different near the surface. In roughly the first 600 metres, the hull responded differently from before dive 80. Deeper down, the readings looked much like they had before. TSB, p. 39 NTSB, PDF p. 53

Both boards saw a pattern consistent with defects between the layers. As the pressure rises, damaged layers can be pressed together and act like a single wall again, so the readings look normal at depth even though the hull is damaged. NTSB, PDF p. 59 TSB, p. 107

Delamination between the layers

Separation between bonded layers is called delamination. Pieces trimmed off Titan’s hull during construction showed wrinkles, porosity, places where wrinkles had been ground down, and gaps in the adhesive between layers. Those gaps weakened the bonds. A recovered piece of the inner layer carried wear and rubbed adhesive, showing that layers had separated before the implosion. NTSB, PDF pp. 41 to 44 and 59

The NTSB concluded that the hull “likely sustained damage after it surfaced at the end of dive 80 in the form of one or more delaminations.” NTSB finding 1, PDF p. 73 Canada’s Transportation Safety Board was more cautious. Because OceanGate didn’t thoroughly inspect the hull after the bang, it said, “it cannot be determined whether this event precipitated the non-linear readings” on the next dive, on July 19. TSB, pp. 39 and 107

Why the sound warnings never triggered

The acoustic system’s warning levels came from earlier hulls and scale models built a different way. OceanGate never showed what the final five-layer construction would sound like as it deteriorated, or how much warning it could give. Sounds picked up while Titan was at the surface didn’t count toward a warning, and the count started from zero on every new dive. NTSB, PDF pp. 31 to 32 and 64 to 65 TSB, pp. 108 to 109 The system also had no audible alarm, so a pilot might not notice a warning right away. TSB, p. 108

On the recorded dives after dive 80, the sensors still picked up activity, but not the kind that would trigger the expected warning. OceanGate’s analysis of that data didn’t recognize that the hull had to come out of service. NTSB, PDF pp. 65 and 73

Two more Titanic dives, then the last descent

Titan reached the Titanic twice more, on dive 81 on July 19, 2022, and dive 82 on July 22. Dive 83 went to 2,954 metres, short of the wreck. NTSB, PDF p. 81 After that season Titan was stored outside, uncovered, in a parking lot in St. John’s, Newfoundland, and moved indoors early in 2023. For the 2023 season it was towed to the site on its launch platform behind the Polar Prince. NTSB, PDF pp. 36 to 38

The NTSB concluded that further damage occurred after dive 82, but its origin is unknown. Freezing temperatures in storage and loads from towing were possible sources, but their effects were never studied, the NTSB says. NTSB, PDF pp. 60 and 63 to 64 The four dives of the 2023 season before June 18 went no deeper than 10 metres. NTSB, PDF pp. 14 and 37

On June 18, 2023, Titan began dive 88 at about 9 a.m. At about 10:47, with Titan last reported at 3,363 metres, a message said it had dropped two weights, which the NTSB calls routine for slowing the descent. Then communications stopped. NTSB, PDF p. 17 The damaged hull likely buckled, losing its shape in one area, and imploded. NTSB findings 2 and 3, PDF pp. 73 to 74

On June 22, a remotely operated vehicle found the debris field, including the tail cone and both titanium domes, about 500 metres from the Titanic’s bow. NTSB, PDF p. 20 Investigators couldn’t recover the monitoring data from the final dive, so the record doesn’t show what warning, if any, the system displayed. NTSB, PDF pp. 45 and 51

What the NTSB and Canada’s TSB concluded

The NTSB’s report, dated October 2, 2025, gives this probable cause:

“OceanGate’s inadequate engineering process, which failed to establish the actual strength and durability of the Titan pressure vessel and resulted in the company operating a carbon fiber composite vessel that sustained delamination damage that was subsequently exacerbated by additional damage of unknown origin, resulting in a damaged internal structure that subsequently led to a local buckling failure of the pressure vessel.” NTSB, PDF p. 74

It names two contributing factors: U.S. and international guidance and rules that were “insufficient to ensure OceanGate adhered to established industry standards,” and “OceanGate’s flawed analysis of their pressure vessel monitoring system data, which led to their continued operation of a damaged pressure vessel.” NTSB, PDF p. 74

Canada’s TSB published its report on June 17, 2026. Its first finding is that Titan’s as-built hull was never checked against the numbers used to design it, so “OceanGate did not know for how long the pressure hull would remain safe when used repeatedly for dives to the depth of the Titanic.” It describes the failure a little differently from the NTSB: a weakened cylinder that likely failed “progressively with damage accumulating during each dive cycle until it imploded.” It also found that risk management at OceanGate was hindered by “the influence of power dynamics and social and psychological factors.” TSB, p. 120

Who could have stopped the dives

Ships and submersibles can be classed. A classification society, an independent organization, reviews the plans and construction, witnesses critical tests and keeps surveying the vessel as it ages. OceanGate planned to class Titan through Lloyd’s Register or DNV, according to its former director of marine operations. In 2018 it gave up. The NTSB says Rush decided classing would take too long. NTSB, PDF pp. 22 to 23 A review like that could have asked OceanGate to show how strong the hull was, how long it would last and whether its monitoring would give enough warning, the questions the company never answered. In our view, a review couldn’t have guaranteed safety, but it was one more chance to stop the dives.

Government oversight was another check. OceanGate called Titan “experimental” and its paying clients “mission specialists.” The NTSB notes that classification societies and the U.S. Coast Guard have “no ‘experimental’ design endorsement for vessels carrying passengers or cargo.” NTSB, PDF pp. 15 and 23 Because Titan was owned in the United States and not registered anywhere else, the NTSB treats it as a U.S. vessel, and carrying passengers for hire required a Coast Guard inspection that OceanGate never arranged. The NTSB doesn’t rule on specific violations. NTSB, PDF pp. 67 to 71

Canada’s investigators found a gap on their side too. Transport Canada knew Titan operated from St. John’s with Canadian support ships, and assumed those ships’ certification meant Titan was properly overseen. It never assessed Titan itself. The TSB found that Canada’s approach “enabled the Titan to operate without any independent verification to identify safety deficiencies.” TSB, pp. 121 and 123

What has changed since

The NTSB recommended that the Coast Guard commission an expert panel on submersibles and other pressure vessels that carry people. The panel would look at, among other things, who counts as crew and who counts as a passenger, and whether every such vessel should be inspected whatever its operator calls it. The NTSB also recommended U.S. regulations based on the panel’s findings, revised Coast Guard guidance and a push to make the international guidelines for passenger submersibles mandatory. These are recommendations M-25-012 to M-25-015; on their own they change no rule. NTSB, PDF p. 75

On April 8, 2026, Transport Canada published a Ship Safety Bulletin setting out requirements for passenger submersibles in Canadian waters. The TSB called it a step in the right direction but noted that “SSBs and policies are intended to guide behaviour and are not enforceable.” It recommended that Canada require all human-occupied submersibles to meet international safety standards, and that submersible operators and their support ships manage safety together. TSB, pp. 126 to 129

We remember the five people who died aboard Titan: Stockton Rush, 61; Paul-Henri Nargeolet, 77; Hamish Harding, 58; Shahzada Dawood, 48; and Suleman Dawood, 19. U.S. Coast Guard Marine Board report, p. 4

The film ends on a question: what should it take to resume passenger dives after an unexplained bang and a change in hull readings? Tell us in the YouTube comments who should sign off, and what evidence you would require.

Primary sources

NTSB page numbers are PDF pages. TSB page numbers are those of the report’s PDF edition. Hearing pages are the September 23 day’s own transcript pages, not the combined file’s PDF pages; the Coast Guard report page is its printed page 4. The hero image and preview artwork are frames of the remotely operated vehicle video released at the Coast Guard’s hearing, not AI images. The film’s narration voice is AI-generated; its diagrams and models are illustrations built from the reports.