News
SpaceX says Crew Dragon capsule exploded due to exotic titanium fire
SpaceX has announced via an official update and conference call the preliminary results of a failure investigation convened immediately after Crew Dragon capsule C201 exploded in the midst of an April 20th static fire test.
Hosted by SpaceX Vice President of Mission Assurance Hans Koenigsmann and NASA Commercial Crew Program manager Kathy Lueders, the call provided some minor additional insight beyond a fairly extensive press release issued just prior. According to the preliminary results from SpaceX’s failure investigation, Crew Dragon’s explosion was unrelated to the spacecraft’s propellant tanks, Draco maneuvering thrusters, or SuperDraco abort engines. Rather, the cause lies in a more exotic and unanticipated chemical/material interaction between a plumbing valve, liquid oxidizer, and a helium-based pressurization system.
When metal burns
According to Hans Koenigsmann, SpaceX is approximately 80% of the way through what is known as the fault tree, essentially meaning that the failure investigation is 80% complete. That additional 20% could certainly throw some curveballs but the SpaceX executive was fairly confident that the results presented on July 15th would be representative of the final conclusion.
The ultimate (likely) cause of Crew Dragon’s extremely energetic and destructive explosion centers around the spacecraft’s extensive SuperDraco/Draco plumbing and its associated pressurization system, which uses helium to keep the pressure-fed engines, propellant tanks, and feed lines around 2400 psi (16.5 megapascals). Necessarily, this method of pressurization means that there is direct contact between the pressurant (helium) and the oxidizer/fuel, thus requiring some sort of valve preventing the pressurized fluid from flowing into the pressurization system.
During flight-proven Crew Dragon capsule C201’s April 20th static fire testing, that is reportedly exactly what happened. Over the course of ground testing, a “check valve” separating the pressurization system and oxidizer leaked what SpaceX described as a “slug” of nitrogen tetroxide oxidizer (NTO) into the helium pressurization lines. Around T-100 milliseconds to a planned ignition of the vehicle’s 8 SuperDraco abort engines, the pressurization system rapidly “initialized” (i.e. quickly pressurized the oxidizer and fuel to operational pressures, ~2400 psi).
To do this, helium is rapidly pushed through a check valve – designed with low-molecular-mass helium in mind – to physically pressurize the propellant systems. Unintentionally, the NTO that leaked ‘upstream’ through that valve effectively was taken along for the ride with the high-pressure burst of helium. In essence, picture that you crash your car, only to discover that your nice, fluffy airbag has accidentally been replaced with a bag of sand, and you might be able to visualize the unintended forces Dragon’s check valve (the metaphorical airbag) was subjected to when a “slug” of dense oxidizer was rammed into it at high speed.
In itself, this sort of failure mode is not hugely surprising and SpaceX may have even been aware of some sort of check valve leak(s) and accepted what it believed to be a minor risk in order to continue the test and perhaps examine Dragon’s performance under suboptimal conditions. What SpaceX says it did not realize was just how energetic the reaction between the NTO and the check valve could be. SpaceX’s understanding is that the high-speed slug of dense NTO was traveling so fast and at such a high pressure that, by impacting the titanium check valve, it quite literally broke the valve and may have chemically ignited the metal, thus introducing a slug of burning NTO into the liberated NTO system itself – effectively a match tossed into a powder keg.
It’s unclear if the ignition came from a chemical reaction between titanium (a technically flammable metal similar to magnesium) and NTO, or if the source came from the titanium valve being smashed apart, perhaps quite literally creating a spark as metal debris violently interacted. Either way, the solution – as SpaceX perceives it – is the same: instead of a mechanical check valve (simple but still not 100% passive), the barrier between pressurant and oxidizer (as well as fuel, most likely) will be replaced with something known as a burst disk. According to Koenigsmann, only a handful (~4) of those valves exist and thus need to be replaced by burst disks, a relatively fast and easy fix.
Burst disks are single-use and inherently unreusable, but they are also completely passive and simply do not leak until subjected to a specific amount of pressure. Because they are single-use, they can’t be directly tested prior to flight, limiting some of the in-principle reliability for the sake of an extremely leak-proof barrier.
Ultimately, both Koenigsmann and Lueders went out of their way to avoid answering any questions about SpaceX’s Crew Dragon upcoming test and launch schedule and what sort of delays the explosion will ultimately incur. Both individuals were nevertheless upbeat and by the sound of it, delays to Crew Dragon will be far less severe relative to delays caused by a pressure vessel or engine failure. For the time being, NASA has published a tentative target of mid-November 2019 for Crew Dragon’s first crewed launch to the International Space Station, while Lueders and Koenigsmann expressed hope in a 2019 launch but refused to give a specific estimate of the odds of that occurring.
Check out Teslarati’s Marketplace! We offer Tesla accessories, including for the Tesla Cybertruck and Tesla Model 3.
Tesla released Full Self-Driving version 14.3.7 yesterday, and after about 90 miles of testing today, it is evident there are some definite fixes from version 14.3.6, which I wrote about last week and called a regression.
Within the first 40 minutes of my drive on v14.3.7, it saved me from getting into an accident with an unaware Dodge Charger driver, and some of the things Tesla seemed to miss in v14.3.6 were definitely improved. All in all, the release so far has some really great performance, and I’m looking forward to testing it further.
For now, here’s everything I noticed with v14.3.7:
Overall Improvement
Just generally speaking from a ride perspective, this was a really great experience. A lot of the hesitancy I experienced on v14.3.6 was gone. There were no instances of brake-stabbing, wheel-jerking, or any uncertain or unconfident movements. It was void of anything that I felt made it timid with v14.3.6.
The one thing I do hope to see down the road is a smaller need to adjust Speed Profiles so often. Because Tesla calls FSD “Supervised,” I’m okay with needing to hit the scroll wheel a few times a drive.
However, I hope that things can be incrementally improved upon with speed. Sometimes it’s too fast; other times it’s too slow. It’s a difficult thing to hone in and refine, but I hope it eventually gets there.
I didn’t notice any significant left lane camping or any behaviors that were completely out of line. I am hopeful that this opinion does not change, but after driving a few days with this version and putting it in a variety of different situations, you are exposed to more behaviors, some of which are not necessarily what I’d prefer.
The big things to notice, at least in my experience thus far, are that the major issues with previous versions — meaning the braking stabbing and wheel jerking — simply weren’t there. That’s enough to already consider this progress compared to .6.
Manual Signal Override is More Responsive
On .6, I had quite a few issues with FSD ignoring my manually input turn signals. If Tesla wants to call it “Supervised,” then the car should not ignore any input the driver gives. If I touch the accelerator on FSD, the car speeds up.
🚨 Tesla FSD v14.3.7 obeying manual turn signals https://t.co/6eqToXpQfC pic.twitter.com/vHBlFQ4PDV
— TESLARATI (@Teslarati) August 2, 2026
The car did a great job of obeying my turn signals when I wanted it to change lanes, which is welcome.
Parking Lot Performance
Before .6, I traditionally took over in nearly every parking lot my car entered, because I knew it would not park somewhere that I wanted, and usually, it was just a tad too timid in this setting.
The one bright spot of .6 was how well it handled parking lots. This continued with v14.3.7:
I’m always really happy to see progress at all, but once parking preferences come to FSD, as long as this performance is still around, that could potentially be the biggest improvement I’ve seen in FSD in the year I’ve been using it personally on a daily basis.
Full Self-Driving Averts Disaster
A Dodge Charger changed into my lane without checking if I was there, running me off the road. FSD made the initial avoidance maneuver; I grabbed the wheel out of instinct, looked in my side mirror to ensure I had nobody following closely behind, hit the brake, and straightened the car back up to avoid a curb:
🚨 Guys this is why you all NEED to stay vigilant behind the wheel, even on Tesla Full Self-Driving
Human drivers are UNHINGED and have no idea what they’re doing anymore. This was a kid obviously younger than 20 years old with zero awareness.
First drive with v14.3.7 https://t.co/1vTbCMpCn8 pic.twitter.com/lz7KKEF6bj
— TESLARATI (@Teslarati) August 2, 2026
There have been quite a few responses to this video stating that I should never have grabbed the wheel. To be honest, I really wish I had not done so, because I do believe FSD would have avoided any sort of collision with anything, including the car or the curb.
However, this was the first time I had ever been this close to being hit while using FSD. My natural reaction was to take over. I think if I had had something like this happen before, my reaction might have been different.
Hitting the brake avoided hitting the curb, while FSD swerved to avoid the car. My concern after the car was clear of my front end was the curb. All in all, I’m really happy with how things turned out, and I think anyone could be a critic of how I handled it. I only had a split second to really make a decision, and thankfully, any damage was avoided.
It is clear FSD managed to avoid the car coming down before I was able to. I truly credit FSD for avoiding the collision.
What Needs to Improve
Better Recognition of Potholes, Uneven Roads, Sharp Changes in Roadway/Bumps
On Friday, my Fianceè and I were in the car, and FSD was driving us. We crossed over a roadway that has a traffic light, and FSD was traveling at 40 MPH on Standard, 5 MPH over the speed limit. Everything was more than reasonable.
However, the road we were crossing at the light has a major bump both as you start and finish crossing it. Without a speed reduction, your car can go airborne. The Tesla did just this on Friday on v14.3.6; it was an uncomfortable bounce that pretty much confirmed I would not ever let FSD go over again unless we were sitting at that intersection when there is a red light.
I even tried scrolling down into Sloth quickly, but I ended up just taking over:
This is that big bounce that I mentioned in the quoted post.
It’s just a tad too drastic to take at the speed FSD wants to go over it. You can see me quickly swipe down into Sloth, but I intervened. https://t.co/K20PK9ysBg pic.twitter.com/81Oc82ZJcZ
— TESLARATI (@Teslarati) August 2, 2026
A few people have said it remains related to the vision-based approach and its difficulty comprehending 3D. This is a huge issue because this can cause serious damage at certain speeds.
Navigation
Nothing new here. I still turn off “Online Routing” quite frequently to get the car to take logical routes from time to time.
Auto Wipers
Auto Wipers are just plain bad. I really hope Tesla just uses a rain sensor. I thought they had improved at one point, but I still get dry wipes, Speed 4 on a drizzle, and Speed 2 on a steady rain. In reality, these should be switched.
You can watch our full review of Tesla Full Self-Driving v14.3.7 below:
🚨 Tesla Full Self-Driving v14.3.7 saved me from an accident! FULL REVIEW: https://t.co/1vTbCMpCn8 pic.twitter.com/9mHmKVoMVA
— TESLARATI (@Teslarati) August 2, 2026
SpaceX will report second quarter results after the market closes on Tuesday, August 4, marking the first time the company has opened its books to the public since its record IPO in June. Management will host a live audio only webcast at 4:30 p.m. ET, streamed on X, with no dial in option.
The debut carries more weight than a typical first quarter as a public company. Two trading days after the release, on August 6, the first tranche of SpaceX’s lockup expires, freeing roughly 911.5 million insider and employee shares, worth well over $100 billion at current prices and the largest such release in Wall Street history. A second, larger tranche tied to the stock trading 30 percent above its $135 IPO price never triggered, since shares have spent most of July trading below that price.
Wall Street’s models point to revenue near $6.9 billion for the quarter, up sharply from the $4.69 billion SpaceX reported in the first quarter, with a narrower per share loss than the $1.27 posted three months earlier, according to estimates compiled by Motley Fool. Those numbers will be the first look at how SpaceX’s three segments, Starlink, launch and AI, are performing independently.
SpaceX scores another massive Pentagon deal to support military satellites
Investors heading into the call have a specific list of questions. How many net new Starlink subscribers did SpaceX add after ending March with 10.3 million, and is average revenue per user holding up as the service expands into lower income markets. How much of the AI segment’s revenue reflects contract signings with Anthropic, Google and Reflection AI this year, deals that combined could annualize to nearly $28 billion if fully ramped. Whether capital expenditures, which nearly doubled in the AI segment alone between 2024 and 2025, are still accelerating or starting to plateau. And whether management offers any forward guidance at all, something SpaceX has never done publicly.
The report will also land days after Elon Musk publicly denied a Wall Street Journal report describing internal planning to separate Tesla’s China business ahead of a potential Tesla-SpaceX merger. Whether Musk or SpaceX executives address that speculation on the call, even indirectly, maybe something investors will be listening for on Tuesday.
As Teslarati reported after Musk’s own warning to short sellers last week, the CEO has made clear he expects skeptics to be proven wrong over time. Tuesday will be the first chance for the numbers themselves to make that case.
Elon Musk
SpaceX’s Starship just got filmed by its own cargo
SpaceX released new footage of Starship in space captured by the Starlink satellites it deployed.
SpaceX released a new video Friday evening showing Starship from an angle showcased by its own Starlink satellites, watching the rocket drift away in orbit.
The 65 second clip, posted on X, stitches together footage from four cameras mounted on a single Starlink V3 satellite. It opens with a close view of Starship’s 171 foot upper stage, still catching sunlight, then pulls back as the two spacecraft separate.
The footage comes from Starship’s 13th flight test, which launched July 24 from Starbase after a scrubbed attempt and an abort caused by an engine issue the week before. When Flight 13 finally flew, it carried the first batch of functional Starlink V3 satellites Starship has ever deployed, twenty of them, with six equipped with cameras meant to scan the ship’s heat shield during reentry.
View of Starship in space from a Starlink V3 satellite on Flight 13.
This composite is made of imagery from four separate cameras on a single satellite. Six of the satellites were equipped with cameras to scan Starship’s heat shield and transmit imagery down to operators to… pic.twitter.com/sNTHHz5vLr
— SpaceX (@SpaceX) July 31, 2026
Flight 13 checked most of its boxes. Starship deployed all 20 satellites, relit a Raptor engine in space, and splashed down softly in the Indian Ocean off Western Australia. Musk’s longer term plan calls for a Starlink V3 constellation of 100,000 satellites, according to a recent FCC filing, with Starship as the only vehicle capable of launching them at the volume that requires. Each Starship flight is designed to carry up to 60 V3 satellites once the vehicle reaches routine service, well beyond what Falcon 9 can carry in a single mission.
Starship is next expected to fly with an attempt at catching the ship itself with the launch tower’s mechanical arms, a maneuver SpaceX has so far reserved for the Super Heavy booster.
Tesla Full Self-Driving v14.3.7 early review: FSD saved me from an accident
SpaceX’s biggest test yet arrives this week and it’s not a rocket launch
