And even Space Shuttle wasn't a demonstration of that, since the propellant tankage (the largest volume of any stage) wasn't reusable.
SpaceX has pretty conclusively shown the Shuttle architecture (stage and a half, using LH2 at launch) was not the right choice.
[1] https://www.space.com/22432-reusable-dc-x-rocket-space-hall-...
The space shuttle was retired for a reason? This was already known. The only thing it shows is that with that knowledge, you can design something better. Hindsight 20/20 etc
This is a complete apples to oranges comparison.
A more apt comparison would be the space shuttle being a kin to Vasco da Gama's discovery of the sea route to India. After that point, the concept was demonstrated and the challenge was in optimizing cost-savings of commercial endeavours.
I mean, orbital lift services have been a thing for how many decades?
How are you measuring "safe enough" in this context? My understanding is that the Shuttle didn't have programmatic probabilistic requirements for mission safety. Ie, they didn't use overall reliability estimates and instead relied on engineering judgement (which the Challenger report showed varied widely depending on who you interviewed). It's hard to make a claim on whether it was safe enough without explicitly defining the safety goal.
Early 1970s documents were given safety requirements of 1:20 flights being successful. [1] That's the closest I've found to an actual requirement, and the shuttle record falls well within that requirement. I do think such an early spec is not a good basis for final performance, though.
[1] page 326: https://ntrs.nasa.gov/api/citations/19700005884/downloads/19...
Shuttle started at 1:10 and ended with 1:90 chance of loss of crew. That was bad and got it canceled as soon as the station was built, and supposedly ISS was finished only because international obligations. The last mission to extend Hubble's life was almost canceled because the NASA admin thought it was too dangerous.
NASA wanted 1:270 (so just "3 times safer than shuttle at the end of the shuttle program") from commercial crew (i.e. SpaceX Crew Dragon and Boeing Starliner).
Current US law for commercial space works on consent: passengers sign a waiver saying they know it's dangerous, and then the company doesn't have to promise safety the way an airline does. The company is on the hook for debris and innocent bystanaders on the ground though.
SpaceX plan to fly Starship a lot and gain so much experience they can find and fix all problems and reach airline safety. At least private jet level of safety, at a start.
There are multiple layers of risk consent, from engineering, safety, medical, and crew, for NASA. Of those, I think crew are the least likely to say no because they tend to come from a very risk tolerant population.
Whatever official spec that did or did not exist is irrelevant.
If you hire a contractor to build a 2000 sqft house and then after it’s built complain to them that it’s not big enough, the problem is literally the requirement you gave them.
I don't think it's that hard. "We're not sure exactly how safe it is, and do not have any actual requirement that it meet any specific level of safety" is clearly not safe enough for manned flight. (At least not outside of war-time or other such times we're willing to take on extra risks)
Would you get on an elevator that has a weight limit listed as "hasn't fallen yet", "has only fallen that one time", or "did fall that one time, but not since the redesign". Would you trust millions of dollars of human training and several lives to such an elevator?
I think we can say that it wasn’t safe.
The point was that wasn’t explicitly defined up front…they seemed to have fallen into the same trap you did of “we’ll know it when we see it”. CCP has a loss of crew threshold of 0.37%.
We do not design civilian, optional, exploratory space missions with a threshold for acceptable fatalities. There is no number of dead astronauts that we can accept as a design goal.
Unforeseeable fatalities can happen, but death rate should not be a predetermined design parameter.
My comment is that a predetermined fatality rate is unacceptable. Not that unforeseen fatalities are unacceptable. Space programs playing Russian roulette is NOT the same thing as letting a few beetle parts into the Rice Krispies. If you can reasonably foresee that people WILL die (as happened on the space shuttle) you fix it.
I thought you were claiming that they were designing for an intentional failure rate when you said this
> The point was that wasn’t explicitly defined up front…they seemed to have fallen into the same trap you did of “we’ll know it when we see it”. CCP has a loss of crew threshold requirement of 0.37%.
I read your statement of a “requirement” of a crew loss threshold as being required losses. You seem to keep ignoring that I’m saying that no one is designing for intentional losses, and asserting that they are designing to hit a certain number of failures as a “requirement”
Ie, a reliability (survivability) requirement is functionally the same as an acceptable failure requirement. The confusion seems to be around whether that defines a maximum or a minimum.
FWIW I wasn’t deliberately ignoring that point. That’s why I made the distinction between “intentional” and “acceptable” in the GPP
That said, I do understand what you’re saying, that we can’t insist that we never start human commercial space travel unless it can be perfectly safe, and we can’t stop it every time there’s an accident. But that’s not quite the same thing as having an expected fatality rate that we find acceptable.
The linked reference does indeed give an acceptable rate. This doesn’t seem to be a curious discussion if you’re jumping to argue instead of engaging the answer provided.
It’s cool man, I think you’re wrong in your overall approach here, but it’s not worth arguing, agreed.
But if you ask a question and literally given the authoritative source but are too lazy to read it, maybe feed it into AI or something. That is, if you actually want the answer.
My pilot license has long expired, but I did work as a reliability engineer in aerospace. I'm telling you, this is how the design work operates. You are either being deliberately obtuse or misunderstanding what I'm saying. Nobody is "letting accidents slide". If a mishap occurs, it will be investigated. But when they are designing aircraft/spacecraft, failure modes are categorized on severity and probability to determine risk. Something that may be "catastrophic" must have a lower probability of occurrence than something that is "hazardous" and so on. If that risk is above a specified threshold the system must be redesigned to fall within the acceptable risk limits or, in the case of some space operations, waived by those with the technical authority to do so.
Since you follow accidents, I'm sure you are aware of the MCAS issue with Boeing. The Seattle Times had some good early reporting on how they misapplied their hazard analysis and characterization of that system if you want a real-world example with some public information available.
NTSB is an agency investigating accidents why would it define acceptable fatalities… Is it something like after an accident report they’d proclaim “enh, only 10 people died, we’re ok with that”? I just don’t see it. But sounds like you know more.
Maybe FAA has some cutoff before penalties kick in, but those are different agencies with different mandates (though NTSB is an appellate authority for FAA, maybe that’s what you meant…)
It had a failure rate of 2/135. Doing quick math its actual failure rate was about 1500/100,00 flights
But those “expected” rates are just guesstimates and not scientific. Feynman put it at about 1% given the engineering data available in 1986.
Yes. That’s why I asked for clarification. It’s like someone saying, “That costs too much“ And someone else asking “How much can you afford?” They are logically sequential statements.
You responding with “this other thing has a different price” doesn’t answer how much they can afford.
I’m sorry this so offended your delicate sensibilities. I suggest you append “do not reply with anything that isn’t exactly what I’ve asked for” to all of your future comments to avoid any more misunderstandings.
I bet there are lots of people willing to fly on Starship even if it had Falcon 9’s level of reliability even without an escape system.
Current (2026-11) FSD in “sloth” mode requires less attention (in software), and, having used it a fair bit myself, I would probably ride in it unsupervised. The Cybercab customers do.
Maybe it’s exactly as you say, maybe they have finally figured this out, but Tesla / Elon have blatantly lied, manipulated data, faked capabilities, and overpromised way too many times over the last decade for me to take them at their word. We’ll see.
It is a joy to use. It drives the way I wish Uber and taxi drivers would drive; it goes exactly the speed limit. A cop drove behind me for like five minutes straight the other day and I turned on FSD in sloth mode and knew it would be unimpeachable, and he left me alone.
Flight 13 launched 20 Starlink v3 sats for a total payload of 34,100 kg.
Flight 14 launched 26 Starlink v3 sats for a total payload of 44,330 kg.
The Space Shuttle's max payload capacity was only 22,780kg (https://i.redd.it/8t8j1n4032lh1.png)
I have yet to be convinced that Starship’s payload capacity is actually a good thing, from a big picture perspective. It looks to me like a possible case of shouting, “I meant to do that!” after a pratfall. On a months-long trip to Mars, you need a vehicle that big for the sake of the human occupants. For launching Starlink, it’s less clear whether the ability to deploy so many satellites at once was worth the vehicle’s development costs. Perhaps it’s just inherited from an earlier iteration of the program that had different goals, and a more modest design would have been more economical. Worse, so far it seems like Starlink might be it. Most other use cases people have proposed for putting 40 tonnes in orbit on a regular cadence are the kind of highly speculative grasping at straws that accompanies any technically impressive solution in search of a problem.
A lot of the vision was actually around building orbital solar arrays (remember the microwave power plant from sim city 2000?) to resolve the oil crisis.
The oil crisis went away and the microwave arrays idea didn't make much progress and shuttle' budget got cut until they could only do one vehicle, so they tried to combine both and have a reusable heavy lift manned space glider.
But because the microwave arrays weren't going to happen nor was the earlier L1 space station idea (and skylab deorbited before shuttle's first launch) they had nowhere to shuttle anything to, so instead they just ended up with a heavy lift LEO launcher that had people onboard for some reason.
Ofc they eventually got the ISS as a destination, much later.
There are many valid points about the difficulties they'll need to overcome for manned flights but there are likely many solutions as well. I wonder if it could be as simple as making a really big crew capsule capable of launch abort with its own thrusters? It would make the second stage partially single use but the capsule itself could be reused like they reuse the Dragon crew capsule.
As for large satellite launches, there's an obvious need for a larger faring volume when you look at the lengths we've had to go to in order to make payloads fit into the tiny ones currently available. Being able to launch a large telescope or solar array that needs minimal folding would be a huge cost and time saver. Having more and cheaper tonnage also saves money on materials used for manufacturing satellites, you can use less expensive metals/alloys and components if you have more weight budget. You can add more fuel to extend operational lifespan. It's also big enough that launching habitat modules becomes much more attractive. There's a lot we didn't do just because it was too expensive or simply impossible with the launch capability we had.
In it's current design, there's no way to do that. Which is also why starship can be a lander on the moon, but it's crew will board after it gets to orbit (for now, at least).
The reason the shuttle was so expensive and costly was because they used ablative cooling. SpaceX came up with some novel material produced entirely in house to make the heat shield hot swappable. They new this was a problem from the start and specifically designed this thing to be rapidly reusable, not just reusable in namesake.
Atleast, on elon time it's super easy to get these haters to eat their words. They are producing one of these every month, so he will see how fast they scale from here. Even if it takes a month to referb and put back on the pad, they are scaling exponetially pretty fast.
Even if they have to do that every flight, which I doubt, SpaceX can achieve a very high flight rate with relatively few Ships.
The whole “catch the giant rocket” thing is (was?) the hard part
Starship has taken this long because of over optimistic project management. Elon doesn't like bad news about difficult issues. He's run a massive rocket project like a mobile app development run with an agile task board. In contrast, Apollo was a triumph of conventional systems analysis style project management.
Now, would I let my children fly on one? I'm hesitant now, and I'd like to see how they handle the flip maneuver with humans aboard. Bit no hard yes or no.
I assume they'll do the latter. If the vehicle is truly rapidly reusable, that's less crazy than it sounds. Just fly it 10,000 times and fix all the problems you find :) Escape systems are only really useful during launch anyway. The only way to protect the crew from a failure during reentry is reliability.
The previous flight launched 20 sats and that is the equivalent (in Starlink bandwidth) to 10x F9 launches.
This one launched 26 sats. That'd be 13x F9 launches in Starlink bw.
F9 internal cost is 20M$.
Starship stack is rumoured to cost ~100M$ to build.
If they can launch a stack once every month, they've already achieved more "capacity" for Starlink than their best year using F9s (last year). At half the price.
So that's already proven, even if nothing else works (i.e. reusability, etc). But of course they'll keep working the problem and make it work. They're also scaling at a crazy pace (3 more pads in various stages of readiness, plus even more planned in the next ~5 years). They've just finished the gigabay and they're planning even more in Louisiana.
Anything they do from here on out is "the cherry on top".
Are we ready to put passengers and NASA interplanetary missions on Starship? No.
Is SpaceX ready to count on Starship for replenishing or expanding their main moneymaking satellite fleet? Yes.
If you just look at incentives combined with the large change in the underlying assumptions of space flight, it’s quite clear that customer base will expand if SpaceX wants it to.
a) EU and China both are actively working on it and
b) > No other nation
Starlink also doesn't belong to a nation, it's owned and run by a private company, which will probably cause issues for the US government at some point in the future
> An insider source told Bloomberg that each fully stacked Starship actually costs hundreds of millions of dollars. How many hundreds of millions? They didn’t say, but this is more evidence to support my estimated build cost of a fully stacked Starship at $500 million. This figure didn’t come out of thin air. It was derived by scaling up Falcon 9 build costs to Starship’s weight to obtain a rough estimate and then verifying it against SpaceX’s Starship expenditure, minus overheads, and dividing it by the number of launches (read more here). This suggested a build cost of $500 million. Now, this is still a very rough estimate, but the fact that it aligns with this SpaceX insider’s statement suggests we are at the very least close.
The Starship program cost is $15B and counting. It will like get to $20-30B+, particularly when you start factoring in aspects like crewed missions, in-orbit refueling and lunar landing and takeoff. You don't have to do all of these capabilities but any you drop remove certain markets.
Likewise, any production line needs to be scaled. There is a cost to that too.
All of which to say is that $500M/launch currently I find entirely plausible.
But what's the market? Because AFAICT it's basically "Starlink". How many launches will that sustain? Well, that depends on the market for Starlink. Starlink receivers obviously have uses for mobile broadband, particularly on planes and ships. But th ereceivers are expensive. How many people will be able to and willing to pay $100/month for mobile broadband?
I was curious about the possibility for a Starlink based phone communication networks and the prospects are... not good. Because physics. 5G phones typically have to communicate only a few miles. Depending on the altitude and it's overhead, a Starlink phone has to communicate with something 300-1000km+ away. You'll lose service indoors. The signal just isn't powerful enough. Raise the power and the phone gets bigger, hotter and more expensive with worse battery life.
It's like SpaceX has decided to keep throwing more and more balls in the air and the solution thus far has been to throw them really hard so they don't come down for awhile. But, given there's a limited return because force is geometric, eventually it's not going to end well. Even the whole Ai angle was Elon using SpaceX to rescue himself and his "investors" from xAI, which itself was used to rescue Elon from paying $44B for Twitter.
[1]: https://wlockett.medium.com/things-arent-adding-up-145732c9c...
I can't read that article, but I've seen some of the author's other stuff and he gets basic facts wrong. For example, he claims that Starlink is unprofitable, and that higher RAM & fuel prices will make it moreso.[1] But he's ignoring that SpaceX's financial filings already account for depreciation, and that propellant & RAM are a tiny fraction of costs. Both could double and only increase SpaceX's costs per satellite by a few percent. He also claims that, "...a large portion of the Starlink constellation consists of V1 and V1.5 satellites..." but they're less than 15% of the bandwidth of the constellation. Over half of the v1 & v1.5 satellites have been deorbited.
As far as I can tell, he's throwing as many criticisms at SpaceX & Tesla as he can think of and hoping some will stick. He's arguing like a lawyer, not a researcher. That's not a good way to figure out what is true or what is likely to happen in the future.
Ultimately spaceX aren't profitable. That's about the best we can say, because the other numbers can be fudged
What matters is the marginal cost of building a new Starship, which (currently) is around $80-100m.
There's also glaring errors - it claims the $70m Falcon 9 customer price actually costs SpaceX $67m internally, implying a 3% profit margin - which is absurd and fiscally impossible given SpaceX's SEC filings. The internal launch cost is closer to $15m.
I'm not aware of government subsidizing starlink. They do pay for starshield tho which is another revenue stream from constellation.
In the IPO, it was clear that the launches were not profitable.
Once Musk established DOGE, SpaceX basically started getting a shitload of contracts, as well as enjoyed the ability to operate with way less regulation.
Starship doesn’t need a 1960's style launch escape system. The whole ship is the crew vehicle. If the booster fails, Starship separates and flies away on its own Raptors. That’s the plan SpaceX has stuck with for years. NASA requires high reliability and failure tolerance for crewed flights, not a regulatorily mandated escape system.
If you extrapolate the safety record of Falcon onto Starship, which seems conservative given that Starship is going to be their focus, it will be far safer than the Space Shuttle.
The Space Shuttle blew up 2 out of 135 flights, 1.5%. Falcon has 3 blow-ups out of 720+ flights. That's more than 20x as safe, and includes the aggressive/early R&D.
Starship is still being developed. Unless they go under somehow, it's going to be safer than the Space Shuttle by similar orders of magnitude.
stop repeating this lie, please
shuttle costs per-launch were on falcon 9 level
the only reason for the cost was lack of payloads (!) to spread per-year costs
there's a reason why spaceX now gets far far FAR largest part of its income from Starlink and most most MOST of its launches are Starlink. It's sattelite company with launch side business
Huh, in what world is that true? SpaceX quoted $75 million to external customers per launch according to their IPO.
Shuttle launch costed $1.0 billion in 1992, or $2.41 billion today, inflation adjusted. That is without R&D cost.
Thus, average cost per flight including development costs and flights through FY 1993 is $1.6 billion. Average cost per flight not including development costs or flights, up to FY 1992, is close to $1.0 billion. - [0]
How much was the maintenance, refurbishing cost for each launch? the cost of a new external tank for each launch was $52 million in 1992, or $120 million today - [1].
How much do you think they could get it down if they could launch more?
Finally, do you think NASA could fly 150 times annually even if there was demand, when it only flew 135 times in the shuttle's 30 year lifetime?
[0] - https://sciencepolicy.colorado.edu/admin/publication_files/1...
why not? if there was demand, especially internally from the government as planned - it would've been accommodated
> Huh, in what world is that true?
in the world where demand for Shuttle was 150 annually and not 135 in 30 years, duh?
> How much do you think they could get it down if they could launch more?
my main point is the per-year costs, like all the crew training complexes
LARGE part of the billon$ number you quote comes from that and would get spread a lot. 50 million$ is a nice floor to exemplify where it would've ended up
Mind to share how did you get that?
> why not? if there was demand, especially internally from the government as planned - it would've been accommodated
Which demand? For how many launches?
> in the world where demand for Shuttle was 150 annually and not 135 in 30 years, duh?
So the demand is in your imagination?
Challenger is largely true (5 years is about the same as CRS-7), but iirc demand disappeared already at the late development stage
I do call it a lie because people mix up program running costs and rocket launch costs. If your factory costs bajillion dollars to keep lights on, but nothing extra to increase output 100-fold, then it is expensive only on low demand. Proclaiming that it's always expensive and thus a failure misunderstands how mass production works
Falcon rockets don't seem to be exploding on ascent anymore, they're doing 2-3 launches per week now, no explosions in a decade.
Perhaps an escape system could be deployed from the dorsal side of Starship, too.
Given their track record of achieving the impossible, I wouldn't bet against SpaceX as they head towards commercial deployment of Starship.
What makes you think humans will be the primary passengers? My bet is they will send Optimus for the majority of "manned" missions, especially 5 years from now
Why does it need to be so much more reliable/safe than any other rocket that's existed before?
Seems unfair to hold it to a different standard than anything else?
Sir, you are a programmer of Java applications for mobile phones. You’ve had an impressive career, and are a person I am honored to look up to, but please have some humility
SpaceX has actually replaced all the heat shield tiles in a couple of days, so even if they all needed replacing it would be a huge problem for rapid reuse.
SpaceX has tested metallic tiles on some flights and determined they didn’t work well.
For example, on flight 14 you claimed that SpaceX fudged their payload numbers by not fully fueling the satellites.[1] I pointed out that this must be false for two reasons: 1. Propellant is less than 5% of the mass of the satellites and 2. The FAA requires accurate masses for debris assessments.
You also suggested that flight 14 wasn't fully fueled to allow for RTLS, but the amount of propellant burned is the same for a water landing and a tower landing. The only difference is when they do the reentry burn.
Other times you've claimed that Starship hasn't met its payload goals. But Starship currently delivers a half payload because when launching from Texas, it must perform a dogleg maneuver to avoid flying over populated areas (such as Cuba). When they start launching from Florida, they they'll be able to put 100 tons into orbit per launch.[2]
You've also claimed that Starlink's addressable market is shrinking because terrestrial wireless will expand.[3] But a cell tower costs about 1/3rd as much as a V2 mini and can't serve nearly as many customers in rural areas. So even with Falcon 9, Starlink outcompetes cell towers in areas with low population density. If Starship lowers launch costs more by being reusable, then the tradeoff will favor Starlink in more areas, not fewer. The only way to reduce Starlink's TAM would be if cell towers somehow became much cheaper.
Even today you've claimed that the ship's engines don't have a heat shield.[4] But if you look at images of Starship or watch videos of reentry, you can see the engines are enclosed by the aft skirt. They already need to shield the engines for reentry because they operate during the landing burn.
These sorts of mistakes make it much harder to lend credence to your conclusions. Yes the Starship program has taken much longer than SpaceX planned, but if the program is going to be successful in the long term, outcomes like flight 15 are exactly what one should expect to see.
1. https://news.ycombinator.com/item?id=49893773
2. https://arstechnica.com/space/2026/11/spacex-now-fundamental... "Dontchev said a single Florida Starship launch will be able to carry up to 20 times the bandwidth of a single Falcon 9." SpaceX currently claims that each Starship launch carries 10x the bandwidth of a Falcon 9 launch, so this would put them at around 100 tons of payload.
I'd argue not quite yet. They've shown that they can catch the upper stage and that of course is a big milestone and improves the chances of getting a reusable system tremendously. A demonstration of a reusable upper stage requires them to successfully re-fly a previously used upper stage. The catch demonstration is a very good step in that direction, but not quite yet (in my opinion).
Part of the importance is due to the fact that while the space shuttle was a technical success in terms of being able to build and operate a reusable upper stage, it was hella expensive. It required a lot of expensive refurbishment between missions. Launch cadence had a fairly low upper bound. And it wasn't safe enough.
The true believers will claim that catching a starship of her stage marks a new era. It marks the possibility of a new era. It proves it is plausible that a vertical landing upper stage could possibly be a better approach than a space plane. But it hasn't proven any of that. There are about a dozen major milestones between here and actually having a rocket system that meets all of its claimed specifications, both technical and financial. Here's just one big hairy example: to fly manned missions, starship either needs to be redesigned to accommodate an escape system, or it needs to show a level of reliability and safety never previously achieved by a rocket.
In my estimation, it will take at least another five years to know what starship really can do and what it will cost.