It's funny how people can look at the exact same program and draw wildly different conclusions. From where I'm sitting, Starship is a deeply flawed and highly risky program with an uncertain future. There are two keys problems:
1. It's not clear what the product-market fit for Starship really is, specifically there is limited demand for these super-heavy payloads. What's my evidence? Falcon Heavy. It averages about 1 launch a year, typically by the military, compared to >100 for Falcon 9.
There are five standard counterarguments to this:
a) "You can carry multiple payloads." Most of the time this simply isn't true. The payloads have to be on almost identical orbits for this to work. Different inclination? No shot. Different altitude? Almost no shot. This brings us to the second;
b) "But Starlink." Yes, Starlink can take advantage of this although to justify its massive expansion, Starlink really has to launch a satellite-based handheld mobile phone network and it has to be pretty much global. It's unclear if this can be from a legal and regulatory standpoint or if it will be able to effectively compete with traditional terrestrial 5G networks. It's unclear what customer density can be supported;
c) "It will induce demand for bigger payloads." Falcon Heavy is relatively cheap, particularly compared to the alternatives from Ariane and ULA yet launch cadence is still particularly slow. Satellites, which are the bread and butter of commercial rocketry, just don't need to be that big;
d) "Starship will replace Falcon 9 launches." Yes, but other companies will fill the Falcon 9 gap if SpaceX abandons it. If you're only carrying one payload to LEO (or even geostationary orbit) and you can do that with a smaller launch system that you can reuse, it's simply going to be better eventually; and
e) "Space based AI data centers." This is never going to happen. it doesn't make any sense. You can't get enough power or distribute the heat. You're subject to cosmic rays and solar radiation, which the atmosphere otherwise mostly protects you from, and you have no way to economically service the inevitable failures.
2. The whole Mars thing is BS, basically, kinda like space-based data centers. It's not happening. There's absolutely no reason to go;
3. Starship isn't suited for lunar missions, which is one reason (IMHO) why Elon tried to scuttle NASA's lunar program [1]. Basically, a 40 meter high vessel that has to land horizontally is hard to land, hard to get in and out of (with your astronauts 40 meters off the surface) and much harder to get off the Moon and back to Earth.
4. Between (2) and (3) a whole lot of new capabilities have to be developed anyway, none of which have even begun yet (eg in-orbit refueling).
Now the one really useful thing from the Starship program (IMHO) is the Raptor engine. For context, the Falcon 9's Merlin engine used oxygen and RP-1 (kerosene, basically). Only the oxygen had to be kept really cold. But it has coking issues (ie solid carbon build up from the RP-1 being consumed). Raptor replaces RP-1 with LNG. No coking but now you have two cryogenic fuels instead of one.
I'll float a use case, which I admit though borders on science fiction. Still, it could be done without breaking the laws of physics.
"Construction of a multi-generational ship, in orbit, designed to travel for hundreds of years between stars".
Discussing this in-depth would take a book, but by necessity this ship would be very (very) large and would require an extraordinary amount of mass to be ferried into orbit.
Of course this unlikely to happen during starship's lifetime, but hey, it's at least a legitimate use case.
They don't need super heavy payloads, it seems pretty clear that they'll be able to launch even just a Falcon 9 payload in this for less than Falcon 9, because they're not throwing away the expensive upper stage.
Also, my understanding is that Starlink v3 just doesn't fit inside F9. The larger phased arrays that that enables lets them make much better use of spectrum than the v2 minis. There are lots of potential uses for larger payload volumes.
The space DC thing has been beaten to death online, with tons of very confident commentators claiming it will never happen for the reasons you've mentioned. Some counterpoints, Google just launched their first ML infra to space with their Project Suncatcher launch, and if you actually run the numbers, the radiators are smaller than the solar arrays, so obviously it's feasible to launch the radiator if it's feasible to launch the solar array to power it. Who knows if it'll make economic sense, but it seems pretty clear that it's getting harder to build DCs on the ground in the US, there's a lot of backlash against it, we just saw what it did to Abbott. Many areas don't have the spare power, and doing it in space makes solar into reliable baseload power rather than 1/8th nameplate and intermittent.
Anyway, I don't work in that industry, I've just been following along with them because I've been an investor since before Starlink was a thing, but I really think people should keep an open mind about whether SpaceX can pull off things that sound a bit nutty.
I don't know why people go on about the radiator, it's pretty obvious to me that the biggest challenge is your solar panels and radiators acting as wind sails that increase atmospheric drag. This is a manageable problem at small scales but it limits the maximum node size.
I'm having trouble going from the facts I know to your conclusion ("Starship is deeply flawed"). I supposed "deeply flawed" is subjective, so I doubt I'll change your mind, but I can't help myself when "someone is wrong on the internet."
You say, "It's not clear what product-market fit for Starship really is" but then you list a pretty good market fit: Starlink. Starship is the only vehicle that can launch Starlink v3 satellites, and one single Starship launch can put up as much bandwidth as 10 Falcon 9 launches.
The internal launch cost of a Falcon 9 launch is ~$20 million, so effectively, one Starship launch is replacing $200 million worth of Falcon 9 launches. Current speculation is that it costs $100 million to build a Starship stack, so even if they never recover either booster or ship, they would still be saving $100 million per launch.
That's pretty good product-market fit, if you ask me.
Next you talk about Mars. I assume the argument is basically: Starship was designed to go to Mars, but there's no reason to go to Mars, ergo Starship has no purpose. But that sounds like a self-refuting argument. If Starship was designed to go to Mars, that implies that Musk wants to go to Mars (which I actually think is true), which means Starship has a purpose: to go to Mars.
Alternatively, let's say Musk is just conning and he doesn't really care about Mars. He's really just building Starship for Starlink or Starmind or whatever. In that scenario (which I don't believe but can't disprove) then Starship still has a purpose: Starlink and Starmind.
Your next argument is that Starship isn't suited for lunar mission. I actually agree with this--Starship was designed for Mars, which is why it runs on methane (for in-situ propellant generation), and relies on a aerobraking to enter Mars orbit. But the same architecture that can take Starship to Mars will be able to land 50 tons on the lunar surface--something that no other lander can do.
[Your arguments about landing difficulties and astronauts descending 40 meters are not very strong. Starship is bottom-heavy because of the engines, so very hard to tip over, and elevators are a 100-year-old technology.]
Lastly, you bring up "a whole lot of new capabilities" to be developed, but you only mention in-orbit refueling. That's true: in-orbit refueling is yet to be demonstrated. But that's a new capability that (a) no other vehicle has performed, and (b) would open up the solar system to exploration. That's a case of Starship being more ambitious than any previous space vehicle--almost the opposite of "deeply flawed". And even if in-orbit refueling never works (which I doubt), Starship would still be revolutionary for being reusable.
We both have wildly different conclusions on Starship. Maybe one way to see who's right is to make predictions. I think Starship will continue to launch Starlink satellites and that SpaceX will continue to make a profit from Starlink. My guess is between 10-20 Starship launches in 2027 and 50-70 launches in 2028. I also predict that Starship will land on the moon, with astronauts, before the end of 2030.
If my predictions come true, I think that would make Starship a very successful program.
It isn't necessary to refuel in orbit. Just add fuel by releasing the empty rocket and attaching a fully fueled one. Rendezvous in orbit has been done since the 60s.
So cost is a funny thing. You have to amortize the program cost across the launches. So far the program cost is at least $15 billion [1]. This is a huge multiple (40x+) of the original Falcon 9 program, which was expendable originally and has gone through many upgrades since obviously. But Starship also isn't done. We also don't know how many uses each Starship stack will get and what the refurbishment cost will be. I think it's safe to say there'll need to be continued development just like there has been with Falcon 9.
As for the whole 10x the payload, well that goes back to a big chunk of my comment. The only real demand for this is Starlink. Nobody else needs this kind of capacity in volume. Exactly how many launches are required depends entirely on how large the constellation gets. that could be anywhere from 10,000 to 40,000+. That'll be driven by demand for satellite broadband. I've seen estimates on Starlink-related Starship launches anywhere from 30 to 150 per year.
But how much demand is there really? The receivers are expensive. Wired broadband is superior wherever it's available. Obviously there are some use cases (eg ships, planes) where being mobile is required.
I spent a little time looking into how realistic phone-based Starlink is and... it's not particularly realistic. Phone transmitters are too weak. You'd lose service inside because the phone is too low powered. And more power means a bigger, hotter phone with less battery life. This is a significant disadvantage to terrestrial 5G and is arguably a limit of physics. After all, an iPhone's signal doesn't need to travel 300-600km.
So there's a market for Starlink but is it enough to sustain a $15-30B program cost and 100 Starship launches a year? The jury is out. And if there isn't that demand, it gets a whole lot more expensive because you need to amortize the program cost over fewer launches.
Oh, and new capabilities means redeveloping the capability to carry crew, assuming we have some space station to replace the ISS. The original Crew Dragon was meant to support up to 6-7 crew. There was no demand for that so it got scaled back to 4. Well, what's a 10x payload going to do if you can't even fill 6 seats on Crew Dragon?
As for your launch cadence predictions, I'd argue they're optimistic but not wildly optimistic. I think it's going to take some time to scale up. But the Moon by 2030? No shot. This requires in-orbit refueling, some kind of lunar lander, some kind of crew vehicle and certification for all of the above. If this happens by 2035 I'll be pleasantly surprised.
> You can't get enough power or distribute the heat
At the very least, one can turn off the data center components, wait till it cools down, then start it again. Nobody says all of it has to run continuously.
> you have no way to economically service the inevitable failures
The data center's components can communicate with each other by radio, meaning failed components can be simply replaced.
The primary cost benefit of orbital data centers is the availability of efficient, around-the-clock solar power. Any time spent idling for heat reasons directly eats into the profitability of the concept so there's no reason* to design a system that isn't optimized to operate in a steady state. You would instead just put fewer GPUs on it or increase the size of your panels and radiators. The only question is whether or not that steady state is cost effective.
*There are reasons, but none that justify very large scale compute clusters in space
1. It's not clear what the product-market fit for Starship really is, specifically there is limited demand for these super-heavy payloads. What's my evidence? Falcon Heavy. It averages about 1 launch a year, typically by the military, compared to >100 for Falcon 9.
There are five standard counterarguments to this:
a) "You can carry multiple payloads." Most of the time this simply isn't true. The payloads have to be on almost identical orbits for this to work. Different inclination? No shot. Different altitude? Almost no shot. This brings us to the second;
b) "But Starlink." Yes, Starlink can take advantage of this although to justify its massive expansion, Starlink really has to launch a satellite-based handheld mobile phone network and it has to be pretty much global. It's unclear if this can be from a legal and regulatory standpoint or if it will be able to effectively compete with traditional terrestrial 5G networks. It's unclear what customer density can be supported;
c) "It will induce demand for bigger payloads." Falcon Heavy is relatively cheap, particularly compared to the alternatives from Ariane and ULA yet launch cadence is still particularly slow. Satellites, which are the bread and butter of commercial rocketry, just don't need to be that big;
d) "Starship will replace Falcon 9 launches." Yes, but other companies will fill the Falcon 9 gap if SpaceX abandons it. If you're only carrying one payload to LEO (or even geostationary orbit) and you can do that with a smaller launch system that you can reuse, it's simply going to be better eventually; and
e) "Space based AI data centers." This is never going to happen. it doesn't make any sense. You can't get enough power or distribute the heat. You're subject to cosmic rays and solar radiation, which the atmosphere otherwise mostly protects you from, and you have no way to economically service the inevitable failures.
2. The whole Mars thing is BS, basically, kinda like space-based data centers. It's not happening. There's absolutely no reason to go;
3. Starship isn't suited for lunar missions, which is one reason (IMHO) why Elon tried to scuttle NASA's lunar program [1]. Basically, a 40 meter high vessel that has to land horizontally is hard to land, hard to get in and out of (with your astronauts 40 meters off the surface) and much harder to get off the Moon and back to Earth.
4. Between (2) and (3) a whole lot of new capabilities have to be developed anyway, none of which have even begun yet (eg in-orbit refueling).
Now the one really useful thing from the Starship program (IMHO) is the Raptor engine. For context, the Falcon 9's Merlin engine used oxygen and RP-1 (kerosene, basically). Only the oxygen had to be kept really cold. But it has coking issues (ie solid carbon build up from the RP-1 being consumed). Raptor replaces RP-1 with LNG. No coking but now you have two cryogenic fuels instead of one.
[1]: https://futurism.com/elon-musk-scrap-nasa-moon-program
"Construction of a multi-generational ship, in orbit, designed to travel for hundreds of years between stars".
Discussing this in-depth would take a book, but by necessity this ship would be very (very) large and would require an extraordinary amount of mass to be ferried into orbit.
Of course this unlikely to happen during starship's lifetime, but hey, it's at least a legitimate use case.
Also, my understanding is that Starlink v3 just doesn't fit inside F9. The larger phased arrays that that enables lets them make much better use of spectrum than the v2 minis. There are lots of potential uses for larger payload volumes.
The space DC thing has been beaten to death online, with tons of very confident commentators claiming it will never happen for the reasons you've mentioned. Some counterpoints, Google just launched their first ML infra to space with their Project Suncatcher launch, and if you actually run the numbers, the radiators are smaller than the solar arrays, so obviously it's feasible to launch the radiator if it's feasible to launch the solar array to power it. Who knows if it'll make economic sense, but it seems pretty clear that it's getting harder to build DCs on the ground in the US, there's a lot of backlash against it, we just saw what it did to Abbott. Many areas don't have the spare power, and doing it in space makes solar into reliable baseload power rather than 1/8th nameplate and intermittent.
Anyway, I don't work in that industry, I've just been following along with them because I've been an investor since before Starlink was a thing, but I really think people should keep an open mind about whether SpaceX can pull off things that sound a bit nutty.
You say, "It's not clear what product-market fit for Starship really is" but then you list a pretty good market fit: Starlink. Starship is the only vehicle that can launch Starlink v3 satellites, and one single Starship launch can put up as much bandwidth as 10 Falcon 9 launches.
The internal launch cost of a Falcon 9 launch is ~$20 million, so effectively, one Starship launch is replacing $200 million worth of Falcon 9 launches. Current speculation is that it costs $100 million to build a Starship stack, so even if they never recover either booster or ship, they would still be saving $100 million per launch.
That's pretty good product-market fit, if you ask me.
Next you talk about Mars. I assume the argument is basically: Starship was designed to go to Mars, but there's no reason to go to Mars, ergo Starship has no purpose. But that sounds like a self-refuting argument. If Starship was designed to go to Mars, that implies that Musk wants to go to Mars (which I actually think is true), which means Starship has a purpose: to go to Mars.
Alternatively, let's say Musk is just conning and he doesn't really care about Mars. He's really just building Starship for Starlink or Starmind or whatever. In that scenario (which I don't believe but can't disprove) then Starship still has a purpose: Starlink and Starmind.
Your next argument is that Starship isn't suited for lunar mission. I actually agree with this--Starship was designed for Mars, which is why it runs on methane (for in-situ propellant generation), and relies on a aerobraking to enter Mars orbit. But the same architecture that can take Starship to Mars will be able to land 50 tons on the lunar surface--something that no other lander can do.
[Your arguments about landing difficulties and astronauts descending 40 meters are not very strong. Starship is bottom-heavy because of the engines, so very hard to tip over, and elevators are a 100-year-old technology.]
Lastly, you bring up "a whole lot of new capabilities" to be developed, but you only mention in-orbit refueling. That's true: in-orbit refueling is yet to be demonstrated. But that's a new capability that (a) no other vehicle has performed, and (b) would open up the solar system to exploration. That's a case of Starship being more ambitious than any previous space vehicle--almost the opposite of "deeply flawed". And even if in-orbit refueling never works (which I doubt), Starship would still be revolutionary for being reusable.
We both have wildly different conclusions on Starship. Maybe one way to see who's right is to make predictions. I think Starship will continue to launch Starlink satellites and that SpaceX will continue to make a profit from Starlink. My guess is between 10-20 Starship launches in 2027 and 50-70 launches in 2028. I also predict that Starship will land on the moon, with astronauts, before the end of 2030.
If my predictions come true, I think that would make Starship a very successful program.
As for the whole 10x the payload, well that goes back to a big chunk of my comment. The only real demand for this is Starlink. Nobody else needs this kind of capacity in volume. Exactly how many launches are required depends entirely on how large the constellation gets. that could be anywhere from 10,000 to 40,000+. That'll be driven by demand for satellite broadband. I've seen estimates on Starlink-related Starship launches anywhere from 30 to 150 per year.
But how much demand is there really? The receivers are expensive. Wired broadband is superior wherever it's available. Obviously there are some use cases (eg ships, planes) where being mobile is required.
I spent a little time looking into how realistic phone-based Starlink is and... it's not particularly realistic. Phone transmitters are too weak. You'd lose service inside because the phone is too low powered. And more power means a bigger, hotter phone with less battery life. This is a significant disadvantage to terrestrial 5G and is arguably a limit of physics. After all, an iPhone's signal doesn't need to travel 300-600km.
So there's a market for Starlink but is it enough to sustain a $15-30B program cost and 100 Starship launches a year? The jury is out. And if there isn't that demand, it gets a whole lot more expensive because you need to amortize the program cost over fewer launches.
Oh, and new capabilities means redeveloping the capability to carry crew, assuming we have some space station to replace the ISS. The original Crew Dragon was meant to support up to 6-7 crew. There was no demand for that so it got scaled back to 4. Well, what's a 10x payload going to do if you can't even fill 6 seats on Crew Dragon?
As for your launch cadence predictions, I'd argue they're optimistic but not wildly optimistic. I think it's going to take some time to scale up. But the Moon by 2030? No shot. This requires in-orbit refueling, some kind of lunar lander, some kind of crew vehicle and certification for all of the above. If this happens by 2035 I'll be pleasantly surprised.
[1]: https://kfgo.com/2026/05/01/exclusive-spacex-spending-on-sta...
At the very least, one can turn off the data center components, wait till it cools down, then start it again. Nobody says all of it has to run continuously.
> you have no way to economically service the inevitable failures
The data center's components can communicate with each other by radio, meaning failed components can be simply replaced.
*There are reasons, but none that justify very large scale compute clusters in space