If the upcoming Starship launch successfully demonstrates a deployment of working Starlink satellites, that will doubtless give the company a boost.
Eh.
The big booster works to orbit. The satellite dispensers work fine on their other rockets. And have worked fine in the few dummy tests on the big booster to date. It’s optimism about everything else about Starship that hasn’t yet been demonstrated that’s supporting the stock price.
Until then it’s smoke, mirrors, and the expectation that somehow Musk has permanently attached himself & SpaceX to the federal teat in a way that ensures no matter what happens technologically or politically, he, and therefore his shareholders, will win and the taxpayers will lose.
That’s all good thinking. Except for the assumption that SpaceX shareholders will share in the largesse just like Musk will. There’s lots of ways for that to not quite happen. Like the folks who voted for you know who in the expectation that all his theft would somehow accrue to that voter’s benefit.
Even if orbital data centers prove a chimera, demonstrating that the launch capacity to orbit them is there would allow the AI bubble to continue that much longer, and there are always ways to cash in in the interim.
What does a launch via Starship demonstrate that the dozen of launches they’ve already accomplished didn’t?
In response to criticism about his space data center pipe dream, Elmo said:
We start flying them next year.
Right! Elmo will have a fully operational hyper-scale data center orbiting the earth next year, consisting of 2,500,000 high-end AI chips and consuming 5 GW of power! Also by next summer, he will have fully colonized Mars, with a million-plus population, all driving Cybertrucks. He will also have further increased his ketamine dosage to the point he might actually believe it’s all true.
The potential for Starship to achieve a vast economy of scale on orbital launch. Basically put ALL the Starlink satellites that will share an orbital plane up in one go, as well as any other applications of Starship’s heavy tonnage capacity.
It isn’t whether Musk believes it’s true, it’s whether the investors believe it’s true.
I don’t think it’s legal to lie to investors. Being delusional is not an excuse.
I don’t think that demonstrates much.
The size of the launch pad explosion?
“Optimism is not securities fraud.” – Elon Musk, probably
Here’s an engineering assessment of data centers in space. It’s pitched for educated lay persons (not too many equations).
Why Thermodynamics Rules Future Orbital Data Centers - IEEE Spectrum
However, the winners in this sector will be determined by the systems architects who most cleverly accommodate the thermodynamics and the companies with sufficient vertical integration to take on the massive costs of operating data centers in orbit. Ultimately, the physics tax is universal. Whether managing heat rejection in the vacuum of low Earth orbit or managing power density in a hyperscale facility in Northern Virginia, the constraint is never the silicon. It’s the thermodynamics.
Even if I believed he’ll be launching data centers into orbit next year, I still wouldn’t understand why anyone would want him to. Orbital data centers barely make more sense than orbital Tesla plants.
Hey, he already did the prototype orbital Tesla. 
“I got a Tesla today.”
“I didn’t know you were interested in getting a Tesla.”
“I wasn’t. It crashed through the roof of my living room while I was having lunch.”
If you take some wacky Libertarian belief that orbit is beyond the reach of Earthly governments it all makes sense.
Or if you take ketamine; lots and lots of ketamine.
The article mostly talks about the key problems with orbital data centers, mainly collecting enough energy, rapid chip degradation by ionizing radiation, and of course the all-important cooling problem. But it also makes an attempt to justify computing power in space, and both reasons seem to me to be pretty lame.
One reason given is that rapid space-borne calculations to avoid satellite collisions in an increasingly crowded LEO space are much faster and more effective than terrestrially-based ones. This seems counterintuitive since the satellite clusters that are proposed for these data centers would themselves be a major contributor to the problem; Musk alone has proposed sending up as many as a million such satellites, but even allowing for his usual hyperbole, even thousands of new satellites is a lot.
The other rationalization is that the development of high-bandwidth links is beneficial for earth observation satellites like those used for things like intelligence gathering or tracking shadow fleets of ships carrying contraband. This seems only very distantly related to the vast overall effort going into building these silly orbital data centers.
Agree. Both those justifications are purely drugs talking.
You’d think that cooling should be easy in space, because space is very cold.
The problem is that cooling usually involves either conduction or convection. Conduction involves some kind of medium, like a heat sink, that draws heat away and disperses it onto something cooler. But where? If you are in a vacuum, like space, that heat doesn’t have anything to leech onto.
Convection involves cycling in something fluid, like a gas or liquid, and replacing the hot fluid with cool fluid. An air conditioner replacing hot air with cool air is a form of convection. But again, you’re in a vacuum.
The only other cooling method available is thermal radiation, and that does work. You have to disperse the heat into a radiator of sorts, using pipes, or fins, or tubes of liquid, something that can then radiate the heat into space as electromagnetic radiation. Because that can travel through the vacuum of space; that’s how we get heat and light from stars, including our sun. It works, but it’s less effective than convection or conduction.
And that sucks, because one of the biggest challenges for datacenters is the need for cooling. And you’re putting them one of the most difficult environments for cooling. You are basically picking “hard mode”. I see the appeal of it, in a way… There is a real NIMBY (not in my backyard) attitude toward datacenters, so it seems like it makes sense to put them in an out-of-the-way area so that people won’t complain. Just like how they pick remote areas far from any people to dump hazardous waste. Except that, again, it’s a very challenging place to try to perform this kind of operation.
I don’t think so.
The author is pointing out the impracticality of data centers in space using current technology while simultaneously pointing out that there are other space based compute applications that make development of technology to reduce cost and increase lifetime of compute in space worthwhile.
The article discusses applications that would justify more computing in space than is currently available, not data centers in space. At no time does the author suggest that this computing requires data centers. He does mention the need for HPC near satellites, but I’ve seen HPC installations for special purposes with less than 16 processors.
However, there are niche applications where the much higher costs of computing in space could be justified.
While training or inference on LLMs in space doesn’t seem economical today, there are other, very compelling applications for computing in space.
The cost of even the equivalent of a laptop processor is a lot higher if it is going to be space based, and its reliability is going to be extremely limited. However, both of the applications he raises are probably worth efforts to make improvements. Getting data to/from the ground as the amount of data rises exponentially with bigger, higher resolution, wider bandwidth sensors in orbit forces decisions on what useful data to throw away and if the latency of processing on the ground and returning the information to the satellite becomes on the order of the time between potential collisions, this raises the probability of collision, with subsequent cascading damage.