The only reason for xAI to join SpaceX is to offload Elon's Twitter debt in the upcoming IPO.
Twitter (X) owed $1.3B in debt every year in interest since Musk's takeover. This was before re-financing in a higher interest rate environment. The company was losing $200MM+ per year on ~$5B in revenue before the takeover, and there are reports that revenues have decreased by round 50%.
Best case scenario if we accept those numbers is that X makes $3B per year and about half of that goes immediately out the door in debt payments before paying a cent for the entire business to function.
However, if SpaceX acquires X, that ~$1.5B in interest is a fraction of the $8B In profits SpaceX is allegedly generating annually. Further, they can restructure the debt if it's SpaceX's debt, and not owned by X. Investors will be more likely to accept SpaceX shares as collateral than X.
X made a profit last year because they cut costs lower than the drop in ad revenue (which is also slowly recovering). The big question is if they will still be profitable in 2026 year without the US election driving big traffic numbers and ads.
You also have to be careful about who said it and what they meant by "profit," because there is gross profit, EBIT, EBITDA, and others.
> Despite a revenue drop from $5 billion in 2021 to roughly $2.7 billion in 2024, the EBITDA margin surged from 13.6% to 46.3% due to drastic cost-cutting measures and restructuring
The staggering amount of money Elon Musk raised for doing AI stuff is quite a bit more than what he ever expended on the Twitter value implosion. I think we can agree that there isn't much left of that. Also, whatever debt was issued for that was issued in dollars. We've had a few years of inflation and dollar devaluation recently. I don't think whatever Twitter debt there was is much of big headache for X at this point.
X.ai is controversial mainly because of Musk. But if you can look beyond that, it does actually have a bit of non trivial IP. Grok is not bad as a LLM. It's not necessarily best in class but it's close enough to be useful. Apple needs to license their AI from Google and OpenAI. MS outsources to OpenAI. Amazon doesn't really have their own models at all. So, as trillion dollar companies go, having your own in house developed model training pipeline that actually works isn't all that common yet.
Musk for all his failings has a talent for looking beyond the current day to day navel gazing that characterizes VC short term thinking and much of the activity in silicon valley. He clearly looks at space as a bit of underused real estate.
Star Link is one of those mad plans that actually seems to make sense now that he has proven that launching thousands of satellites into space isn't that big of a deal and can actually be profitable if you get a few million people to spend billions per month on reliable data connections.
AI data centers in space are similarly ludicrous unless you have a newly developed 100+ ton to orbit reusable launch capability at your disposal. Also, the nature of doing stuff in space is that it is a very people hostile environment. So having some in house AI capability isn't the worst idea for a space company with ambition, which like it or not SpaceX clearly has. I wouldn't call X.ai a bar gain. But what's the alternative if you are semi serious about controlling an armada of space craft across the solar system?
Total investment in xAI are around $30B-40B (including the latest E round). Twitter purchase price at $44B was more than that. Out of that 44B, ~$25B was debt financing.
> Star Link is one of those mad plans that actually seems to make sense now [...] AI data centers in space are similarly ludicrous unless you have a newly developed 100+ ton to orbit reusable launch capability at your disposal
I don't think these two are comparable. Starlink obviously makes sense if you can put thousands of satellites in LEO cheaply, which (only) SpaceX could. The challenge there was to actually build and put them there.
For data centers, even if you can launch for free, the physics and economics don't make sense. Solar is free but the amount of solar arrays (and cooling radiators) required means it's just easier and cheaper to build out the same thing on Earth, and that's without thinking about maintenance of either the data center or the required support equipment.
In theory it can be done. In practice, I humbly propose that putting the same engineering brains on solving the hard questions of keeping people alive in space (so they can, eg, get to Mars and back) would align more closely with the SpaceX mission.
> But what's the alternative if you are semi serious about controlling an armada of space craft across the solar system?
"X Combinator" for space tech (life support, stations, habitats, etc - everything that SpaceX itself isn't focusing on). Refueling depots at strategic locations that are good launching points for deep space (Mars+) missions.
Not a friggin' LLM.
I would argue that they have earned their own controversy independent of Musk with all the shenanigans they pulled building out their data centers, namely their illegal use of gas turbines to power the whole thing.
I'm actually not that upset about AI data center energy usage. I see this as a short term and costly scaling measure with a minor impact (considering overall wasteful energy practices) that is an obvious target for large and rather obvious cost reductions the second this market gets profitable. The only reason that isn't happening from day 1 is all the red tape currently being put in place to actively slow down the demise of fossil fuel based generation.
Cost reductions here mean switching to a cleaner form of energy for the reason that that can be a lot cheaper than burning expensive gas in an expensive generator. Any large scale user of energy is going to be optimizing their energy opex if it saves them lot of money. If they survive long enough to matter, of course. If you are using energy by the tens/hundreds of gwh per year that is not going to be small amounts.
The point of Star link was orders of magnitude reduction in cost of launching thousands of satellites. Musk is talking indirectly about another order magnitude of further reduction of that via star ship; sorry if that wasn't clear.
> the physics and economics don't make sense
This is a popular assertion that despite all the experts chiming in is not that black and white. Clearly investors and Elon Musk beg to differ. Similar arguments were used against Star Link when that was still science fiction. And now it isn't. It actually seems like a good idea that at this point is being copied by others. And SpaceX is getting a lot of the launching business, for now.
I think it's mainly the economics that are the challenge here; not the physics. Implicit in the assertion is that launching the amount of mass needed would be prohibitively expensive. There are lots of engineering challenges as well.
> it's just easier and cheaper to build out the same thing on Earth
Maybe; but it seems challenging to scale there. Permitting and scaling energy generation are a problem right now. But I agree, it's more logical to fix that. But one does not exclude the other. We might end up with a lot of in orbit computation regardless. It's not an either or proposition.
the problem is that this market becomes commoditized, there are tons of not bad open weight LLMs available, and it is not clear if Grok IP is that not trivial. They even totally can run others LLMs under the hood, it is well known that they used Claude output at the beginning.
Of course it's the economics, not the physics, that are the challenge. We have thousands of existence proofs that you can launch a computer into orbit and have it work. The question is not, can you do it. The question is, why would you do it instead of putting that same computing capacity in a building somewhere?
It's not an either/or proposition in general but it is at the small scale. If I'm going to spend a million dollars on computing equipment, I can spend it on a terrestrial installation, or a space installation, or put some if it towards each, but anything I put towards one takes away from the other. If the economics of a terrestrial installation are much better than a space installation, why would I allocate any of my money to space?
Most of the conversation here seems to boil down to:
"Putting a data center in space is very hard and makes no sense."
"Putting a data center in space is actually pretty easy."
But by "hard" we mean "difficult to the point of being extremely economically uncompetitive with putting computers in a building," and by "easy" they mean "technologically feasible and the basic concept of computers in space has been done many times already."