Elon: To Infinity and Beyond! Again!
Earlier this week SpaceX acquired xAI. Elon’s explanation immediately triggered memories, and set off alarm bells.
“SpaceX has acquired xAI to form the most ambitious, vertically-integrated innovation engine on (and off) Earth,” Musk, the top executive at both companies, said in the memo.
Just like his justification for the Tesla acquisition of SolarCity, which I excoriated at the time. I wrote that the vaunted “vertical integration” benefits then were chimerical, and indeed they proved to be exactly that. (Tesla’s subsequent performance does not gainsay that).
Elon used vertical integration as a distraction from the real motivation for the tie-up, namely, to bail out a related company (managed by relatives, to boot), the failure of which would have punctured Elon’s image as a colossus and genius.
As my X post (ironic, eh?) indicates, I had the same suspicions about the SpaceX absorption of xAI. I am not alone. Most of the subsequent commentary on it recalls the Tesla-SolarCity deal, and questions Elon’s vertical integration rationale. Examples here, here, and here.
The vertical integration story here strains credulity far farther than even the Tesla-SolarCity story. You see, energy is the huge constraint on AI and the data centers that it relies on. Terrestrial sources of power will be inadequate. Hence, space based generation and data centers will be necessary. Hence, synergy!
Elon also said the tie up will facilitate SpaceX’s true purpose–interplanetary travel. More synergy!
“The capabilities we unlock by making space-based data centers a reality will fund and enable self-growing bases on the Moon, an entire civilization on Mars and ultimately expansion to the Universe,”
To infinity and beyond!
Someday. Maybe. Here and now and the foreseeable future? Not so much. Think of all the technical obstacles that must be surmounted, and the time it will take to do so. This is a vision that will take decades to realize. At least.
NASA’s Artemis project, troubled and oft-delayed, envisions 9 Moon missions over the next decade, only the last of which involves astronauts remaining there for an extended period of time. Just think how long after that it will take to put data centers there. To call this pie in the sky is almost literal and a serious understatement.
This is all typical Elon razzle-dazzle to distract attention from fundamental business realities. X qua X is a dog of a business: the main benefit xAI gets out of that piece of the business is as a data source to train Grok.
AI is a massive capital suck. Moreover, the AI business is extremely competitive right now, with multiple players plunging huge amounts of capital into it. Since (like search) it is likely to be winner take all, or at least a few take all, when it matures, xAI’s success is by no means assured. It is therefore highly uncertain whether xAI will be able to obtain the necessary capital as a standalone entity.
By merging xAI with what is effectively a monopoly business, SpaceX, xAI can get capital from its dominant owner and reduce its need to tap the capital markets.
One could possibly tell an “internal capital market” story. External capital is costly. It is likely to be especially costly for xAI given the substantial competition and potential for asymmetric information (the company has a much better understanding of its prospects than outsiders). Given its robust near monopoly business, SpaceX’s capital market frictions are less, and it will potentially throw off free cash flow that it can invest in xAI. Thus, it may be efficient for SpaceX, rather than external capital markets, to fund xAI.
Maybe. But I’m skeptical. Methinks this is more likely a reprise of SolarCity. Use a healthy company to prop up a shakier one, as I posited in my X post above.
In terms of synergy, a Tesla-xAI tie-up would be far less fantastical. Self-driving and robotics (the new focus of Tesla) have a much closer relationship to AI than data centers in space do. Especially in the here and now.
But–and it is a big but–Tesla is a public company, meaning that there would be many more obstacles, legal and otherwise, to a merger with xAI. In contrast, Elon could combine SpaceX and xAI almost by fiat.
In sum, discount the ElonHype. In particular, whenever Elon says “vertical integration” run, don’t walk, to the nearest exit. The phrase a smokescreen intended to give a seemingly plausible business rationale for combinations that in fact have a very different motivation: keeping Elon’s empire intact.
Note: The phrase “To infinity and beyond!” was originally and inadvertently included in the quote box for Elon’s Moon, Mars, Universe statement. Although it was believable in that context, I have moved it outside the box.
Solar City technology will very likely be efficient and cost-effective in space or on the Moon. Maybe Elon is betting on the come.
Comment by Pat Frank — February 6, 2026 @ 6:11 pm
I’m trying not to be any more cynical than I absolutely have to, and I’ll approach this semi-rationally. A data center in space will need thumb suck twenty-five square kilometres of solar panels. If those panels are the same kind you put on your roof the mass is thirty tons a square kilometre. Add the mass of the GPU hardware and round it off to a thousand tons. At $100 per kilogram to low earth orbit that’s $100 million. Add a zero for good luck and the transport cost is a billion dollars. But once it’s up there, the electricity is free. So it’s not as crazy as it first seems.
The next problem is earth shadow. This can be overcome two ways. One is batteries. One is batteries with polar orbit, and the space data center will only experience earth shadow for half of every year, for half of every orbit during that half-year. There’s a mass to orbit penalty, but it’s not as high as you’d think. My thumb says half a billion dollars.
Remember that Meta is paying two billion or a bit less, same ballpark, to restart TMI.
I don’t know the revenue, but back of an envelope there’s a business case for space data centers.
Comment by Michael van der Riet — February 7, 2026 @ 12:57 am
He’s quite the card, Mr Musk.
Still, if you must have gazillionaire tax-farmers then it’s not a bad idea that they provide as much entertainment as Barnum and Bailey.
Comment by dearieme — February 7, 2026 @ 6:52 am
Hmm, no mention of the astronomical (geddit?!) valuation of the combined enterprise? You can have some fun if you ask AI (geddit?!) about this – Gemini described it as “satirical”. The singularity is already here.
@Pat: Wait what?? You like solar panels, but just not here on Earth?
@Riet: What about transmission costs to/from this floating behemoth? Gonna be bigly.
Comment by David Mercer — February 8, 2026 @ 2:25 am
@4 Unbiased freethought, David.
You might try it.
Geosynchronous orbit solves MvdR’s earth shadow problem. So would orbits at Lagrange Points L1 & L2.
Zero gravity structures are far less massive than their surface homologues.
I’ll do no calculation, but unfiltered solar radiation at Earth is 1.36 kW/m² with the full suite of ultraviolet. And no dust, no clouds. They’d be far more efficient than their ground-based counterparts. They’d produce power 24/7.
Comment by Pat Frank — February 9, 2026 @ 8:32 am
I appreciate the ‘unbiased freethought,’ David, but the engineering here is missing quite a few reality checks.
“Geosynchronous orbit solves MvdR’s earth shadow problem.”
It solves the shadow but creates a distance problem. At 35k km distance , physics (the Inverse Square Law) dictates that your power beam will spread out massively. To catch it, you’d need a ground receiver the size of a small county. Orbits at L1/L2 aren’t magical sunlight zones either—halo orbits require continuous station-keeping, and illumination isn’t perfectly uninterrupted.
“Zero gravity structures are far less massive…”
You save on steel, but you pay in thermal management. Panels in a vacuum can’t shed heat via air; they’ll bake and lose efficiency unless you add massive, heavy radiators. Also, sunlight exerts physical pressure—without heavy thrusters and fuel, your “massless” farm literally sails away.
“Unfiltered solar radiation… 1.36 kW/m²… far more efficient.”
This ignores conversion losses. Converting DC to microwaves, beaming them through the ionosphere, and converting them back on Earth loses 50–70% of the energy. A “less efficient” panel in Nevada is still better than one that loses half its power on the commute.
“cost-effective in space or on the Moon.”
The moon has a 14-day night. Unless you’ve invented a battery the size of a crater, you’re in the dark for half the month. Even if you hide at the poles for ‘eternal light,’ you’re stuck building vertical towers on jagged terrain to catch a sun that never rises more than 1.5 degrees above the horizon. Plus, lunar dust is basically tiny, magnetic glass shards that shred panel coatings and stay lofted forever. There are no squeegees on the Moon, David.
In short=> sunlight isn’t the limiting factor—distance, diffraction, thermal management, conversion losses, lunar nights, and dust are the real blockers.
Comment by [email protected] — February 9, 2026 @ 12:43 pm
Manufacturing is in space.
Metal-Mylar coating is very light and highly heat-reflective.
Nowhere is beaming power back to Earth mentioned.
Regions of both lunar poles receive nearly perpetual insolation.
Comment by Pat Frank — February 9, 2026 @ 1:56 pm
My dear Patrick — your ideas seem to be evolving faster than the engineering constraints I am highlighting are being addressed, so let’s unpack the new additions.
“Manufacturing is in space.”
That moves the supply chain; it doesn’t repeal physics. High-efficiency photovoltaics require extreme material purity, precision fabrication, structural stiffness, attitude control, thermal management, and micrometeoroid protection regardless of where they’re manufactured. Relocating the factory to orbit is an impressive ambition, but it doesn’t make the downstream engineering problems politely disappear.
“Metal-Mylar coating is very light and highly heat-reflective.”
Reflective surfaces reduce absorbed energy; they do not provide heat rejection. Solar arrays still convert a significant fraction of incident power into waste heat, and in vacuum that heat must be radiated away through dedicated radiator area and thermal control systems. Spacecraft engineers have been trying to make “lightweight reflective film replaces radiators” work for decades — they still carry radiators.
“Regions of both lunar poles receive nearly perpetual insolation.”
“Nearly” is doing a remarkable amount of work there. Polar illumination is intermittent, occurs at very low solar elevation angles, and introduces long shadowing, complex siting, distributed infrastructure, and ongoing storage or transmission requirements. Extended sunlight is helpful; it is not the same thing as continuous power.
Comment by [email protected] — February 9, 2026 @ 2:46 pm
You appear unfamiliar with the lightweight photovoltaics already in use, in space.
Heat is the transfer of energy from a warm body to a cold. A material that reflects energy does not become hot.
Lunar obliquity is about 1.5°. Grazing polar sunlight is virtually full intensity.
Comment by Pat Frank — February 9, 2026 @ 3:42 pm
Patrick, your confidence in basic thermodynamics is refreshing, but it’s unfortunately unburdened by the Second Law. Let’s tighten up the physics:
1. Reflective materials: You’ve confused a mirror with a refrigerator. A material that reflects incident sunlight prevents external heating, but it does absolutely nothing to shed the internal waste heat generated by the photovoltaic process itself. Even the most advanced ‘lightweight’ space-grade cells operate at roughly 30–35% efficiency. The remaining 65%+ of that intercepted energy is converted directly into heat within the substrate. In a vacuum, that heat has nowhere to go but through a radiator. If your plan is to rely solely on reflection, your ‘lightweight’ array will effectively become a very expensive, one-time-use space heater before the electronics melt.
2. Lights at the poles: Yes, the lunar obliquity is ~1.5°. That is precisely why your ‘full intensity’ grazing light is an engineering liability, not an asset. To capture that flux, your arrays must be oriented vertically. This creates a cosine loss nightmare or requires massive, tower-mounted tracking systems. More importantly, at a 1.5° angle, a single pebble can cast a shadow that spans an entire crater. Unless your plan involves building towers multiple meters high to stay above the ‘grazing’ shadows of the lunar topography, your ‘perpetual’ power is intermittent at best.
3. Manufacturing: I am quite familiar with thin-film and IMM (Inverted Metamorphic Multi-junction) cells. They are marvels of engineering. However, ‘lightweight’ refers to the active layers, not the integrated system. You still have to solve for structural rigidity, power conditioning, and high-voltage distribution. A sheet of Mylar floating in a vacuum isn’t a power plant; it’s a sail.
If you’ve discovered a way to bypass Kirchhoff’s Law of thermal radiation or found a way to make shadows transparent, the Nobel committee is waiting. Until then, the radiators stay 🙂
Comment by [email protected] — February 9, 2026 @ 4:00 pm
https://www.polytechnique-insights.com/en/columns/space/ultra-lightweight-high-performance-foldable-advances-in-space-photovoltaics/
In other news, never mind, libte.
Comment by Pat Frank — February 9, 2026 @ 7:41 pm
Pat — thanks for the link. Lightweight photovoltaic materials are indeed advancing, and thin-film/IMM technologies are impressive developments.
That said, none of those advances remove the underlying systems constraints we’ve been discussing:
• Reflective coatings reduce absorbed flux but do not eliminate internal waste heat. Radiative cooling remains mandatory in vacuum regardless of panel efficiency.
• Lightweight active layers do not remove the need for structural stiffness, pointing control, power conditioning, and distribution infrastructure.
• Extended polar illumination on the Moon reduces darkness intervals but does not eliminate shadowing, low solar elevation challenges, or the need for storage and system-level redundancy.
In other words, improvements in cell mass or folding geometry are valuable — but they address only one component of a much larger engineering system.
Sunlight has never been the limiting variable. Closing the full thermal, structural, and operational design loop remains the harder part.
Comment by [email protected] — February 10, 2026 @ 5:37 am
@7: Who said anything about beaming power back to Earth?? I’m talking comms – to and from these behemoths to (presumably) millions of users here on Earth, guessing via ground stations dotted around the globe to ensure continuity of coverage. Unless of course Elon envisages us all having satphones and, if so, is he in talks with Apple et al about this? I suspect any Chinese manufacturers will play ball just for LOLz to see the whole enterprise eat itself, while their compatriots churn out endless cheap-as-chips competitor terrestrial AI systems.
Please do keep us apprised of your upcoming investment in SpaceX/xAI. From your fulsome praise of the proposed enterprise I hope you are going all-in on this, no half-measures and all that.
Comment by David Mercer — February 10, 2026 @ 10:47 am
On and in other Musk/space-related news, I’m reading Elon has just had a reality check on his whole “let’s plant the Stars and Stripes on Mars” US taxpayer financed/DOGE exempt junket. Apparently NASA has finally persuaded him to try something more err realistic/affordable/rational. Oh, and some serious space scientists are beginning to fret about something called the Kessler Syndrome (an actual thing), which, if it is an actual thing, may or may not have a material impact on any upcoming space/vanity-based mega-ventures.
Comment by David Mercer — February 10, 2026 @ 11:05 am
@13, libte introduced beamed power in #6.
My interest was not whatever sort of transmission you had in mind(?), which you left ambiguous in any case and directed at MvdR.
I admire your ability to smirk while the UK crashes down around your ears. But then, its your sort doing the crashing, innit.
Comment by Pat Frank — February 10, 2026 @ 2:42 pm
Wake me when someone – anyone! – can send a mammal of any size through and beyond the Van Allen Belts without its radiation-fried body turning luminescent green.
Comment by Ex-Global Super-Regulator on Lunch Break — February 10, 2026 @ 10:08 pm
@16 – didn’t Apollo astronauts travel through the Van Allen belts, to the Moon and back? 11 times?
Comment by Pat Frank — February 10, 2026 @ 11:12 pm
@15 So are you going to invest, or are you all talk*?
Funnily enough a commentator made a comparison between this and another mega merger of yore, between Vodafone and Mannesmann, which resulted in huuuge destruction of shareholder value. But hey, it was a different era when valuations actually meant something, and involved proper European (grr!) companies who actually did something of worth.
*Or broke, having spaffed all of your surplus hard-earned on those futile self-publishing efforts?
Comment by David Mercer — February 11, 2026 @ 7:32 am
@18 – irrelevant blather as your ilk betrays the UK David.
All peer-reviewed. All analytically sound. All overturning one or another progressive pseudo-scientific apple cart.
Comment by Pat Frank — February 11, 2026 @ 9:57 am
@Pat when was the last time anyone went through the belts?
Comment by Ex-Global Super-Regulator on Lunch Break — February 11, 2026 @ 5:00 pm
@20 Ex-G – Apollo 17, mid-December 1972.
https://www.nasa.gov/mission/apollo-17/
Comment by Pat Frank — February 11, 2026 @ 11:25 pm
Hey! It’s the annual SWP post where I can actually share some expertise!
Much as I hate to find myself agreeing with Pat Frank 😛 He’s absolutely right that there are full-sun earth orbits available. It’s one potential reason all these companies are jumping on the bandwagon – there’s a large but not infinite number of such orbits, so this might be something of a gold rush.
I recommend the recent youtube video about this by the very-sensible Scott Manley. He points to an online tool that anybody can use to guess at the viability of space compute, with sliders to adjust launch costs per kg, etc. depending on what you think will happen to the industry in the near future. If we expect Starship to be a success (and much as I dislike Musk, his company shows every sign that they will achieve it), Manley’s personal guess is that in-orbit compute would cost about 3 times what it does on earth. Whether that’s a dealbreaker or not, I leave to the more business-minded commentators (but as the Prof points out, AI is currently a very competitive market, so that markup seems unlikely to be viable for now).
For those worried about radiation or cooling, Manley quite sensibly points out that SpaceX already has satellites with 28kW of solar and high data throughput, so it’s not impossible that this could be scaled up with ~100 satellites flying in constellation. SpaceX already has a unique laser satellite-to-satellite comms system, 100 satellites is a small production run by their standards, and they should soon have an extremely cheap launcher… I dislike Musk greatly, and he does indeed get up to some obvious dead-end ideas, but I would NOT bet against him on this one…
Finally, Manly points out that even if this is all ridiculous, let’s let the billionaires have at it, because by doing this, they’re developing technologies that would be very useful to future space exploration!
Comment by HibernoFrog — February 12, 2026 @ 4:59 am
@Pat So, over 53 years ago.
It hasn’t happened in my lifetime, would you believe. And I’m no spring chicken (anymore). It’s just ‘Boomer Mythology’ to me.
I’ll leave it to you to draw an obvious conclusion.
Comment by Ex-Global Super-Regulator on Lunch Break — February 12, 2026 @ 3:30 pm
@23 Ex-G – The obvious conclusion is – presuming the Van Allen Belts are as they were before – that humans having several times crossed them, can cross them again.
It doesn’t take age to know about the Moon landings.
Just to say, there’s no such thing as a “Boomer.” Or a Gen-Xer or a Millennial, or whatever. There’s only people.
The generational labels are a media fiction. They’re made to cause alienation of children from adults, and to give newsfolk a way to write lazy facile stories about fake generational conflict instead of doing the real work of investigative journalism.
Comment by Pat Frank — February 12, 2026 @ 6:13 pm
@Pat Another obvious conclusion is that, assuming we ever were able to cross through them safely, we long ago forgot how to, so that no one has ventured beyond low earth orbit since late 1972, and are now having to rediscover how to do so.
Comment by Ex-Global Super-Regulator on Lunch Break — February 12, 2026 @ 11:46 pm
All this assumes that the moon landings were not faked in a Hollywood studio.
Was this notion fabricated to make fools of the credulous chumps who adopted it i.e was it a false flag conspiracy theory?
Is that suggestion a meta-conspiracy-theory?
Comment by dearieme — February 13, 2026 @ 5:18 am
Thousands of amateur astronomers worldwide watched the moon shots. The astronauts left reflective targets on the surface for laser distancing. Satellites in low Lunar orbit have spotted landing vehicles and other materials left behind.
A more complete discussion, including Ex-G’s van Allen radiation belts:
https://www.iop.org/explore-physics/moon/how-do-we-know-we-went-to-the-moon
Comment by Pat Frank — February 13, 2026 @ 4:35 pm
@Pat, it’s a conundrum, yes?
Humans went fifty years ago, but since then humans haven’t been able to leave low-earth orbit, and even with a dedicated 21st century moonshot programme are still having trouble going beyond it.
What gives?
I love a mystery!
Comment by Ex-Global Super-Regulator on Lunch Break — February 13, 2026 @ 8:05 pm
@28 Ex-G – The original moon program (Mercury, Gemini, Apollo) was driven by Cold War politics. Beat the Russians to the moon.
When that program ended, the race was won and Federal money disappeared.
Launches were then too expensive for private companies. The U.S. started the shuttle program. It was ostensibly to raise satellites to low Earth orbit with a return vehicle. But that wasn’t popular, wasn’t particularly useful, and didn’t have much support – the least expensive, lowest technology was used.
So, it wasn’t that the technology was forgotten. It’s that there was no support for an expensive project, and no political vision to push investment.
It’s only now, with satellite communications bringing in consumer money, that there’s investment. That, and – thank-you Musk – launches are affordable. And he has the vision that’s been lacking for so long. Who knows: maybe SpaceX will get to the moon before NASA. It’s all now quite doable.
Comment by Pat Frank — February 14, 2026 @ 1:19 am
Yeah. Like I said, wake me when a mammal goes through the belts safely.
Comment by Ex-Global Super-Regulator on Lunch Break — February 14, 2026 @ 4:33 pm
@31 Ex G – Oh, yeah, I forgot. All the moon shot astronauts were marsupials.
Comment by Pat Frank — February 16, 2026 @ 12:37 am
You woke me. Has it happened? O, I guess not. LOL!
Comment by Ex-Global Super-Regulator on Lunch Break — February 16, 2026 @ 1:18 pm