SpaceX IPO confirmed for 2026 Elon personally confirmed SpaceX will go public in 2026.
Raiing massive capital for space-based AI data centers.
Expected to be one of the largest IPOs ever ($1.5 to $2 trillion)
Space-based AI data centers are now the main mission.
Earth CANNOT build power or chips fast enough → move AI compute to orbit, then Moon, then Mars.
Brian Wang is a Futurist Thought Leader and a popular Science blogger with 1 million readers per month. His blog Nextbigfuture.com is ranked #1 Science News Blog. It covers many disruptive technology and trends including Space, Robotics, Artificial Intelligence, Medicine, Anti-aging Biotechnology, and Nanotechnology.
Known for identifying cutting edge technologies, he is currently a Co-Founder of a startup and fundraiser for high potential early-stage companies. He is the Head of Research for Allocations for deep technology investments and an Angel Investor at Space Angels.
A frequent speaker at corporations, he has been a TEDx speaker, a Singularity University speaker and guest at numerous interviews for radio and podcasts. He is open to public speaking and advising engagements.
Uh, probably not worth the latency delays, says Grok:
Me: If a typically sized data center could be built in low Earth orbit, would it be able to keep up with requests to, and responses from, AI like yourself (Grok) using microwave technology at non-harmful levels? OR would latency doom such a project?
Grok: A hypothetical typically sized data center in low Earth orbit (LEO, around 400–600 km altitude) could handle interactive AI queries like those to Grok or similar conversational models, but latency would noticeably degrade the user experience compared to ground-based systems — though it wouldn’t completely “doom” the project for all use cases.Latency BreakdownThe fundamental limit comes from the speed of light (~300,000 km/s in vacuum, slightly slower in atmosphere).One-way propagation delay to/from LEO: Approximately 1.3–2 ms at zenith (direct overhead), up to ~4–5 ms at lower elevations.
Round-trip for a single query-response: ~10–20 ms minimum (uplink + downlink), often 20–40 ms in practice due to slant range and processing.
Full interaction: User → ground gateway → orbital data center → gateway → user adds at least 20–40 ms pure propagation delay.
Real-world analogs like Starlink (satellites at 550 km) achieve ~20–50 ms median latency for internet access, including multiple hops and routing. Direct communication to a dedicated orbital data center could approach the lower end (20 ms round-trip).Comparison to Current AI PerformanceConversational AIs like Grok or ChatGPT deliver:First-token latency (time to start responding): Often <200–500 ms on ground-based systems.
Total response time: 1–5 seconds for typical queries, depending on model and load.
Adding 20–40 ms orbital delay is small (~5–10% increase) relative to inference/compute time. Users already tolerate 50–100 ms network delays in terrestrial setups without noticing. However:It would feel slightly less "instant" than local data centers (e.g., <10 ms intra-region).
For voice mode or real-time interactions, cumulative delays could make conversations less fluid.
Microwave Technology and SafetyMicrowaves (typically in Ku/Ka bands for satellite comms) are standard for high-bandwidth links and pose no issue here.Power levels: Directional beams from ground stations and orbital antennas keep ground exposure thousands of times below FCC/ICNIRP safety limits (non-ionizing, no harm at regulated intensities).
Bandwidth: Easily supports high-throughput for AI requests/responses (gigabits/sec possible).
Many orbital data center proposals (e.g., from Axiom Space, Lumen Orbit, or EU's ASCEND studies) note ~20 ms round-trip latency as viable, especially for batch processing or non-real-time AI training, but less ideal for ultra-low-latency interactive use.VerdictLatency would add a perceptible but tolerable delay for chat-based AI interactions — similar to using Starlink for video calls today (workable, but not as seamless as fiber). It wouldn't doom the project outright, especially if the data center serves global/remote users evenly (reducing terrestrial routing delays) or focuses on less time-sensitive workloads. Primary barriers are cost, power/cooling in orbit, and launch logistics, not latency or microwave safety. For truly interactive, real-time AI like Grok, ground-based remains superior today.
Me: Details on ASCEND study
Grok: ASCEND Feasibility Study OverviewThe ASCEND (Advanced Space Cloud for European Net zero emission and Data sovereignty) study is a European Commission-funded feasibility project under the Horizon Europe program, launched in 2023 and completed in mid-2024. It was coordinated by Thales Alenia Space (a joint venture between Thales and Leonardo) with a consortium of partners including:Environmental analysis: Carbone 4, VITO
Cloud computing/IT: Orange Business, CloudFerro, Hewlett Packard Enterprise
Launchers: ArianeGroup
Orbital systems: German Aerospace Center (DLR), Airbus Defence & Space
The study received approximately €2 million in funding and lasted about 16 months.Key ObjectivesCompare the full lifecycle environmental impacts (CO₂ emissions, water usage, etc.) of space-based versus terrestrial data centers.
Assess the technical feasibility of designing, launching, deploying, assembling, and operating large-scale data centers in orbit.
Evaluate economic viability and potential contributions to Europe's digital sovereignty (reducing reliance on non-EU data centers) and the EU Green Deal's net-zero carbon goal by 2050.
Main Findings (Released June 2024)Technical Feasibility: Confirmed as viable using existing or near-term technologies, including:Modular, distributed architecture with robotic in-orbit assembly (building on Europe's EROSS IOD demonstrator, planned for 2026).
High-throughput optical (laser) communications for Earth links.
Radiation-hardened hardware and large deployable radiators for passive cooling in vacuum (no water needed).
Environmental Benefits: Space data centers could significantly reduce Earth's carbon footprint and eliminate water cooling requirements, but only if supported by a new reusable launcher 10 times less emissive over its lifecycle than current ones. Unlimited solar power in orbit (no night/cloud interruptions) is a major advantage.
Economic Viability: Promising, with potential multi-billion euro return on investment by 2050. The global data center market is projected to reach 23 GW by 2030; ASCEND targets deploying 1 GW in orbit by 2050.
Proposed Orbit and Scale: Medium Earth orbit at ~1,400 km altitude (lower debris risk than LEO). Initial phase: 13 modular building blocks totaling ~10 MW capacity by 2036 for commercialization start. Full structures could be larger than twice the International Space Station.
Advantages HighlightedNear-constant solar energy harvesting.
Natural radiative cooling in space.
Enhanced data security and sovereignty (harder to physically access or sabotage).
Reduced latency for space-based assets (e.g., satellites processing data on-orbit).
Challenges NotedNeed for major advancements in low-emission heavy-lift reusable launchers.
Radiation protection for electronics.
High initial costs and in-orbit maintenance (robotic only).
Potential security/political risks in space.
The study concludes that orbital data centers are a realistic path for sustainable, sovereign European cloud infrastructure, potentially transforming the digital sector while aligning with climate goals. No full implementation has started yet—this was purely a feasibility assessment. For more, see the official project site: ascend-horizon.eu or Thales Alenia Space announcements.
Actually, no amount of crystal ball gazing will ever fully cover this momentous event, the time that we humans finally decided to venture out into the cosmos. There is untold wealth to be had in precious metals mining, AI data centers, transport, propellant production, and energy production. It is not the technology that is holding us back, that has been available for forty odd years, it is a “mindset” thing……
Manned exploration of outer space comes next, in our lifetimes!
I am reminded of the Klondike gold rush….
https://en.wikipedia.org/wiki/Klondike_Gold_Rush
OFC I’ll bury a few hundred credits (they’re not ‘dollars’ anymore since the great inflation) in this IPO.
Space computing will have a point in the future, but I don’t see it for now.
The reason is mixed. As other commentator said, dissipate energy requires radiation, as void makes conduction impossible, and that’s bulky and expensive.
And move materials there is even more expensive.
On other side, SPECIAL DESIGNED CPUs can use a very low working temperature, and dissipate a lot less per calculation unit. At very low temperature, you generate very low heat.
So stay at a very low temperature can be a huge advantage. AND, if the GPU hardware don’t evolve quickly, most of the cost are concentrated in energy, which is the part you can optimize on space.
In summary. You don’t have the designed low temperature CPUs. You don’t have a stable architecture to amortize hardware over decades. Compute technology still change too fast.
So, you aren’t ready to be competitive with Earth computing by far.
Still, in a far future, I can see the economics of doing computing datacenters on the Moon underground, remotely controlled with robots, on the cost spots where cold is extremely cheap and you can drive solar energy from near places were almost every time light reach it. If necessary, it can be imported from different fields ensuring 24×7 energy and some backup nuclear reactor.
You can think that even with that, move electronics there is too costly, and that’s true, but if most infrastructure is reused and have high IRSU, I can see a point where the combined energy of using that CPU on Earth at room temperature can be higher than the energy to put that CPU on the Moon, if most infrastructure is reused.
But that’s a far away project, not for fast return. Most technology don’t exists.
I’ve had some long conversations with Grok, and it seems to be very upbeat about cooling solutions. There has been a great deal of innovation recently.
Please explain to me like I’m five years old how moving data centers to orbit will solve either the “can’t build chips fast enough” or “can’t build power fast enough” problems. Bonus: explain how cooling will work for all the waste heat.
You use a radiator. Since they’ll be in sun-synchronous orbit over the terminator, everything will be completely stable. The position of the sun, the heat from the earth, the direction in which to radiate. There won’t be any thermal cycling like happens to the ISS, or need to keep adjusting the radiator.
Well, sure, you can avoid thermal cycling in the right orbit.
The real problem is that you’re limited by the Stefan-Boltzmann Law. You pretty much have to reject heat at a low enough temperature to not cook your electronics into premature failure. Let’s say 150F. Which is a shame, because thermal radiation increases as the 4th power of temperature. You’re going to be radiating under a thousand watts per square meter.
Well, you’re using solar panels that might be delivering a few hundred watts per square meter, so that’s inflating the size of the over-all system, but it does mean that your server is spread out over a large area, (Say 3 square meters per kw of server.) and will have substantial speed of light delays built in.
It’s been decades since I was designing digital logic, and maybe that’s not actually an issue for the sort of computation being done here. I’m pretty sure Musk has been talking with people who’ve run the numbers, and he seems to think it’s feasible.
You can potentially afford a lot of downsides to escape micro-management by government, I guess.
I’d buy into this IPO, if it weren’t for the fact that the government strictly regulates who can buy into IPOs, so only the rich get richer, not the middle class.
They can get to higher temperatures by using standing waves in a resonator cavity to localize the heat and pipe it away. I read an article about it somewhere, but I can’t find it now. That introduces a mechanical vulnerability, but I think it can be managed.
Yes, you can refrigerate the chips, and then run your radiator at a higher temperature, and maybe that’s worth doing.
I’m just bummed that IPOs like this are reserved for the aristocracy, us peons are barred from them.
If Musk wants to give owners of TSLA shares the opportunity, he likely can find a method. This has the side effect of being a catalyst to TSLA share price and hence to his own net worth. It seems like the main reason TSLA was up significantly today while most of tech was down significantly.