Moving The Cloud to orbit

[Vultur]

but there's nothing at all to prevent buildout of terrestrial datacentres alongside buildout of terrestrial power plants in the middle of nowhere (i.e. land that is commercially worthless) to run them independent of the local grid.

In theory, no.

In practice, it's not always quite that easy. Many datacenters are trying to build new terrestrial power plants, and they are running into 3, 5, even 7 year wait times ordering the gas turbines.

To use solar ... Well unlike SSO or high orbit you need massive batteries. Really massive...

And land in the middle of nowhere is not always free of local opposition either. Some of those "commercially worthless" places have local communities that do value their land. A lot of people aren't happy with wind farms in parts of Texas with extremely low population density, because of their effect on the aesthetic landscape; noise issues with data center cooling can be similar.

Also, "commercially worthless" private land in the US almost always is desert or at least semi-desert. Thus cooling water is a huge problem.

[Paul451]

I will take your bet if you think it’s unlikely.

Assuming that was replying to my comment, what would I be betting?

That SpaceX doesn't launch at least 300,000 high-end Nvidia AI GPUs or equivalent by a certain date?

[Robotbeat]

I will take your bet if you think it’s unlikely.

Assuming that was replying to my comment, what would I be betting?

That SpaceX doesn't launch at least 300,000 high-end Nvidia AI GPUs or equivalent by a certain date?

Sure, let’s say 2035. And instead of specifying “high end NVidia,” let’s just give them a power rating. I say 50MW of GPUs.

BTW, the idea that terrestrial GPUs burn out in 18 months is false. The MTBF is far, far higher. (I suspect you’re going off some sandbagged and/or cherry picked numbers, assuming zero redundancy so just looking at the entire datacenter and how long it takes for maybe one of them to fail, etc).

[Vultur]

Them launching 50MW of GPUs in the next decade seems very likely.

Hundreds of GW significantly less so, TW scale even less.

I do not believe AI demand growth will continue forever (or even until 2035).

[StraumliBlight]

PowerBank and Smartlink AI (“Orbit AI”) to Launch the First “Orbital Cloud” for AI Infrastructure into Space [Nov 19]

Rocket launch of DeStarlink Genesis-1 satellite expected December 2025.

Orbit AI integrates DeStarlink (the first decentralized Starlink-style network) and DeStarAI (orbital AI data centers) into a unified space infrastructure where data can flow, compute, and verify in Low Earth Orbit (“LEO”) powered by Solar Energy.

[...]

Orbit AI is developing DeStarlink, the first decentralized low-Earth-orbit network for global connectivity, and DeStarAI, a suite of orbital AI data centers powered by solar arrays and naturally cooled in space. Together, these systems form the Orbital Cloud, a unified infrastructure layer designed to enable sovereign, censorship-resistant connectivity and in-orbit compute services.

[Robotbeat]

PowerBank and Smartlink AI (“Orbit AI”) to Launch the First “Orbital Cloud” for AI Infrastructure into Space [Nov 19]

Rocket launch of DeStarlink Genesis-1 satellite expected December 2025.

Orbit AI integrates DeStarlink (the first decentralized Starlink-style network) and DeStarAI (orbital AI data centers) into a unified space infrastructure where data can flow, compute, and verify in Low Earth Orbit (“LEO”) powered by Solar Energy.

[...]

Orbit AI is developing DeStarlink, the first decentralized low-Earth-orbit network for global connectivity, and DeStarAI, a suite of orbital AI data centers powered by solar arrays and naturally cooled in space. Together, these systems form the Orbital Cloud, a unified infrastructure layer designed to enable sovereign, censorship-resistant connectivity and in-orbit compute services.

No offense, but this article reads like it was written by AI (and not a new model, either).

[StraumliBlight]

No offense, but this article reads like it was written by AI (and not a new model, either).

Q4 2025: Launch of Genesis-1 with Ethereum wallet, blockchain node, and initial AI inference payload.

I was just curious about which rocket its launching on next month.

[Robotbeat]

I can’t tell if it’s real. It sounds completely hallucinated.

[Vultur]

They do have a website
https://www.orbitai.global/

It claims a "Genesis" mission November 29 (with a countdown clock). Farther down it says "Launch partner: Galactic Energy". And "First satellite with onboard blockchain wallet performing the world's first in-orbit blockchain transaction signature."

This feels like a pile of techy buzzwords.

The DeStarAI section also has some bizarre claims like "vacuum provides natural cooling" and "Compute Capacity: ∞ TFLOPS".

[Asteroza]

They do have a website
https://www.orbitai.global/

It claims a "Genesis" mission November 29 (with a countdown clock). Farther down it says "Launch partner: Galactic Energy". And "First satellite with onboard blockchain wallet performing the world's first in-orbit blockchain transaction signature."

This feels like a pile of techy buzzwords.

The DeStarAI section also has some bizarre claims like "vacuum provides natural cooling" and "Compute Capacity: ∞ TFLOPS".

Smells like a premined scamcurrency pump and dump, seeing as words like a decentralized network drop by. Even the classic blockchain genesis block. It speaks to the "defi" (decentralized finance) crowd.

In the past there was some talk of spaceborne blockchain nodes to try to reduce government regulation/manipulation. Perhaps someone though with such a high concentration of GPU's might allow some idle compute time capture for blockchain mining.


Though the decentralized bit got me thinking, could a heterogeneous cluster of unequal partner free participation LEO sats self assemble a routable overlay network without central coordination? The ultimate ugly mobile station mesh network problem?

[StraumliBlight]

Will Marshall (Planet Labs CEO) talks about Google's Suncatcher at 8:45.

Planet Labs Q3 Report [Dec 10]

Planet is contracted to deploy two prototype satellites, targeting launch in 2027.

[InterestedEngineer]

Will Marshall (Planet Labs CEO) talks about Google's Suncatcher at 8:45.

His ideas on how to prevent war are very stupid and ignorant.

It doesn't surprise me now that suncatcher's paper had some errors in it.

[JayWee]

https://twitter.com/elonmusk/status/1997706687155720229?t=kH2YWQNatyJ9PoOOCtoWWg&s=19

... replying to SX valuation ...
A major additional factor should be considered.

Satellites with localized AI compute, where just the results are beamed back from low-latency, sun-synchronous orbit, will be the lowest cost way to generate AI bitstreams in <3 years.

And by far the fastest way to scale within 4 years, because easy sources of electrical power are already hard to find on Earth. 1 megaton/year of satellites with 100kW per satellite yields 100GW of AI added per year with no operating or maintenance cost, connecting via high-bandwidth lasers to the Starlink constellation.

The level beyond that is constructing satellite factories on the Moon and using a mass driver (electromagnetic railgun) to accelerate AI satellites to lunar escape velocity without the need for rockets. That scales to >100TW/year of AI and enables non-trivial progress towards becoming a Kardashev II civilization.

[Vultur]

But who is paying for that much AI use?

Progress toward a Kardashev II civilization doesn't really make any sense in a world where the human population is predicted to peak at maybe 10B ish. Even Kardashev I energy levels don't make sense ... That's way more energy per person than can realistically be used.

The Dysonian idea of endless growth in energy use made sense in 1960 where population growth was fast. Once world population peaks and most of the world's population has access to the industrial/technological economy... demand growth basically stops. At that point, efficiency matters, and energy use probably *drops*.

(Using AI just for the sake of using AI isn't valuable. It only makes sense if it's cheaper than doing the same thing without AI, or gives real benefits that aren't otherwise attainable.)

Elon Musk has been incredibly successful, but not everything he says makes sense.

--
Also, the Navy has learned that railguns aren't actually practical due to rail erosion issues. In a world where Earth launch is cheap enough to do this stuff at all, there's no way railguns in the Moon are cost effective due to limited lifetime.

[Lee Jay]

But who is paying for that much AI use?

Progress toward a Kardashev II civilization doesn't really make any sense in a world where the human population is predicted to peak at maybe 10B ish. Even Kardashev I energy levels don't make sense ... That's way more energy per person than can realistically be used.

The Dysonian idea of endless growth in energy use made sense in 1960 where population growth was fast. Once world population peaks and most of the world's population has access to the industrial/technological economy... demand growth basically stops. At that point, efficiency matters, and energy use probably *drops*.

(Using AI just for the sake of using AI isn't valuable. It only makes sense if it's cheaper than doing the same thing without AI, or gives real benefits that aren't otherwise attainable.)

Elon Musk has been incredibly successful, but not everything he says makes sense.

--
Also, the Navy has learned that railguns aren't actually practical due to rail erosion issues. In a world where Earth launch is cheap enough to do this stuff at all, there's no way railguns in the Moon are cost effective due to limited lifetime.

Energy use should be dropping like a rock already.  Energy efficiency is VERY powerful.  I'm approaching a 90% reduction in my personal energy use just with energy efficiency.

[Twark_Main]

in a world where the human population is predicted to peak at maybe 10B ish.

The Quiverfull shall inherit the Earth.

Using AI just for the sake of using AI isn't valuable. It only makes sense if it's cheaper than doing the same thing without AI, or gives real benefits that aren't otherwise attainable.

"Otherwise attainable" at what price?

Current AI is cripplingly flawed, a technology in its infancy. But I suspect it won't be that way for too much longer.

Also, the Navy has learned that railguns aren't actually practical due to rail erosion issues. In a world where Earth launch is cheap enough to do this stuff at all, there's no way railguns in the Moon are cost effective due to limited lifetime.

I think "railgun" is being used here as a catch-all for any electric linear accelerator tech, as opposed to rocket launches, which when you pencil it out clearly wouldn't work.

[Vultur]

But who is paying for that much AI use?

Progress toward a Kardashev II civilization doesn't really make any sense in a world where the human population is predicted to peak at maybe 10B ish. Even Kardashev I energy levels don't make sense ... That's way more energy per person than can realistically be used.

The Dysonian idea of endless growth in energy use made sense in 1960 where population growth was fast. Once world population peaks and most of the world's population has access to the industrial/technological economy... demand growth basically stops. At that point, efficiency matters, and energy use probably *drops*.

(Using AI just for the sake of using AI isn't valuable. It only makes sense if it's cheaper than doing the same thing without AI, or gives real benefits that aren't otherwise attainable.)

Elon Musk has been incredibly successful, but not everything he says makes sense.

--
Also, the Navy has learned that railguns aren't actually practical due to rail erosion issues. In a world where Earth launch is cheap enough to do this stuff at all, there's no way railguns in the Moon are cost effective due to limited lifetime.

Energy use should be dropping like a rock already.  Energy efficiency is VERY powerful.  I'm approaching a 90% reduction in my personal energy use just with energy efficiency.

Until this year (datacenter boom) electricity consumption in the US has been very close to totally flat for 20 years, with energy efficiency improvements matching population growth.

Using AI just for the sake of using AI isn't valuable. It only makes sense if it's cheaper than doing the same thing without AI, or gives real benefits that aren't otherwise attainable.

"Otherwise attainable" at what price?

Current AI is cripplingly flawed, a technology in its infancy. But I suspect it won't be that way for too much longer.

"At what price" is exactly the question.

Right now, given the amount of investment money going toward AI, end users are generally not paying the 'real' cost of it.

If vastly more energy/hardware efficient means are found, of course that changes. But vastly less energy use means the purpose of moving it to orbit goes away.

I think "railgun" is being used here as a catch-all for any electric linear accelerator tech, as opposed to rocket launches, which when you pencil it out clearly wouldn't work.

Coil guns have major problems too... Is there any linear accelerator that would clearly be practical at multiple km/s for long enough lifetimes to compete with (say) $10-20/kg Earth launch costs?

I agree rocket launches for Moon resources wouldn't work ... I don't think mining the Moon makes sense in any scenario except for use by a Moonbase that exists for reasons other than mining.

[Eric Hedman]

Energy use should be dropping like a rock already.  Energy efficiency is VERY powerful.  I'm approaching a 90% reduction in my personal energy use just with energy efficiency.

In the last 25 years, the biggest improvement in efficiency has been in lighting.  Lighting used to consume about 22 percent of electricity put on the grid.  The last twenty five years has been the transition from incandescent lighting to LED bulbs Most of that transition has been done cutting that percent significantly.  Most homes built in the last thirty years are properly insulated so that transition is done.  Electric motors have always been very efficient so there isn't much to be gained with fans, pumps, machine tools, etc.  There are still some efficiency improvements, but the low hanging fruit has been mostly harvested.

AI is just exploding.  I use it with copilot in Microsoft Office.  It goes out to data centers to process what I'm doing.  We use it in my business to create marketing videos with avatars.  It can take quite a while for the results to come back from the data centers processing it.  We are starting to explore using AI in software development.

If you look at how professional sports are using AI for every advantage it is becoming insane.  NFL teams are having AI tools pour through the videos of thousands of college football games trying to find the rare diamond in the rough players.  They also use AI to analyze every move every player in their games are doing.  Brokerage firms are using AI to scour for the best investment opportunities.  I know people in banks that are being replaced by AI software.

Everyone seams to be jumping on the bandwagon.  This growth rate will flatten out, but it might not be for a decade or two.  That's where the opportunity lies and the demand for insane amounts of power.  For SpaceX, Blue Origin and others, this can be one heck of an opportunity to grow.  I see more and more communities saying no to data centers so the only option may be to look up.

[Lee Jay]

Energy use should be dropping like a rock already.  Energy efficiency is VERY powerful.  I'm approaching a 90% reduction in my personal energy use just with energy efficiency.

In the last 25 years, the biggest improvement in efficiency has been in lighting.

For me, lighting improvements have been quite a lot less than 1% of my gains.  Nearly all are in thermal control and transportation.

[Vultur]

Everyone seams to be jumping on the bandwagon. 

This is certainly true today ... The question is cost. The real cost of LLM queries (energy, hardware depreciation, etc) is currently not really being charged to users - otherwise you couldn't get Google AI Overviews, free ChatGPT use, Gemini on Android phones, etc. I think the key question is how many uses would/will survive if/when real costs have to be charged to end users.

(Or whether dramatically more energy efficient [non-LLM?] AI technologies will be developed, which might remove the entire issue but also remove the desirability of putting AI in orbit.]