You're right. Because it is. Twitter was hemmorhaging money . It made a little bit of sense too with the Grok integration. This helps when you can then just fold an unprofitable branch in an...
You're right. Because it is. Twitter was hemmorhaging money and buying it with xAI took Twitter private instead of public. It made a little bit of sense too with the Grok integration. This helps when you can then just fold an unprofitable branch in an otherwise profitable company without having to publicly report on the profitability. Now, without any clear connection between xAI and SpaceX the move just makes me wonder if xAI isn't profitable either, and to be clear most AI companies aren't, so folding it into SpaceX makes it part of a profitable whole.
Instead of letting your companies go bankrupt, you fold it into others. Like a shell game.
I'm going to wager that space-datacenters is just a cloud cover hiding the collapse.
Edit: Some clarification.
Edit2: Twitter became private when Elon bought it, not when xAI bought it.
Wasn't X already private? My understanding was that having xAI acquire X was to give it access to the investment money xAI was getting. It's not like there was much demand from private investors...
Twitter was hemmorhaging money and buying it with xAI took Twitter private instead of public.
Wasn't X already private? My understanding was that having xAI acquire X was to give it access to the investment money xAI was getting. It's not like there was much demand from private investors to invest in X, and they didn't have other significant revenue coming in, so that was the easiest way for Musk to keep it going.
Twitter itself was public. But you're correct that X was already private by the time he incorporated it into xAI. Musk made X private almost immediately upon buying Twitter through a full share...
Twitter itself was public. But you're correct that X was already private by the time he incorporated it into xAI. Musk made X private almost immediately upon buying Twitter through a full share acquisition. Good catch. My mistake!
This is interesting in the sense that SpaceX is the one of Musk's businesses that has been most successful at insulating itself from Musk, whereas Twitter and xAI have been pretty much personal...
This is interesting in the sense that SpaceX is the one of Musk's businesses that has been most successful at insulating itself from Musk, whereas Twitter and xAI have been pretty much personal fiefdoms.
Yeah, feels like a bailout to rescue his shitty AI with one of the only businesses' that is actually doing good work. I hope it doesn't drag down SpaceX too much.
Yeah, feels like a bailout to rescue his shitty AI with one of the only businesses' that is actually doing good work. I hope it doesn't drag down SpaceX too much.
Could you write little bit more about it? What's the difference between these companies regarding his involvement in each?
one of Musk's businesses that has been most successful at insulating itself from Musk, whereas Twitter and xAI have been pretty much personal fiefdoms.
Could you write little bit more about it? What's the difference between these companies regarding his involvement in each?
Not OP but I mentioned it a bit in another comment. SpaceX's president, Gwynne Shotwell, runs a lot of the business and keeps Musk more in line. It also helps that SpaceX operates in a highly...
Not OP but I mentioned it a bit in another comment. SpaceX's president, Gwynne Shotwell, runs a lot of the business and keeps Musk more in line. It also helps that SpaceX operates in a highly regulated environment. If you want to do a bunch of crazy stuff you are pretty confined to what you can do, or otherwise have to plan it out months in advance to get approvals from the FAA and other agencies.
Thank you. Knowing that SpaceX gets money mostly from government subsidies/contracts, which means it gets government money, that also means to me that Musk found a way to get government money for...
Thank you.
Knowing that SpaceX gets money mostly from government subsidies/contracts, which means it gets government money, that also means to me that Musk found a way to get government money for his idiotic projects such as X. I hope I'm mistaken, but for sure one thing will happen - Spacex enshittification, because lots of its money will be spent on Xai. That's what I think.
Oof, space-based data-centers won’t just be competing with normal data-centers. If a small offline server can securely and effectively achieve the same results albeit somewhat slower, you’ll have...
Oof, space-based data-centers won’t just be competing with normal data-centers. If a small offline server can securely and effectively achieve the same results albeit somewhat slower, you’ll have to convince people getting work done in seconds rather than minutes or even hours is worth the upload to the space-cloud and extra cost.
I think xAI in space will fail for the same reasons people don’t mind flying for 8 hours across the Atlantic even though we have the technology to make the journey in 4. All in all, do enough people need the speed to be productive?
True, but that’s what this acquisition is signaling to the world. SpaceX is bullish on servers in space. To me it looks like a big uninformed gamble, but I guess shareholders are convinced it’s...
True, but that’s what this acquisition is signaling to the world. SpaceX is bullish on servers in space. To me it looks like a big uninformed gamble, but I guess shareholders are convinced it’s the future.
On the flip-side, it’s a good starting point for a space-based civilization.
It could be something I guess, but the entire idea of space based data centers seems incredibly flawed. Data centers produce extreme amounts of heat. Getting rid of heat is a massive problem in...
It could be something I guess, but the entire idea of space based data centers seems incredibly flawed.
Data centers produce extreme amounts of heat. Getting rid of heat is a massive problem in space. You can only radiate it away so fast. And that's before getting into shielding from cosmic rays and repairability. RAM is expensive enough before you have to put it on a rocket to get it to your server farm for replacement.
The entire idea seems like something that 'sounds sci-fi' so it makes for good sound bites of "maybe we could do this" to suggest alternatives to LLMs taking up power and water on Earth.
One thing that the last 20 years of investors in Musk's companies should have shown us, it's that shareholders are very, very stupid. It's extremely easy for Musk to trick them by invoking sci fi...
One thing that the last 20 years of investors in Musk's companies should have shown us, it's that shareholders are very, very stupid. It's extremely easy for Musk to trick them by invoking sci fi ideas with no feasibility to drum up excitement.
He drummed up investment in SpaceX by promising crewed mars missions by 2024. He drummed up investment in Tesla by promising self driving in 2014.
He's a proven liar who consistently makes things up to boost stock prices to his advantage and people keep falling for it.
Data centers in space is not only an uninformed gamble, it's a completely unworkable idea from a technical standpoint, let alone an economic standpoint.
It would literally make more sense to invest in a company that promises to build factories at the bottom of the ocean, or farms in Antarctica.
This is such a disappointment for SpaceX. It was doing a pretty decent job at staying focused on a sensible but difficult goal of enabling cheaper and more performant space travel, while doing it...
This is such a disappointment for SpaceX. It was doing a pretty decent job at staying focused on a sensible but difficult goal of enabling cheaper and more performant space travel, while doing it profitably and for less than competitors. In large part because Shotwell isolated it from Musk for the day to day operations. But this will just drain SpaceX dry and divert resources away from better projects. Also how does SpaceX owning Twitter make any sense?
Extremely disappointing. I am a HUGE fan of SpaceX and what it is trying to accomplish, but this is likely the beginning of the end. I can't see any of this being anything but negative. What a shame.
Extremely disappointing. I am a HUGE fan of SpaceX and what it is trying to accomplish, but this is likely the beginning of the end. I can't see any of this being anything but negative. What a shame.
Suspiciously close to the rumored IPO? I'm sure it's just a coincidence, of course. Good that it may be backfiring a bit now that France authorities raided their offices...
Suspiciously close to the rumored IPO? I'm sure it's just a coincidence, of course. Good that it may be backfiring a bit now that France authorities raided their offices https://www.bbc.com/news/articles/ce3ex92557jo
It makes sense to do it before the IPO to avoid the regulatory hassles that come with publicly listed companies. While they're both still private, elon has zero accountability and can easily do...
Suspiciously close to the rumored IPO?
It makes sense to do it before the IPO to avoid the regulatory hassles that come with publicly listed companies. While they're both still private, elon has zero accountability and can easily do this, whereas after, he'd end up getting sued by shareholders and have to endure a long legal battle
Also, xAI is absolutely hemmoraging money with no path to profitability in sight. If he takes it public again, he can use his legendary bullshitting skills to get retail shareholders to burden a...
Also, xAI is absolutely hemmoraging money with no path to profitability in sight. If he takes it public again, he can use his legendary bullshitting skills to get retail shareholders to burden a lot of that risk like they do with Tesla. Then he doesn't actually have to deliver good products via either SpaceX or X anymore.
He just has to string people along to pump the share price when he wants to, which is the main thing he's good at.
From the article: [...] A reason this might not be nuts is that SpaceX has launched more satellites than any other company. Their Starlink satellites do work, so maybe they can figure out power,...
From the article:
The merger could be seen as an indication of the cash xAI needs to compete in the fast-growing field of AI, as well as the technology’s importance in future of space exploration.
[...]
“Current advances in AI are dependent on large terrestrial data centers, which require immense amounts of power and cooling,” Musk wrote in a post on SpaceX’s website. “The only logical solution therefore is to transport these resource-intensive efforts to a location with vast power and space.”
Musk said he estimates that the “lowest cost way to generate AI compute will be in space” within two to three years.
SpaceX sought permission on Friday from the Federal Communication Commission to launch a constellation of 1 million satellites into orbit. The filing said the aim is to provide a network of solar-powered data centers to “accommodate the explosive growth of data demands driven by AI.”
A reason this might not be nuts is that SpaceX has launched more satellites than any other company. Their Starlink satellites do work, so maybe they can figure out power, cooling, and logistics for GPU's?
There is no logical way this would cost less than literally anywhere else, even if you could just snap you fingers and solve the power/cooling problems tomorrow. This is another "and that's why i...
so maybe they can figure out power, cooling, and logistics for GPU's?
There is no logical way this would cost less than literally anywhere else, even if you could just snap you fingers and solve the power/cooling problems tomorrow.
This is another "and that's why i need congress to fund me and bend laws" grift. I NEED all this funding to put processors in space that could go on any empty swath of land or even underwater easier.
I can think of one way, though it's rather dubious: if NIMBY opposition to new data centers goes up enough that they can't build them on Earth for a reasonable amount of money.
I can think of one way, though it's rather dubious: if NIMBY opposition to new data centers goes up enough that they can't build them on Earth for a reasonable amount of money.
You could literally build them cheaper in the Shara, the ocean, or the arctic. Let alone the thousands of square miles in various other countries who would gladly take investment for land. It...
You could literally build them cheaper in the Shara, the ocean, or the arctic. Let alone the thousands of square miles in various other countries who would gladly take investment for land. It would take a literally catastrophic level of NIMBY. That's before you even look at the US.
The level of money required for this is going to be just astronomically stupid and it's trying to entwine his AI/Datacenter needs with his lift capacity company.
It's completely stupid on at least three separate fronts: Launching. Take a low margin commodity (running computing hardware) and do it in the literal most difficult and expensive location outside...
It's completely stupid on at least three separate fronts:
Launching. Take a low margin commodity (running computing hardware) and do it in the literal most difficult and expensive location outside of putting them on another planet.
Cooling. Vacuums insulate heat, and computers generate a ton. It takes an absurd amount of costly equipment to keep the International Space Station habitable, and that's not full of servers with high-wattage GPUs. A single server is roughly analogous to a 1KW space heater, and they're stacked on racks 42 high, and there will be hundreds of racks in a small data center.
Electricity. A single H100 GPU, excluding the rest of the server, pulls 700W. Lets just say a single server is still a 1kW spec heater. 42 high, 100 racks = 4200kW, and that's ignoring the larger cost of cooling (each of those 100 racks needs about a half dozen window AC units worth of cooling on earth, and it's harder to cool things in space). 1kW of solar generation is about 2 square meters of space, so we're talking kilometers...
And this is all contextualizing on a small scale. These data centers they're building out with their own power plants dwarf this. Realistically, we don't produce enough solar to meet humanity's needs on earth, and if we could increase that enough to make space servers viable on that single front, we'd have far more need of that to avoid fucking climate change.
For comparison, the ISS, the biggest and most expensive space construction project in human history by far, requring dozens of countries decades to build and over $120 billion dollars generates......
Electricity. A single H100 GPU, excluding the rest of the server, pulls 700W. Lets just say a single server is still a 1kW spec heater. 42 high, 100 racks = 4200kW, and that's ignoring the larger cost of cooling (each of those 100 racks needs about a half dozen window AC units worth of cooling on earth, and it's harder to cool things in space). 1kW of solar generation is about 2 square meters of space, so we're talking kilometers...
For comparison, the ISS, the biggest and most expensive space construction project in human history by far, requring dozens of countries decades to build and over $120 billion dollars generates... 120kw.
I could buy a generator for $50,000 that can supply that, or solar panels here on earth for about $200,000. That's 150 GPUs, ignoring cooling. Large AI data centers run hundreds of thousands of them.
Yeah, IIRC, AI-focused data centers are hundreds of megawatts in some cases, and some of the largest planned are pushing into gigawatt territory. That's powering an entire city territory. A medium...
Yeah, IIRC, AI-focused data centers are hundreds of megawatts in some cases, and some of the largest planned are pushing into gigawatt territory. That's powering an entire city territory. A medium US city can be like 10GW (roughly one St Louis), and Tokyo (one of the largest cities in the world) has something like 60GW of generation capacity.
It really is a staggeringly stupid waste, hurting strides made against climate change...all to suppress wages/eliminate jobs by making shit-quality graphics, visuals, text and code and/or enable bullshitters to more effectively waste the time of people doing actual work.
Yeah I find it quite hard to believe running this type of operation in space to be viable. The impression I get more is that Elon wants to monopolize space, like the space in Earth's orbit, and...
Yeah I find it quite hard to believe running this type of operation in space to be viable. The impression I get more is that Elon wants to monopolize space, like the space in Earth's orbit, and this is a justification for that. Yes I know the Earth is huge and there's a lot of orbital space around Earth, but because of the nature of wireless signals and managing wireless signals, managing objects in space and managing launches to get into space and so on, the more stuff there is in space, the harder it will be for other people to launch more stuff into space. The early entrants into utilizing space in Earth's orbit will be able to make claims against latecomers that more stuff in orbit is too dangerous, or harms their interests etc.
At that point it doesn't really even matter if the data center idea is a dud, if you want approval to get your stuff into space, you might have to pay Elon for the privilege.
Most earth orbits, especially the easy to get to ones, also eliminate the one possible advantage of having things in space. If something is flying around the earth, about half the time its...
Most earth orbits, especially the easy to get to ones, also eliminate the one possible advantage of having things in space. If something is flying around the earth, about half the time its orbiting, the earth is occluding its view of the sun. You could put something in the orbit of the sun instead, but that would require wayyyyy more launch energy to do.
The US alone is a big place. And there are always places and people who are willing to take some paltry handouts (or maybe its the other way around, giving tax incentives) and the promise of jobs...
they can't build them on Earth for a reasonable amount of money.
The US alone is a big place. And there are always places and people who are willing to take some paltry handouts (or maybe its the other way around, giving tax incentives) and the promise of jobs (however few) or whatever, even if electricity costs skyrocket. Maybe places like Northern Virginia are getting fed up with data centers, but other places are still willing to host them.
No ones backyard is in the middle of the Mojave, which would be a way cheaper and easier place to build data centers. Most large scale data centers are already in the middle of nowhere, and if...
No ones backyard is in the middle of the Mojave, which would be a way cheaper and easier place to build data centers. Most large scale data centers are already in the middle of nowhere, and if NIMBY opposition grew, which is entirely reasonable, they would move even further to the middle of nowhere. The US has an absolutely insane amount of land area that's available.
To be fair, I don't really want that land eaten up by data centers either. But the idea that space is somehow better or cheaper is incredibly illogical
To be fair, I don't really want that land eaten up by data centers either. But the idea that space is somehow better or cheaper is incredibly illogical
Neither do I, but I hear this argument a lot about running out of space, usually in the context of solar panels taking up a lot of space. People don't comprehend how much empty space exists in the...
Neither do I, but I hear this argument a lot about running out of space, usually in the context of solar panels taking up a lot of space. People don't comprehend how much empty space exists in the US, especially if they haven't spent much time in rural areas in the western part of the country.
"Figure out cooling" is hand-waving a huge problem away. The ISS requires massive radiators to dissipate heat, while a single datacenter can pump heat to many homes and provide liveable...
"Figure out cooling" is hand-waving a huge problem away. The ISS requires massive radiators to dissipate heat, while a single datacenter can pump heat to many homes and provide liveable temperatures in cold climates.
It seems like many people with no physics/engineering understanding are vastly underestimating how much heat is produced by a datacenter, and/or overestimating the rate of heat exchange in near-vaccuum.
It's this. I know technologically minded people who don't understand that vacuum insulates. They think, "space is cold, right?" and never consider that you need some form of physical contact to...
... and/or overestimating the rate of heat exchange in near-vaccuum.
It's this. I know technologically minded people who don't understand that vacuum insulates. They think, "space is cold, right?" and never consider that you need some form of physical contact to dissipate the majority of the heat electronics produce.
As someone who works in the space sector, the constant hand-waving about cooling for these proposed space based data centres makes me want to scream. It's a near-insurmountable problem that is...
As someone who works in the space sector, the constant hand-waving about cooling for these proposed space based data centres makes me want to scream. It's a near-insurmountable problem that is constantly being ignored as if it's trivial. It's absolutely insane.
Sattelites worked before SpaceX launched them. They already knew they worked, and we knew how to launch them. They also launch them because there are no feasibile alternatives to provide worldwide...
Sattelites worked before SpaceX launched them. They already knew they worked, and we knew how to launch them.
They also launch them because there are no feasibile alternatives to provide worldwide radio coverage. Google tried using balloons, terrestrial cell providers use towers, but all of those have significant drawbacks compared to sattelites if your goal is covering the entire world.
Data centers are not in the same situation. No one has built a data center in space before. We have good alternatives to putting them in space. There aren't significant constraints to building them on earth, and there a lot of contraints to putting them into space (as there is for putting anything into space).
There also aren't any significant advantages to putting them in space. Yes, they use a lot of power on earth. They'd use a lot of power in space too though. The power requirements don't magically go away because they're in space. Yes, solar panels are a lot more efficient in space, but solar panels are extremely cheap now. Launching them into space is not. Any efficiency gains are totally wiped out by that fact, and that would be the only possible advantage.
Launching data centers into space seem like a great way to save money on hardware replacement parts and paying for techs to do hardware maintenance. This is a real money saving solution
Launching data centers into space seem like a great way to save money on hardware replacement parts and paying for techs to do hardware maintenance. This is a real money saving solution
I can't tell if you're being ironic, but saving money on maintenance by throwing the hardware away if it fails isn't obviously a way to save money. But I suppose that could be a forcing function...
I can't tell if you're being ironic, but saving money on maintenance by throwing the hardware away if it fails isn't obviously a way to save money. But I suppose that could be a forcing function to make the whole thing more reliable.
Much like Starlink doesn't compete with fiber on price or capacity, but has its niche, perhaps there's a niche for this.
I was being ironic, and probably should have made my tone a bit more obvious. I think it is really short sighted since as parts fail, best case the servers become less powerful (a RAM stick...
I was being ironic, and probably should have made my tone a bit more obvious.
I think it is really short sighted since as parts fail, best case the servers become less powerful (a RAM stick failing, but the machine stays operational since there is other RAM) or worst case the machine is broken because of an easily replaceable non-optional part (which would be even worse if it was a cheap part). I would hazard a guess that server maintenance is a really small line item of a data center's budget, but being unable to do any maintenance since it is in space would be even worse.
I think it being a way of forcing function to make the whole thing more unreliable is not an outcome that would happen. Outside of the Intel CPUs that were defectively burning themselves out or the NVidia GPU power connectors, there does not seem to be any major issues with reliability of hardware. You could argue that hardware lifespans have the potential to be lengthened, and that space data centers would be an incentive to do that. However, I think the reason components do not have longer lifespans is it is harder to predict what the market needs 7 years in advance. Which then raises the question, how do you accurately build out a data center for AI that you can not change the hardware? What happens if the resource demands for AI drastically changes after you launch the hardware?
Because I don't want to add a full-on post for this, here's Elon's blog post confirming the acquisition and the "rationale" behind the purchase: https://www.spacex.com/updates#xai-joins-spacex Of...
Because I don't want to add a full-on post for this, here's Elon's blog post confirming the acquisition and the "rationale" behind the purchase: https://www.spacex.com/updates#xai-joins-spacex
Of course, 0 mention of what we're using the compute for, why we should be eating up that much power just because, and NOT EVEN A WHISPER of how they'd solve the cooling problem.
XAI owns Twitter and now Space X buys xAI....
This feels like a shell game.
You're right. Because it is. Twitter was hemmorhaging money
and buying it with xAI took Twitter private instead of public. It made a little bit of sense too with the Grok integration. This helps when you can then just fold an unprofitable branch in an otherwise profitable company without having to publicly report on the profitability. Now, without any clear connection between xAI and SpaceX the move just makes me wonder if xAI isn't profitable either, and to be clear most AI companies aren't, so folding it into SpaceX makes it part of a profitable whole.Instead of letting your companies go bankrupt, you fold it into others. Like a shell game.
I'm going to wager that space-datacenters is just a cloud cover hiding the collapse.
Edit: Some clarification.
Edit2: Twitter became private when Elon bought it, not when xAI bought it.
Wasn't X already private? My understanding was that having xAI acquire X was to give it access to the investment money xAI was getting. It's not like there was much demand from private investors to invest in X, and they didn't have other significant revenue coming in, so that was the easiest way for Musk to keep it going.
Twitter itself was public. But you're correct that X was already private by the time he incorporated it into xAI. Musk made X private almost immediately upon buying Twitter through a full share acquisition. Good catch. My mistake!
I would not be surprised to see 'Tesla buys SpaceX' somewhere down the line to help meet the requirements for Musk's trillion dollar pay package.
This is interesting in the sense that SpaceX is the one of Musk's businesses that has been most successful at insulating itself from Musk, whereas Twitter and xAI have been pretty much personal fiefdoms.
Yeah, feels like a bailout to rescue his shitty AI with one of the only businesses' that is actually doing good work. I hope it doesn't drag down SpaceX too much.
Oh I hope it kills SpaceX, that would be hilarious.
Because it is. Xai was reported to lose a billion per month.
Could you write little bit more about it? What's the difference between these companies regarding his involvement in each?
Not OP but I mentioned it a bit in another comment. SpaceX's president, Gwynne Shotwell, runs a lot of the business and keeps Musk more in line. It also helps that SpaceX operates in a highly regulated environment. If you want to do a bunch of crazy stuff you are pretty confined to what you can do, or otherwise have to plan it out months in advance to get approvals from the FAA and other agencies.
Thank you.
Knowing that SpaceX gets money mostly from government subsidies/contracts, which means it gets government money, that also means to me that Musk found a way to get government money for his idiotic projects such as X. I hope I'm mistaken, but for sure one thing will happen - Spacex enshittification, because lots of its money will be spent on Xai. That's what I think.
Oof, space-based data-centers won’t just be competing with normal data-centers. If a small offline server can securely and effectively achieve the same results albeit somewhat slower, you’ll have to convince people getting work done in seconds rather than minutes or even hours is worth the upload to the space-cloud and extra cost.
I think xAI in space will fail for the same reasons people don’t mind flying for 8 hours across the Atlantic even though we have the technology to make the journey in 4. All in all, do enough people need the speed to be productive?
That's assuming "space based data centers" have possibility to exist at all.
True, but that’s what this acquisition is signaling to the world. SpaceX is bullish on servers in space. To me it looks like a big uninformed gamble, but I guess shareholders are convinced it’s the future.
On the flip-side, it’s a good starting point for a space-based civilization.
It could be something I guess, but the entire idea of space based data centers seems incredibly flawed.
Data centers produce extreme amounts of heat. Getting rid of heat is a massive problem in space. You can only radiate it away so fast. And that's before getting into shielding from cosmic rays and repairability. RAM is expensive enough before you have to put it on a rocket to get it to your server farm for replacement.
The entire idea seems like something that 'sounds sci-fi' so it makes for good sound bites of "maybe we could do this" to suggest alternatives to LLMs taking up power and water on Earth.
One thing that the last 20 years of investors in Musk's companies should have shown us, it's that shareholders are very, very stupid. It's extremely easy for Musk to trick them by invoking sci fi ideas with no feasibility to drum up excitement.
He drummed up investment in SpaceX by promising crewed mars missions by 2024. He drummed up investment in Tesla by promising self driving in 2014.
He's a proven liar who consistently makes things up to boost stock prices to his advantage and people keep falling for it.
Data centers in space is not only an uninformed gamble, it's a completely unworkable idea from a technical standpoint, let alone an economic standpoint.
It would literally make more sense to invest in a company that promises to build factories at the bottom of the ocean, or farms in Antarctica.
This is such a disappointment for SpaceX. It was doing a pretty decent job at staying focused on a sensible but difficult goal of enabling cheaper and more performant space travel, while doing it profitably and for less than competitors. In large part because Shotwell isolated it from Musk for the day to day operations. But this will just drain SpaceX dry and divert resources away from better projects. Also how does SpaceX owning Twitter make any sense?
Extremely disappointing. I am a HUGE fan of SpaceX and what it is trying to accomplish, but this is likely the beginning of the end. I can't see any of this being anything but negative. What a shame.
Suspiciously close to the rumored IPO? I'm sure it's just a coincidence, of course. Good that it may be backfiring a bit now that France authorities raided their offices https://www.bbc.com/news/articles/ce3ex92557jo
It makes sense to do it before the IPO to avoid the regulatory hassles that come with publicly listed companies. While they're both still private, elon has zero accountability and can easily do this, whereas after, he'd end up getting sued by shareholders and have to endure a long legal battle
Also, xAI is absolutely hemmoraging money with no path to profitability in sight. If he takes it public again, he can use his legendary bullshitting skills to get retail shareholders to burden a lot of that risk like they do with Tesla. Then he doesn't actually have to deliver good products via either SpaceX or X anymore.
He just has to string people along to pump the share price when he wants to, which is the main thing he's good at.
From the article:
[...]
A reason this might not be nuts is that SpaceX has launched more satellites than any other company. Their Starlink satellites do work, so maybe they can figure out power, cooling, and logistics for GPU's?
But I expect it will be late.
There is no logical way this would cost less than literally anywhere else, even if you could just snap you fingers and solve the power/cooling problems tomorrow.
This is another "and that's why i need congress to fund me and bend laws" grift. I NEED all this funding to put processors in space that could go on any empty swath of land or even underwater easier.
I can think of one way, though it's rather dubious: if NIMBY opposition to new data centers goes up enough that they can't build them on Earth for a reasonable amount of money.
You could literally build them cheaper in the Shara, the ocean, or the arctic. Let alone the thousands of square miles in various other countries who would gladly take investment for land. It would take a literally catastrophic level of NIMBY. That's before you even look at the US.
The level of money required for this is going to be just astronomically stupid and it's trying to entwine his AI/Datacenter needs with his lift capacity company.
It's completely stupid on at least three separate fronts:
Launching. Take a low margin commodity (running computing hardware) and do it in the literal most difficult and expensive location outside of putting them on another planet.
Cooling. Vacuums insulate heat, and computers generate a ton. It takes an absurd amount of costly equipment to keep the International Space Station habitable, and that's not full of servers with high-wattage GPUs. A single server is roughly analogous to a 1KW space heater, and they're stacked on racks 42 high, and there will be hundreds of racks in a small data center.
Electricity. A single H100 GPU, excluding the rest of the server, pulls 700W. Lets just say a single server is still a 1kW spec heater. 42 high, 100 racks = 4200kW, and that's ignoring the larger cost of cooling (each of those 100 racks needs about a half dozen window AC units worth of cooling on earth, and it's harder to cool things in space). 1kW of solar generation is about 2 square meters of space, so we're talking kilometers...
And this is all contextualizing on a small scale. These data centers they're building out with their own power plants dwarf this. Realistically, we don't produce enough solar to meet humanity's needs on earth, and if we could increase that enough to make space servers viable on that single front, we'd have far more need of that to avoid fucking climate change.
But Elmo is a moron, we all know that.
For comparison, the ISS, the biggest and most expensive space construction project in human history by far, requring dozens of countries decades to build and over $120 billion dollars generates... 120kw.
I could buy a generator for $50,000 that can supply that, or solar panels here on earth for about $200,000. That's 150 GPUs, ignoring cooling. Large AI data centers run hundreds of thousands of them.
Yeah, IIRC, AI-focused data centers are hundreds of megawatts in some cases, and some of the largest planned are pushing into gigawatt territory. That's powering an entire city territory. A medium US city can be like 10GW (roughly one St Louis), and Tokyo (one of the largest cities in the world) has something like 60GW of generation capacity.
It really is a staggeringly stupid waste, hurting strides made against climate change...all to suppress wages/eliminate jobs by making shit-quality graphics, visuals, text and code and/or enable bullshitters to more effectively waste the time of people doing actual work.
Yeah I find it quite hard to believe running this type of operation in space to be viable. The impression I get more is that Elon wants to monopolize space, like the space in Earth's orbit, and this is a justification for that. Yes I know the Earth is huge and there's a lot of orbital space around Earth, but because of the nature of wireless signals and managing wireless signals, managing objects in space and managing launches to get into space and so on, the more stuff there is in space, the harder it will be for other people to launch more stuff into space. The early entrants into utilizing space in Earth's orbit will be able to make claims against latecomers that more stuff in orbit is too dangerous, or harms their interests etc.
At that point it doesn't really even matter if the data center idea is a dud, if you want approval to get your stuff into space, you might have to pay Elon for the privilege.
Most earth orbits, especially the easy to get to ones, also eliminate the one possible advantage of having things in space. If something is flying around the earth, about half the time its orbiting, the earth is occluding its view of the sun. You could put something in the orbit of the sun instead, but that would require wayyyyy more launch energy to do.
The US alone is a big place. And there are always places and people who are willing to take some paltry handouts (or maybe its the other way around, giving tax incentives) and the promise of jobs (however few) or whatever, even if electricity costs skyrocket. Maybe places like Northern Virginia are getting fed up with data centers, but other places are still willing to host them.
No ones backyard is in the middle of the Mojave, which would be a way cheaper and easier place to build data centers. Most large scale data centers are already in the middle of nowhere, and if NIMBY opposition grew, which is entirely reasonable, they would move even further to the middle of nowhere. The US has an absolutely insane amount of land area that's available.
To be fair, I don't really want that land eaten up by data centers either. But the idea that space is somehow better or cheaper is incredibly illogical
Neither do I, but I hear this argument a lot about running out of space, usually in the context of solar panels taking up a lot of space. People don't comprehend how much empty space exists in the US, especially if they haven't spent much time in rural areas in the western part of the country.
"Figure out cooling" is hand-waving a huge problem away. The ISS requires massive radiators to dissipate heat, while a single datacenter can pump heat to many homes and provide liveable temperatures in cold climates.
It seems like many people with no physics/engineering understanding are vastly underestimating how much heat is produced by a datacenter, and/or overestimating the rate of heat exchange in near-vaccuum.
It's this. I know technologically minded people who don't understand that vacuum insulates. They think, "space is cold, right?" and never consider that you need some form of physical contact to dissipate the majority of the heat electronics produce.
As someone who works in the space sector, the constant hand-waving about cooling for these proposed space based data centres makes me want to scream. It's a near-insurmountable problem that is constantly being ignored as if it's trivial. It's absolutely insane.
of course
Sattelites worked before SpaceX launched them. They already knew they worked, and we knew how to launch them.
They also launch them because there are no feasibile alternatives to provide worldwide radio coverage. Google tried using balloons, terrestrial cell providers use towers, but all of those have significant drawbacks compared to sattelites if your goal is covering the entire world.
Data centers are not in the same situation. No one has built a data center in space before. We have good alternatives to putting them in space. There aren't significant constraints to building them on earth, and there a lot of contraints to putting them into space (as there is for putting anything into space).
There also aren't any significant advantages to putting them in space. Yes, they use a lot of power on earth. They'd use a lot of power in space too though. The power requirements don't magically go away because they're in space. Yes, solar panels are a lot more efficient in space, but solar panels are extremely cheap now. Launching them into space is not. Any efficiency gains are totally wiped out by that fact, and that would be the only possible advantage.
It's really a nonsensical idea.
Launching data centers into space seem like a great way to save money on hardware replacement parts and paying for techs to do hardware maintenance. This is a real money saving solution
I can't tell if you're being ironic, but saving money on maintenance by throwing the hardware away if it fails isn't obviously a way to save money. But I suppose that could be a forcing function to make the whole thing more reliable.
Much like Starlink doesn't compete with fiber on price or capacity, but has its niche, perhaps there's a niche for this.
I was being ironic, and probably should have made my tone a bit more obvious.
I think it is really short sighted since as parts fail, best case the servers become less powerful (a RAM stick failing, but the machine stays operational since there is other RAM) or worst case the machine is broken because of an easily replaceable non-optional part (which would be even worse if it was a cheap part). I would hazard a guess that server maintenance is a really small line item of a data center's budget, but being unable to do any maintenance since it is in space would be even worse.
I think it being a way of forcing function to make the whole thing more unreliable is not an outcome that would happen. Outside of the Intel CPUs that were defectively burning themselves out or the NVidia GPU power connectors, there does not seem to be any major issues with reliability of hardware. You could argue that hardware lifespans have the potential to be lengthened, and that space data centers would be an incentive to do that. However, I think the reason components do not have longer lifespans is it is harder to predict what the market needs 7 years in advance. Which then raises the question, how do you accurately build out a data center for AI that you can not change the hardware? What happens if the resource demands for AI drastically changes after you launch the hardware?
It was very clearly obvious to me already, just to provide some feedback.
Because I don't want to add a full-on post for this, here's Elon's blog post confirming the acquisition and the "rationale" behind the purchase:
https://www.spacex.com/updates#xai-joins-spacex
Of course, 0 mention of what we're using the compute for, why we should be eating up that much power just because, and NOT EVEN A WHISPER of how they'd solve the cooling problem.