Summarizer

We also shouldn't overlook the fact that the proposal entirely glosses over the implication of the alternative benefits we might realize if humanity achieved the incredible engineering and technical capacity necessary to make this version of space AI happen. Think about it. Elon conjures up a vision of the future where we've managed to increase our solar cell manufacturing capacity by two whole orders of magnitude and have the space launch capability for all of it along with tons and tons of other stuff and the best he comes up with is...GPUs in orbit? This is essentially the superhero gadget technology problem, where comic books and movies gloss over the the civilization changing implications of some technology the hero invents to punch bad guys harder. Don't get me wrong, the idea of orbiting data centers is kind of cool if we can pull it off. But being able to pull if off implies an ability to do a lot more interesting things. The problem is that this is both wildly overambitious and somehow incredibly myopic at the same time.

That’s not the point of the person you are replying to. They are saying if we somehow come up with the tech that makes harnessing the sun a thing, the best we can still do is put a bunch of GPUs in space? It makes no sense.

> It kinda does make sense if you consider that solar panels in space have been used for a very long time (to power satellites). It stops making sense the second you ask how you’d dissipate the heat any GPU would create. Sure, you could have vapour chambers. To where? Would this need square kilometers of radiators on top of square kilometers of solar panels? All this just to have Grok in space?

The plan seems to be for lots and lots of smaller satellites. For inferencing it can work well. One satellite could contain a handful of CPUs and do batch inferencing of even very large models, perhaps in the beginning at low speeds. Currently most AI workloads are interactive but I can't see that staying true for long, as things improve and they can be trusted to work independently for longer it makes more sense to just queue stuff up and not worry about exactly how high your TTFT is. For training I don't see it today. In future maybe. But then, most AI workloads in future should be inferencing not training anyway.

The problem Elon is trying to address is a societal one, not a technical one. The amount of push back on clean energy generation and manufacturing prevents data centers on earth from being as feasible as they should be. He only got his newly opened xAI data center open using temporary generators on trailers and skirting the permitting process by using laws designed for things like traveling circuses.

OK, so what are they? Scaling photovoltaic production doesn't seem likely to have many broader implications on its own. At best, it makes it easier to change the grid to renewable power, if you ignore the intermittency problem that still exists even at huge scales. PV fabs aren't really reusable for other purposes though, and PV tech is pretty mature already, so it's not clear what scaling that up will do. Scaling rocketry has several fascinating implications but Elon already covered many of them in his blog post. Scaling AI - just read the HN front page every day ;) What are we missing here? Some combinatoric thing?

Scaling up PV production to the point where we could convert the entire Earth's electricity generation to solar is incredibly significant. Yes there's the problem of intermittency, varying sun availability and so forth - which is why solar will never provide 100% of our power and we'll also need grid-scale storage facilities and domestic batteries and all sorts of stuff - but just imagine being able to make that many panels in the first place! Literally solar on every roof, that's transformative. But sure, let's send it all to space to power questionable "AI" datacentres so we can make more fake nudes.

Not to go heads I win, tails you lose, but even if we go down this path - it's the same story because militaries are investing heavily in LLM stuff, both overtly and covertly. Outside of its obvious uses in modeling, data management, and other such things - there also seems to be a fairly widespread belief, among the powers that be, that if you just say the magic words to somebody, that you can make them believe anything. So hyper-scaling LLM potential has direct military application, same as Starlink and Starship.

> This is a question that analysts don't even ask On the contrary, data centers continue to pop up deploying thousands of GPUs specifically because the numbers work out. The H100 launched at $30k GPU and rented for $2.50/hr. It's been 3 years since launch, the rent price is still around $2.50. During these 3 years, it has brought in $65k in revenue.

You really can't grasp that GPUs scaled at this level is the most ambitious thing possible? That it will be the foundation of unfathomable technological innovation?

"In space" is the new blockchain.

Will it, though?

> the most ambitious thing possible really?

Honestly, there's not a lot else I can think of if your goal is find some practical and profitable way to take advantage of relatively cheap access to near-Earth space. Communication is a big one, but Starlink is already doing that. One of the things space has going for it is abundant cheap energy in the form of solar power. What can you do with megawatts of power in space though? What would you do with it? People have thought about beaming it back to Earth, but you'd take a big efficiency hit. AI training needs lots of power, and it's not latency sensitive. That makes it a good candidate for space-based compute. I'm willing to believe it's the best low-hanging fruit at the moment. You don't need any major technological advances to build a proof-of-concept. Whether it's possible for this to work well enough that it's actually cheaper than an equivalent terrestrial datacenter now or in the near future is something I can't answer.

I don't think it's insane. It might not work or be competitive but it's not obviously insane. In a frictionless economy governed by spherical cows it'd be insane. But back here on Earth, AI is heavily bottlenecked by the refusal or inability of the supply chain to scale up. They think AI firms are in a bubble and will collapse, so don't want to be bag holders. A very sane concern indeed. But it does mean that inferencing (the bit that makes money) is constantly saturated even with the industry straining every sinew to build out capacity. One bottleneck is TSMC. Not much that can be done about that. The other is the grid. Grid equipment manufacturers and CCGT makers like Siemens aren't spinning up extra manufacturing capacity, again because they fear being bag holders when Altman runs out of cash. Then you have massive interconnection backlogs, environmentalists attacking you and other practical problems. Is it easier to get access to stable electricity supplies in space? It's not inconceivable. At the very least, in space Elon controls the full stack with nearly no regulations getting in the way after launch - it's a pure engineering problem of the sort SpaceX are good at. If he needs more power he can just build it, he doesn't have to try and convince some local government utility to scale up or give him air permits to run generators. In space, nobody can hear you(r GPUs) scream.

Hate to say this, but manufacturing bitcoin would make the most sense. And hard to see how even that would work.

But everyone is crazy about GPU’s right now. Why not ride that wave for extra investment? All the benefits transfer to all the other things we can do with it.

The distinction is that what they are doing for Webb is trying to dissipate small amounts of heat that would warm up sensors past cryogenic temperatures. Like on the order of tens or hundreds of watts but -100C. Dissipating heat for an AI datacenter is a different game. A single AI inference or training rack is going to be putting out somewhere around 100kW of waste heat. Temps don't have to be cryogenic but it's the difference between chiselling a marble or jade statue and excavating a quarry.

> Oh wait, that didn’t actually happen Not sure how you can say that. Nothing lasts forever, especially in the face of Chinese market dumping, but for a while there Tesla really was the undisputed king of EV manufacturing, that flywheel is how he got there, he did release all the patents because he said from day one he didn't anticipate or aim for 100% market share for Tesla and assumed there'd always be lots of EV manufacturers in future. All that sounds like - mission accomplished? As for Waymo being ahead, maybe today. But Waymo's tech stack is largely pre-DL, they rely heavily on unscalable techniques like LIDAR and continuous mapping. Tesla is betting big on the "scale up neural networks" model we know works well and their FSD can drive everywhere. They're perhaps behind Waymo in some ways, but they're also in different markets - Waymo won't sell anyone a self driving car and Tesla will. I wouldn't count them out. Their trajectory is the right one. > I’m not sure why anyone should bet on Elon pulling it off. PayPal, SpaceX existing at all, then doing reusable rockets, Tesla, FSD, large scale battery manufacturing, Starlink, X ("he can't fire 80% of employees it'll crash immediately"), robotics, training a SOTA LLM so fast even Jensen Huang was shocked ... the man consistently pulls off impossible seeming things in the face of huge skepticism. How many examples does it take before people start taking the guy seriously? Infinity examples?

We also shouldn't overlook the benefits we might realize if humanity achieved the incredible engineering and technical capacity necessary to make this version of porcine flight happen. IDK, what about the side-benefits of applying the "incredible engineering and technical capacity" to something useful instead? Rather than finding rationalisations for space spambots.

What have the engineers at XAI accomplished? From the ground level, it seems they followed the same research all the other LLM chatbot companies did. They followed along and made a sassy mecha hitler who makes revenge porn. XAI isn't a serious venture.

So, the much lauded xAI is overhyped, underwhelming and ... kind of evil? In stark contrast to every other AI company, I suppose? And people are using it for revenge porn? I haven't seen that. I've just seen that grok pioneered really good deep web search, is less woke than other LLMs and grok imagine has really good video generation and pretty good image gen.

spacex is one thing but xai accomplished what? the most racist csam prone llm?

I'm not aware of this - What's that?

Probably shouldn't speak to the brilliance of xAI engineers when you've never heard of their work

Is whatever that is their work?

AI clusters are heavily interconnected, the blast radius for single component failure is much larger than running single nodes -- you would fragment it beyond recovery to be able to use it meaningfully. I can't get in detail about real numbers but it's not doable with current hardware by a large margin.

How are you doing pci express x16 with lasers without fiber optics? Have you touched data center hardware in your life?

> [...] and initial capital investment. This is the big one - Musk knows that if he convinces enough people, they will invest the billions / trillions necessary, making him stupendously rich. But anyone investing in that is... not a good investor, to be politically correct, because what's the expected return on investment? Who are the customers? What is the monetization? Or bar that, how does it benefit humanity? It's throwing money down the drain. If you're an investor and are considering this, consider investing in earth instead. Real projects with real benefits. There's enough money to fix hunger, poverty, housing, education, and everything. Enough money to buy and / or fund politicians to make the necessary changes.

> So the proposal includes building a planetary-scale railway network on bumpy lunar terrain. And that’s from a fascist who barely managed to dig ONE small one lane tunnel under Las Vegas and called it a revolution. I’m sorry to be rude but people who are still giving musk any credit are stupid at this point. Oh boy, IA data centers in space. It’s not only ridiculous, but it’s also boring and not even exciting at all.

We seem to be using 100% of our DRAM manufacturing for AI. So it's not completely out of the question.

unless demand comes from space

the demand is pretty much fake and AI isn't actually making money, just gobbling investors money

You don't even need a particularly large scale, it's efficient resource utilization. Humanity has a finite (and too small) capacity for building solar panels. AI requires lots of power already. So the question is, do you want AI to consume X (where X is a pretty big chunk of the pie), or five times X , from that total supply? Using less PV is great, but only if the total cost ends up cheaper than installing 5X the capacity as terrestrial PV farms, along with daily smoothing batteries. SpaceX is only skating to where they predict the cost puck will be.

Those are for video. AI Chat workflows use a fraction of the data.

That's silly on so many levels. 1. the latency is going to be insane. 2. AI video exists. 3. vLLMa exist and take video and images as input. 4. When a new model checkpoint needs to go up, are we supposed to wait months for it to transfer? 5. A one million token context window is ~4MB. That's a few milliseconds terrestrially. Assuming zero packet loss, that's many seconds 6. You're not using TCP for this because the round trip time is so high. So you can't cancel any jobs if a user disconnects. 7. How do you scale this? How many megabits has anyone actually ever successfully sent per second over the distances in question? We literally don't know how to get a data center worth of throughput to something not in our orbit, let alone more than double digit megabits per second.

Grok doesn’t have video as far as I know. I don’t think it’s so absurd. I don’t know how you scale this. But it seems pretty straightforward.

Of course, we are stripping the earth bare to build word-guessers GPUs in orbit, but aliens are definitely the problem.

Yeah I don't know where this delusion of "Servers in space" came from, I think of it as the new NFTs But I bet those pushing for it are very interested in feeding the grift

>Robotaxi, your Optimus, your lunar lander, your space datacenter etc. And the list keeps getting longer instead of shorter... Lets go through this one by one [1]Robotaxi. Someone just drove coast to coast USA fully on autopilot. I drive my tesla every day, and i literally NEVER disengage autopilot. It gets me to work and back home without fail, to the grocery store, to literally anywhere i need. Whats not full self driving about that? I got in two crashes before i got my Tesla cause i was a dumb teen, but i'm sure my Tesla is a much better driver than my younger sister. Politically it's not FSD, but in reality, it has been for a while. [2] Optimus has gone through three revisions and has hand technology that is 5+ years ahead of the competition. Even if they launched it as a consumer product now, i'm sure a million people would buy it just as a cool toy/ gadget. AKA a successfull product. [3] Lunar Lander Starship, a fully reusable, 2 stage rocket that has gone through 25 revisions and is 95% flight proven and has even deployed dummy starlinks. 10+ years ahead of everyone except maybe stoke. [4]Space Datacenter Have you ever used starlink? They have all the pieces they need... Elon build a giant datacenter in 6 monmths when it takes 3-4 years usually. He has more compute than anybody and Grok is the most intelligent AI by all the metrics outside googles. Combine that with Starship, which can launch 10X the capacity for 10% of the cost, and what reason do you have to doubt him here? Granted... it always takes him longer than he says, but he always eventually comes through.

Okay, but a human being represents what, 200 W of power? The ISS has a crew of 3, so that's less than a beefy single user AI workstation at full tilt. If the question is whether it's practical to put 1-2 kW worth of computing power in orbit, the answer is obviously yes, but somehow I don't think that's what's meant by "datacenter in space".

I argue that if you have literal hundreds of billions of hard cash to burn for stupid things like AI datacenters, you could afford to make the lives of millions of starving people not suck instead, pretty easily so. But to do that, you'd have to try, and that would mean actually doing something good for humanity. Can't have that as a billionaire.

Who has hundreds of billions of hard cash for data centers? All of the AI spending has been in IOUs between Nvidia, OpenAI, Coreweave, etc. And even if you did have hard cash, how will you spend those billions? No one actually seems to have a sound plan, like I said. They just claim it can be done.

It's not just "very challenging", it's "very challenging and also solves no actual problem we face".

Google is currently working on AI data centers in space. https://blog.google/innovation-and-ai/technology/research/go...