This is a masterful piece of financial engineering by Google and SpaceX.
Google purchased 10% of SpaceX over a decade ago. After dilution they probably own around 5%.
SpaceX is valued at a whopping 94x revenue. This deal increases SpaceX's revenue by $11 billion per year. If SpaceX maintains this revenue multiplier, then the single deal boosts SpaceX's valuation by 94 x 11 billion = $1 trillion dollars. Google owns 5% of SpaceX, so they make 50 billion dollars. Google spends 10 billion and makes 50 billion, $40 billion profit.
The even better part is that because of this deal, SpaceX is now profitable. The S&P requires companies to demonstrate 12 months of profits before they can enter the S&P 500 index. SpaceX lobbied to have this profitability requirement removed, but S&P said no and refused to rewrite the rules.
Now with this incredible deal, SpaceX is now GAAP profitable under the existing rules, and they get to join the index next year without a rule change.
> SpaceX is trading at a whopping 94x revenue. This deal increases SpaceX's revenue by $11 billion per year. If SpaceX maintains this revenue multiplier, then the single deal boosts SpaceX's valuation by 94 x 11 billion = $1 trillion dollars.
That final number doesn't make sense: if you're trading shares at $X revenue, increasing the revenue by $Y multiplier doesn't increase the share price by the same multiplier.
A huge chunk of SoaceX value in their filing is attributed to their AI technology (aka Grok). I believe it’s 90% or more… Now, it seems they’re leasing the infrastructure required for Grok to scale to Anthropic and Google. I wonder how that math works…
But what is xAI? I thought that was the company that had the compute + Grok, the AI company? Since when does SpaceX (which I thought was a space company?) own AI-compute hardware and/or can do model hosting? Are all of Musks companies just one big thing now where the names no longer matter, or how is it supposed to work?
Edit: seems I'm just a bit behind: "xAI — now part of SpaceX ", seems really strange for a space company to buy an AI company, but I guess rather that, than the other way around.
Not really strange... if the goal is to go to mars, you probably need robots, those need intelligence -> ai. It fits pretty well, especially because you want to own all the core technologies as a company.
Why 4-5 companies instead of one then? I thought the goal of SpaceX was to get to Mars, why does xAI need to have that same goal? Or he didn't think xAI was suitable for that goal, then changed his mind so merged the companies?
Google renting infra from xAI, I did not see that coming. My understanding of what computers are doing, what companies are doing and what governments are doing seems to be getting worse day by day.
> Google renting infra from xAI, I did not see that coming.
Actually that seems to be fairly logical? Hardware is what xAI has, and it's in great demand. So sell what makes you money. The real story here is that that xAI hardware is going to be running Gemini and not Grok. Which is to say: Grok basically failed as a frontier AI and they need to pivot to a business model which makes money.
Obviously not everything Musk did was wrong. xAI bought a ton of compute when it was possible to get it. But the product they were going to build with it failed and so now they're deciding to be a landlord.
This IPO is just insane. No way do you justify a $trillion+ valuation based on what amounts to a bunch of commoditized rent seeking endeavors. Datacenters are buildings and chips, and everyone can build those. Starlink is just an ISP with lots of competition at scale (they have the high bandwidth mobile market cornered, but that's a very small market!). Mars is at best a grift on public funding. Even satellite launch services are commoditized and competetive these days.
> Which is to say: Grok basically failed as a frontier AI and they need to pivot to a business model which makes money.
They can just run Grok as a local AI inside Tesla cars. It's actually really efficient as a compute platform because the Tesla cars are in motion at highway speeds, which gives you lots of free airflow for shedding waste heat via the car radiator. Way more efficient than trying to run AI on space satellites.
The future needs more AI compute. No one has enough AI compute.
Memory chip vendors are betting hard on this being a temporary state of affairs that doesn't last, and doesn't warrant commissioning a shitton of new memory foundries.
Musk is betting hard on this staying that way, and is putting the next Colossus into the last place not corrupted by NIMBYs... SPACE!
Is this admission that google’s proprietary chips etc. are not cutting it? Why would you need a bunch of nvidia GPUs if you have your own silicon? (AFAIK they have their own for both inference and training do they not?)
How do you come to this conclusion? All it means is that spacex has compute and google does not.
Suppose tpus were theoretically a million times better, but cannot be produced due to supply chain constraints, this action would still be rational.
My personal take is that this really shows how bottlenecked the entire supply chain is. For such an important commodity there are shockingly few players ready for scale.
3 years is quite a short horizon when it comes to semiconductor fabs. Also this article is a dupe, when it was previously discussed it surfaced that after some time either party can cancel with only 90 days notice.
Its because none of the promised Data Center and NVIDIA hardware deployments described in NVDA earnings calls have actually happened. Once more Ed Zitron has the goods: https://youtu.be/zbKDmkJPVvI?t=482
Kinda; while it does show that overall Google's proprietary chips etc. are not cutting it, it doesn't say if the problem is the hardware itself or the factories to make more of the hardware. Without more information, it could be that Google's hardware is 100x the energy efficiency per token, but they can only make enough hardware for 1% of the tokens there's currently demand for: 1% of your product being 0.01% of your costs isn't nothing, but it leaves the other 99% at full price.
Supply and demand? Bubblists seem to think there's an infinite supply of chips, power, and water to make as many chat bots as possible; physics, as usual, dictates limits.
Or alternatively there is simply a huge demand for compute and this is helping them fill a short-term need. Keep in mind if you saw in the article there is a 90-day cancellation clause. This is a nothing burger.
Is there any data on whether Google, Amazon, Microsoft, Anthropic, OpenAI etc are most cost efficient in getting datacenter compute online and operating it?
I'd be interested in how large the range is here across company and region and specific data center and how it relates to companies like Hetzner if at all.
Well, Elon seems to take the fastest path possible to these DCs. One can envision a future where these get shut down for the severity of the pollution, not to mention being built and operated illegally [0].
> Is there any data on whether Google, Amazon, Microsoft, Anthropic, OpenAI etc are most cost efficient in getting datacenter compute online and operating it?
Well considering that ~80% of the price is hardware deprecation, I don't know why they'd be considerably worse than anyone else at negotiating hardware deals.
Typically when you buy in bulk, you have more sway.
Companies like Google also have in-house chips like TPUs that are substantially cheaper for inference for them to make than anyone else can get through Nvidia.
I’ve seen some numbers related to datacenters in Ireland and they would stress price per MW as a way to see where to build them. But then you have depreciation of equipment as well. Depreciation can be played with when filing taxes though.
I don't think they are most efficient for small GPUs. I think they might only be the one which have capex and certainty required for multimillion dollar purchase of GB200 NVL72 or something of that scale.
> As part of that deal, Anthropic agreed to pay SpaceX $1.25 billion per month through 2029 to rent all the available compute from its Colossus 1 data center near Memphis, Tennessee, that xAI — now part of SpaceX — originally built for its own artificial intelligence efforts.
I don't get why SpaceX is going public. But anyway, well played, the whole crypto mining that dried out GPUs back in the day seems tiny now.
These deals are part of how the AI economy operates. Amodei has explained this in his recent Patel podcast.
1. Building datacenters takes time. Months, if not years. They take billions of investment.
2. AI revenue is highly unpredictable. Sure, you can make predictions, but maybe your competitor releases a better model 2 weeks after your release, maybe the new model you built isn't as much better, maybe the chinese models steal your show, etc.
3. AI revenue grows a lot. Anthropic's case is 10x per year.
4. So if you are off by just a year in terms of how much GPU you actually need, then that means a 90% of your compute capacity is wasted, and you go bankrupt.
As a solution, companies buy compute from each other! If one company's model did well, they can buy compute from the company whose model didn't do well (like in the case of grok). It's beneficial for both sides, so positive sum game. So deals like this aren't something bad in itself.
It's nothing new either. In SAAS deals, you often commit to a certain revenue and then pay extra if your revenue exceeds that amount. And power market is cut in two as well: longer term deals plus spot markets. Spot prices are way higher than the longer term deal prices.
Given it's SpaceX of course there is financial engineering involved: the GPUs aren't actually owned by SpaceX but a daughter company, and it's been financed via loans that are backed by pension funds. So it's already the case that pension funds back bear the risks associated with SpaceX's operations.
Right now, the bulk of the AI bubble sits in such debt statements and not in public markets.
“If we fail to deliver access to the committed amount of GPUs by September 30, 2026, then following a one-month grace period, Google may immediately terminate the agreement or accept the number of GPUs provided, with a corresponding pro rata reduction in the monthly fees. After December 31, 2026, the agreement may be terminated by either party upon 90 days' notice.”
SpaceX has recently started pitching itself as an orbital datacenter company.
If you buy into that business model (or pretend to), it makes sense for SpaceX to start selling compute early. Their "earthside compute" clients of today are "skyside compute" clients of tomorrow.
A part of Musk's old pitch for Starlink was: space-based solar makes perfect sense for powering space assets, and no sense whatsoever for powering Earth assets. So you have to find a way to use that power in space to do something economically useful. Comms were the only scalable way to do that, so Starlink it was.
I can see how space-based datacenters would follow the same logic. If SpaceX can make them economical, that is. There's no guarantees of that - but if anyone at all can make space-based datacenters economical, it's SpaceX.
> if anyone at all can make space-based datacenters economical, it's SpaceX
Let's hope burning ten thousand tons of toxic e-waste annually in upper atmoshphere never becomes economical. Or mankind gets to senses and bans externalizing your e-waste problem by burning in atmosphere...
There are no NIMBYs in space. No government permitting on land use. And solar power is plentiful. It's like having a dollar store Dyson sphere.
Making use of that is predicated entirely on being able to put a lot of hardware into space cheaply. SpaceX is the undisputed best at that, no one comes close. The question is whether their "best" is good enough to make space datacenters economical.
The data link between earth and space has so much bandwidth.
There are sensors in space that send data to earth it gets processed and then the data is sent back to space then to the end user back on earth. If you do the compute in space you save the space-earth transfer time twice. Latency and availability of bandwidth are both factors.
There may be limited utility for this outside of military.
Evaporative cooling is the way it happens down on earth - and that shuttles h2o molecules from dense useful clumps like aquifers and rivers to a less useful form spread out in the air. But evaporating h2o isn’t an option in space afaik - since there’s a shortage of air to take up the h2o. In fact I think radiative cooling is the only actual option in space.
Can anyone explain how the thermals will work? One of the biggest challenges on Earth is cooling the data center, and it's at least as challenging in space.
The earthbound equivalent would be strapping each chassis to the back of a dedicated solar panel and having the panel double as a giant heat sink. The problem is that doesn't work on the surface due to (at least) rain, the day/night cycle, and the cost of real estate.
There are asteroids with concentrations of precious metals more valuable than earth's entire economy. Why don't we just send up spaceships to mine them and send the haul back to earth? What country would say no to free money?
After all, it's just an engineering challenge, not impossible.
The numbers on that are at least somewhat questionable. Even ignoring that you'd crash the market (thus it's not actually worth what it first appears to be) what is the total fuel cost to adjust the orbit of the target asteroid to land the entire thing back on the earth? Because that's what you're doing bit by bit as you shuttle loads of ore back.
Now if you have space based manufacturing or fuel production on the other hand ...
When I hear space I think "that's the perfect location for a data center", since data centers are lightweight, small, require little power, don't need human intervention, have lifetimes measured in decades and don't have to reject heat. Since space easily satisfies these requirements, space is an ideal deployment location for data centers.
Isn't the Vegas Loop just a car tunnel? As far as I know, there aren't any actual hyperloops[1] involved, just a narrow highway, even if they deceivingly brand it "Loop".
It seems off at first glance but actually appears to work out if you do the math. You can model a solar panel as a flat, opaque rectangle. You can calculate power generation and equilibrium temperature for it based on surface area. If you require additional radiative surface area to achieve the desired equilibrium temperature you can place a flat triangle orthogonal to and behind the solar panel in its shadow.
Compute is "free" at that point because waste heat is coming out of the total energy flux which was already accounted for (because we modeled it as opaque).
Of course swapping out the equipment poses a bit of a challenge. The "helping hands" rate is entirely unaffordable and wait until you see this new DC's physical access policies. 0/10 would not rack with them again.
This is all just the typical Elon hate. What's desperate about getting paid $920,000,000 per month? If that's desperation, I'd love to start groveling more!
Given extreme supply constraints, it's very unlikely that Google or Anthropic will just suddenly cancel right after the IPO unless their own demand collapses. And even if this were true, what value would that provide Musk? Could you imagine if your newly public company suddenly received termination notices from your two largest compute customers? Disaster.
I have no love or hate for Elon Musk. I wish him luck with his space endeavours.
What's desperate is announcing a temporary (allegedly) doubling of revenues days before an IPO that has been criticised for being overpriced at 93 times sales.
These data centers were supposed to serve xAI. Now suddenly they get rented out to others. Why the sudden change of plans?
It's either an emergency accounting gimmick or the effective shutdown or repurposing of xAI.
It’s a repurposing of xAI to be a commodity service provider during a crunch for that commodity. It would be dumb if xAI had any quality or market traction, but they have neither, so it’s actually a rational fallback. But it writes off any high margin future in favor of low margin scale.
And once the compute crunch is over, they’ll have a lot of overprovisioned data centers with no business to soak up the capacity.
Didn't Anthropic pull the same in both ways? you pull me up I pull you up kind of deal? Sounds like SpaceX bought themselves some time up to Q4, which is not the case of Anthropic and even worse for OpenAI. Not counting that none of them got their S&P500 fast-track ticket.
If you look at the IPO filings you’ll see that Spacex as we know it is just a small part of the expected revenue generator. It is supposedly Grok and AI, hence Google competitor.
All companies are now AI companies.
Just like a while ago all companies were suddenly Ads companies.
The entire tech sector is one big FOMO - once you reach certain scale you do exactly the same thing as everyone else.
Their satellite internet business is the only thing which makes them money, which is enabled by their orbital launch business which is as of right now not profitable and I have no idea of if it ever will be but without it they would not be able to launch enough satellites.
Their stupidity with AI and buying X mostly seems to be about scamming investors to make Musk even richer. Like this particular deal is just them doing what CoreWeave does at a SpaceX valuation.
Launch isn't profitable simply because ongoing Starship R&D is eating into it. A lot of opex, capex, and pre-revenue.
If they start running Starship anywhere near the way they do Falcon 9, it'll flip into profits. A lot of big bets SpaceX made ride on Starship coming online. I'm honestly surprised Starlink is already so profitable without it.
One of their big named bets includes: orbital datacenters. Which puts this specific deal into perspective.
80% of the space launch business is putting starlink satellites into orbit, so it's all internal funny money. They could very well be letting the space launch business take losses to make the satellite internet business look better (only profitable part of the whole thing).
Wasn't starship supposed to be funded by the NASA contract?
Sorry, I was unclear. With that I did not talk about this particular deal. This particular deal seems sane. XAI built more compute that they can use themselves since Grok is not very successful so to not just have the hardware standing there they rent it out to competitors. Makes total sense.
It is other things Musk has gone with Twitter and SpaceX which are shady.
That's only about 35% more than the main telecom operator here Belgium (Proximus: $7.2B revenue in 2025, $2.5B market cap, positive earnings for 15+ years).
Obviously Starlink can and will growth. I'm just pointing out how insane the market cap is, when compared to similar scale "connectivity" businesses.
Starlink terminals are popular, they put them on drones to avoid jamming (Starling jamming exists but not that easy for now). It might be their sales are inflated due to its use at war.
This feels actually like a pretty safe bet for Google, they secure the compute in case it works (I doubt that the described volume will be available in the near future), while if SpaceX doesn’t manage to provide there is not much loss.
I see it more as another way of blowing up SpaceX valuation on paper…
Does this mean that SpaceX are the only company that really did build some datacenters to put all the million of GPU/TPU/whatever they all talk about everyday?
I mean, Google, Amazon, Meta and Microsoft told investors they spent more than $1B per day last year in CapEx... why on Earth do they (well, Google and Anthropic at least) need to rent compute to SpaceX, of all companies?
Other companies built data centers but also built products that soak up their data center capacity.
xAI built data centers, and products that are mostly good for nonconsensual porn and confirming a small group’s biases. So they have a lot of excess capacity, and might as well rent it to the adults.
No, CoreWeave for example also rent compute to the big AI companies. This likely just means Grok has few users so they need to rent their extra capacity to their competitors.
Yes but someone will be along shortly to defuse what sounds like giving the bad mars man credit where it’s due. Like everything else he does that works out, it was just luck, timing, actually a mistake that worked out, or someone else behind the scenes that he got lucky in hiring at the right time (by accident).
If your dad had owned an emerald mine I am sure you could also have been that dumb.
But to be more serious: It is impossible to say if this is good or bad for XAI without more numbers. What if they bought their compute way over market price and sell it at a loss?
This is a masterful piece of financial engineering by Google and SpaceX.
Google purchased 10% of SpaceX over a decade ago. After dilution they probably own around 5%.
SpaceX is valued at a whopping 94x revenue. This deal increases SpaceX's revenue by $11 billion per year. If SpaceX maintains this revenue multiplier, then the single deal boosts SpaceX's valuation by 94 x 11 billion = $1 trillion dollars. Google owns 5% of SpaceX, so they make 50 billion dollars. Google spends 10 billion and makes 50 billion, $40 billion profit.
The even better part is that because of this deal, SpaceX is now profitable. The S&P requires companies to demonstrate 12 months of profits before they can enter the S&P 500 index. SpaceX lobbied to have this profitability requirement removed, but S&P said no and refused to rewrite the rules.
Now with this incredible deal, SpaceX is now GAAP profitable under the existing rules, and they get to join the index next year without a rule change.
Truly a brilliant deal for everyone involved.
> SpaceX is trading at a whopping 94x revenue. This deal increases SpaceX's revenue by $11 billion per year. If SpaceX maintains this revenue multiplier, then the single deal boosts SpaceX's valuation by 94 x 11 billion = $1 trillion dollars.
That final number doesn't make sense: if you're trading shares at $X revenue, increasing the revenue by $Y multiplier doesn't increase the share price by the same multiplier.
A huge chunk of SoaceX value in their filing is attributed to their AI technology (aka Grok). I believe it’s 90% or more… Now, it seems they’re leasing the infrastructure required for Grok to scale to Anthropic and Google. I wonder how that math works…
But what is xAI? I thought that was the company that had the compute + Grok, the AI company? Since when does SpaceX (which I thought was a space company?) own AI-compute hardware and/or can do model hosting? Are all of Musks companies just one big thing now where the names no longer matter, or how is it supposed to work?
Edit: seems I'm just a bit behind: "xAI — now part of SpaceX ", seems really strange for a space company to buy an AI company, but I guess rather that, than the other way around.
I think some justify it as SpaceX plans to offer hosting in space, and then use Starlink to distribute it.
Not really strange... if the goal is to go to mars, you probably need robots, those need intelligence -> ai. It fits pretty well, especially because you want to own all the core technologies as a company.
Why 4-5 companies instead of one then? I thought the goal of SpaceX was to get to Mars, why does xAI need to have that same goal? Or he didn't think xAI was suitable for that goal, then changed his mind so merged the companies?
Wow…sounds like some kindergarten stuff
Google renting infra from xAI, I did not see that coming. My understanding of what computers are doing, what companies are doing and what governments are doing seems to be getting worse day by day.
> Google renting infra from xAI, I did not see that coming.
Actually that seems to be fairly logical? Hardware is what xAI has, and it's in great demand. So sell what makes you money. The real story here is that that xAI hardware is going to be running Gemini and not Grok. Which is to say: Grok basically failed as a frontier AI and they need to pivot to a business model which makes money.
Obviously not everything Musk did was wrong. xAI bought a ton of compute when it was possible to get it. But the product they were going to build with it failed and so now they're deciding to be a landlord.
This IPO is just insane. No way do you justify a $trillion+ valuation based on what amounts to a bunch of commoditized rent seeking endeavors. Datacenters are buildings and chips, and everyone can build those. Starlink is just an ISP with lots of competition at scale (they have the high bandwidth mobile market cornered, but that's a very small market!). Mars is at best a grift on public funding. Even satellite launch services are commoditized and competetive these days.
> Which is to say: Grok basically failed as a frontier AI and they need to pivot to a business model which makes money.
They can just run Grok as a local AI inside Tesla cars. It's actually really efficient as a compute platform because the Tesla cars are in motion at highway speeds, which gives you lots of free airflow for shedding waste heat via the car radiator. Way more efficient than trying to run AI on space satellites.
Grok is just DOA. No need to beat a dead horse. Even Musk got that thus the reason why he is renting the stuff it planned for Grok.
They're struggling.
The future needs more AI compute. No one has enough AI compute.
Memory chip vendors are betting hard on this being a temporary state of affairs that doesn't last, and doesn't warrant commissioning a shitton of new memory foundries.
Musk is betting hard on this staying that way, and is putting the next Colossus into the last place not corrupted by NIMBYs... SPACE!
When as appears inevitable Google decides to stop using this capability what will it do to the SpaceX stock value?
Is this admission that google’s proprietary chips etc. are not cutting it? Why would you need a bunch of nvidia GPUs if you have your own silicon? (AFAIK they have their own for both inference and training do they not?)
How do you come to this conclusion? All it means is that spacex has compute and google does not.
Suppose tpus were theoretically a million times better, but cannot be produced due to supply chain constraints, this action would still be rational.
My personal take is that this really shows how bottlenecked the entire supply chain is. For such an important commodity there are shockingly few players ready for scale.
It's a very long contract (till 2029) for just covering themselves for supply.
3 years is quite a short horizon when it comes to semiconductor fabs. Also this article is a dupe, when it was previously discussed it surfaced that after some time either party can cancel with only 90 days notice.
No its not.
"Both SpaceX and Google have the option to terminate the agreement with 90 days’ notice after December 31, 2026"
Its because none of the promised Data Center and NVIDIA hardware deployments described in NVDA earnings calls have actually happened. Once more Ed Zitron has the goods: https://youtu.be/zbKDmkJPVvI?t=482
Kinda; while it does show that overall Google's proprietary chips etc. are not cutting it, it doesn't say if the problem is the hardware itself or the factories to make more of the hardware. Without more information, it could be that Google's hardware is 100x the energy efficiency per token, but they can only make enough hardware for 1% of the tokens there's currently demand for: 1% of your product being 0.01% of your costs isn't nothing, but it leaves the other 99% at full price.
At this point you are not buying a particular chip. You are buying whatever compute you can get.
Not necessarily, just that they don't have as many as they can make use of, and that xAI can't make more valuable use of them than renting them out.
Or it’s paying to make sure competition can’t buy said compute. Also isn’t Google an investor in SpaceX anyways?
Supply and demand? Bubblists seem to think there's an infinite supply of chips, power, and water to make as many chat bots as possible; physics, as usual, dictates limits.
Yes, it is issue of scale, google had to restrict usage because hardware are not available, regardless of what kind of hardware that is
It could be, or simply we are so far away into chip shortage that even google needs to buy from other people’s pot.
Or alternatively there is simply a huge demand for compute and this is helping them fill a short-term need. Keep in mind if you saw in the article there is a 90-day cancellation clause. This is a nothing burger.
Is there any data on whether Google, Amazon, Microsoft, Anthropic, OpenAI etc are most cost efficient in getting datacenter compute online and operating it?
I'd be interested in how large the range is here across company and region and specific data center and how it relates to companies like Hetzner if at all.
Well, Elon seems to take the fastest path possible to these DCs. One can envision a future where these get shut down for the severity of the pollution, not to mention being built and operated illegally [0].
[0] https://www.selc.org/news/xai-built-an-illegal-power-plant-t...
> Is there any data on whether Google, Amazon, Microsoft, Anthropic, OpenAI etc are most cost efficient in getting datacenter compute online and operating it?
Well considering that ~80% of the price is hardware deprecation, I don't know why they'd be considerably worse than anyone else at negotiating hardware deals.
Typically when you buy in bulk, you have more sway.
Companies like Google also have in-house chips like TPUs that are substantially cheaper for inference for them to make than anyone else can get through Nvidia.
I’ve seen some numbers related to datacenters in Ireland and they would stress price per MW as a way to see where to build them. But then you have depreciation of equipment as well. Depreciation can be played with when filing taxes though.
I don't think they are most efficient for small GPUs. I think they might only be the one which have capex and certainty required for multimillion dollar purchase of GB200 NVL72 or something of that scale.
that's asking the cart before the horse; is there any data on what compute actually results in real GDP improvements?
Nothing other than vendor promises and white papers.
380 dollars per second... Good to know I could afford my own data center for an appreciable fraction of a minute.
plus $473 per second from Anthropic
> As part of that deal, Anthropic agreed to pay SpaceX $1.25 billion per month through 2029 to rent all the available compute from its Colossus 1 data center near Memphis, Tennessee, that xAI — now part of SpaceX — originally built for its own artificial intelligence efforts.
I don't get why SpaceX is going public. But anyway, well played, the whole crypto mining that dried out GPUs back in the day seems tiny now.
For context, Alphabet earns ~$12k/sec.
These deals are part of how the AI economy operates. Amodei has explained this in his recent Patel podcast.
1. Building datacenters takes time. Months, if not years. They take billions of investment.
2. AI revenue is highly unpredictable. Sure, you can make predictions, but maybe your competitor releases a better model 2 weeks after your release, maybe the new model you built isn't as much better, maybe the chinese models steal your show, etc.
3. AI revenue grows a lot. Anthropic's case is 10x per year.
4. So if you are off by just a year in terms of how much GPU you actually need, then that means a 90% of your compute capacity is wasted, and you go bankrupt.
As a solution, companies buy compute from each other! If one company's model did well, they can buy compute from the company whose model didn't do well (like in the case of grok). It's beneficial for both sides, so positive sum game. So deals like this aren't something bad in itself.
It's nothing new either. In SAAS deals, you often commit to a certain revenue and then pay extra if your revenue exceeds that amount. And power market is cut in two as well: longer term deals plus spot markets. Spot prices are way higher than the longer term deal prices.
Given it's SpaceX of course there is financial engineering involved: the GPUs aren't actually owned by SpaceX but a daughter company, and it's been financed via loans that are backed by pension funds. So it's already the case that pension funds back bear the risks associated with SpaceX's operations.
Right now, the bulk of the AI bubble sits in such debt statements and not in public markets.
“If we fail to deliver access to the committed amount of GPUs by September 30, 2026, then following a one-month grace period, Google may immediately terminate the agreement or accept the number of GPUs provided, with a corresponding pro rata reduction in the monthly fees. After December 31, 2026, the agreement may be terminated by either party upon 90 days' notice.”
https://www.sec.gov/Archives/edgar/data/1181412/000162828026...
It’s only to boost the IPO price. The agreement will last only a few months on paper. I doubt it is a real transaction.
Either that or SpaceX is permanently turning xAI's assets into a neocloud because xAI itself has no traction.
The whole thing looks rather desperate. I wonder what SpaceX's margins are on these contracts.
SpaceX has recently started pitching itself as an orbital datacenter company.
If you buy into that business model (or pretend to), it makes sense for SpaceX to start selling compute early. Their "earthside compute" clients of today are "skyside compute" clients of tomorrow.
A part of Musk's old pitch for Starlink was: space-based solar makes perfect sense for powering space assets, and no sense whatsoever for powering Earth assets. So you have to find a way to use that power in space to do something economically useful. Comms were the only scalable way to do that, so Starlink it was.
I can see how space-based datacenters would follow the same logic. If SpaceX can make them economical, that is. There's no guarantees of that - but if anyone at all can make space-based datacenters economical, it's SpaceX.
> if anyone at all can make space-based datacenters economical, it's SpaceX
Let's hope burning ten thousand tons of toxic e-waste annually in upper atmoshphere never becomes economical. Or mankind gets to senses and bans externalizing your e-waste problem by burning in atmosphere...
Why would it ever be more economical to put datacenters in orbit, rather than on some dirt cheap land?
There are no NIMBYs in space. No government permitting on land use. And solar power is plentiful. It's like having a dollar store Dyson sphere.
Making use of that is predicated entirely on being able to put a lot of hardware into space cheaply. SpaceX is the undisputed best at that, no one comes close. The question is whether their "best" is good enough to make space datacenters economical.
The data link between earth and space has so much bandwidth.
There are sensors in space that send data to earth it gets processed and then the data is sent back to space then to the end user back on earth. If you do the compute in space you save the space-earth transfer time twice. Latency and availability of bandwidth are both factors.
There may be limited utility for this outside of military.
Because dirt cheap land usually does not have dirt, cheap water or dirt cheap electricity.
Water in orbit: famously cheap.
Ah yes: computation. Famous for annihilating water. Every bit you flip consumes an H2O molecule.
Well, how do you cool servers in space then?
Evaporative cooling is the way it happens down on earth - and that shuttles h2o molecules from dense useful clumps like aquifers and rivers to a less useful form spread out in the air. But evaporating h2o isn’t an option in space afaik - since there’s a shortage of air to take up the h2o. In fact I think radiative cooling is the only actual option in space.
Space-based datacenters simply won't work. That people are talking about them shows Musk is the greatest snake oil salesman the world has ever seen.
Can anyone explain how the thermals will work? One of the biggest challenges on Earth is cooling the data center, and it's at least as challenging in space.
The earthbound equivalent would be strapping each chassis to the back of a dedicated solar panel and having the panel double as a giant heat sink. The problem is that doesn't work on the surface due to (at least) rain, the day/night cycle, and the cost of real estate.
Thermals are one among many really big challenges that require costly solutions.
> won’t work
A datacenter (earthbound or space) itself is a fantastical idea until a mix of events and inventions made it feasible to build them to sell compute.
You think the military can’t or won’t dump billions into this to make killing people with drones more effective?
It’s a engineering challenge not impossible.
There are asteroids with concentrations of precious metals more valuable than earth's entire economy. Why don't we just send up spaceships to mine them and send the haul back to earth? What country would say no to free money?
After all, it's just an engineering challenge, not impossible.
The numbers on that are at least somewhat questionable. Even ignoring that you'd crash the market (thus it's not actually worth what it first appears to be) what is the total fuel cost to adjust the orbit of the target asteroid to land the entire thing back on the earth? Because that's what you're doing bit by bit as you shuttle loads of ore back.
Now if you have space based manufacturing or fuel production on the other hand ...
When I hear space I think "that's the perfect location for a data center", since data centers are lightweight, small, require little power, don't need human intervention, have lifetimes measured in decades and don't have to reject heat. Since space easily satisfies these requirements, space is an ideal deployment location for data centers.
Yeah... What am I missing? Like why isn't this just laughed at when it's proposed?
I felt the same way about the "tube with an air hockey table in it." But here I am fifteen years later eating crow as I take the hyperloop to Vegas.
Isn't the Vegas Loop just a car tunnel? As far as I know, there aren't any actual hyperloops[1] involved, just a narrow highway, even if they deceivingly brand it "Loop".
https://en.wikipedia.org/wiki/Hyperloop
It seems off at first glance but actually appears to work out if you do the math. You can model a solar panel as a flat, opaque rectangle. You can calculate power generation and equilibrium temperature for it based on surface area. If you require additional radiative surface area to achieve the desired equilibrium temperature you can place a flat triangle orthogonal to and behind the solar panel in its shadow.
Compute is "free" at that point because waste heat is coming out of the total energy flux which was already accounted for (because we modeled it as opaque).
Of course swapping out the equipment poses a bit of a challenge. The "helping hands" rate is entirely unaffordable and wait until you see this new DC's physical access policies. 0/10 would not rack with them again.
This may be one of the rare instances where the sarcasm is obvious without using the sarcasm font
This is all just the typical Elon hate. What's desperate about getting paid $920,000,000 per month? If that's desperation, I'd love to start groveling more!
Given extreme supply constraints, it's very unlikely that Google or Anthropic will just suddenly cancel right after the IPO unless their own demand collapses. And even if this were true, what value would that provide Musk? Could you imagine if your newly public company suddenly received termination notices from your two largest compute customers? Disaster.
Try logic.
I have no love or hate for Elon Musk. I wish him luck with his space endeavours.
What's desperate is announcing a temporary (allegedly) doubling of revenues days before an IPO that has been criticised for being overpriced at 93 times sales.
These data centers were supposed to serve xAI. Now suddenly they get rented out to others. Why the sudden change of plans?
It's either an emergency accounting gimmick or the effective shutdown or repurposing of xAI.
It’s a repurposing of xAI to be a commodity service provider during a crunch for that commodity. It would be dumb if xAI had any quality or market traction, but they have neither, so it’s actually a rational fallback. But it writes off any high margin future in favor of low margin scale.
And once the compute crunch is over, they’ll have a lot of overprovisioned data centers with no business to soak up the capacity.
Didn't Anthropic pull the same in both ways? you pull me up I pull you up kind of deal? Sounds like SpaceX bought themselves some time up to Q4, which is not the case of Anthropic and even worse for OpenAI. Not counting that none of them got their S&P500 fast-track ticket.
why would Google help a competitor like that, though?
Google (Alphabet) owns 6% of SpaceX which they bought for $12B in 2015. They want to maximize the value of their investment.
How is Google competing with SpaceX?
If you look at the IPO filings you’ll see that Spacex as we know it is just a small part of the expected revenue generator. It is supposedly Grok and AI, hence Google competitor.
I can’t believe serious people use Grok. It has to be propped up by Twitter usage/Musk fans right? It really strikes me as the worst one.
They’re both AI companies
All companies are now AI companies. Just like a while ago all companies were suddenly Ads companies. The entire tech sector is one big FOMO - once you reach certain scale you do exactly the same thing as everyone else.
I'd push back slightly — not on the conclusion, but on the reasoning. There's a simpler explanation that accounts for the same observations.
One of which has more capacity and wants even more, one of which has less capacity and is renting it out.
In the way that Michael Jordan and myself are both basketball players
One is an ad company the other a lifestyle venture?
In the sense that if you want to sell anything whatsoever today it must an AI story.
The article mentions Google is heavily invested in it.
Maybe common investors want to sell stocks to retail
They are not. The amount of conspiracy in high ranked HN comments for AI companies is insane.
Feels like these IPOs are thankfully the top coming before the AI crash and we get back to the real world.
One can hope, but that sentiment is quite unpopular on HN.
It's all about the money.
What exactly is SpaceX's core business?
Their satellite internet business is the only thing which makes them money, which is enabled by their orbital launch business which is as of right now not profitable and I have no idea of if it ever will be but without it they would not be able to launch enough satellites.
Their stupidity with AI and buying X mostly seems to be about scamming investors to make Musk even richer. Like this particular deal is just them doing what CoreWeave does at a SpaceX valuation.
Launch isn't profitable simply because ongoing Starship R&D is eating into it. A lot of opex, capex, and pre-revenue.
If they start running Starship anywhere near the way they do Falcon 9, it'll flip into profits. A lot of big bets SpaceX made ride on Starship coming online. I'm honestly surprised Starlink is already so profitable without it.
One of their big named bets includes: orbital datacenters. Which puts this specific deal into perspective.
80% of the space launch business is putting starlink satellites into orbit, so it's all internal funny money. They could very well be letting the space launch business take losses to make the satellite internet business look better (only profitable part of the whole thing).
Wasn't starship supposed to be funded by the NASA contract?
Orbital datacenters sound like a dangerous bet. I couldn't think of a worse place for a lot of delecate electronics.
Boomers and luddites won't let them be built on earth so no other option really
I can understand this being a move to increase valuation, but I can't connect with the stupidity and scamming investors argument.
Sorry, I was unclear. With that I did not talk about this particular deal. This particular deal seems sane. XAI built more compute that they can use themselves since Grok is not very successful so to not just have the hardware standing there they rent it out to competitors. Makes total sense.
It is other things Musk has gone with Twitter and SpaceX which are shady.
I'm pretty sure xAI is just Musk throwing a tantrum after being played for a fool by Lying Sam.
Smoking crack and investment fraud.
With a light sprinkling of space.
Harvesting energy from the convulsions of people who got tds / tiktok psychosis during covid
Its main business is connectivity. Starlink generated over 10B last year.
Becoming a broader infrastructure company with xAI.
That's only about 35% more than the main telecom operator here Belgium (Proximus: $7.2B revenue in 2025, $2.5B market cap, positive earnings for 15+ years).
Obviously Starlink can and will growth. I'm just pointing out how insane the market cap is, when compared to similar scale "connectivity" businesses.
> Starlink generated over 10B last year
An entire one-hundredth of their proposed valuation!
Yeah, crazy for a company with nothing but the largest civilian satellite network and what amounts to a monopoly on space flight.
Starlink terminals are popular, they put them on drones to avoid jamming (Starling jamming exists but not that easy for now). It might be their sales are inflated due to its use at war.
A datacenter that also provides connectivity/Internet
This feels actually like a pretty safe bet for Google, they secure the compute in case it works (I doubt that the described volume will be available in the near future), while if SpaceX doesn’t manage to provide there is not much loss. I see it more as another way of blowing up SpaceX valuation on paper…
I knew GCP was third banana but what is even happening?
[dupe] Discussion: https://news.ycombinator.com/item?id=48417490
If you can't buy DRAM, you gotta rent your compute infrastructure.
Sorry, what?
Does this mean that SpaceX are the only company that really did build some datacenters to put all the million of GPU/TPU/whatever they all talk about everyday?
I mean, Google, Amazon, Meta and Microsoft told investors they spent more than $1B per day last year in CapEx... why on Earth do they (well, Google and Anthropic at least) need to rent compute to SpaceX, of all companies?
They overbuilt capacity for grok but no one wants to use grok for several reasons
Other companies built data centers but also built products that soak up their data center capacity.
xAI built data centers, and products that are mostly good for nonconsensual porn and confirming a small group’s biases. So they have a lot of excess capacity, and might as well rent it to the adults.
Scarcity. It's becoming difficult to plan for new data centers. They will rent where capacity is available. Grok hasn't gain the expected popularity.
No, CoreWeave for example also rent compute to the big AI companies. This likely just means Grok has few users so they need to rent their extra capacity to their competitors.
Plus it’s not like some absolutely enormous data centers, only 300MW.
Yes but someone will be along shortly to defuse what sounds like giving the bad mars man credit where it’s due. Like everything else he does that works out, it was just luck, timing, actually a mistake that worked out, or someone else behind the scenes that he got lucky in hiring at the right time (by accident).
People with access to enormous wealth tend to get a lot of chances at the betting table.
If he’s so smart why isn’t grok using all that capacity?
Building excess capacity from the start and selling it for a billion a month to constrained competitors. I only wish I could be so dumb.
If your dad had owned an emerald mine I am sure you could also have been that dumb.
But to be more serious: It is impossible to say if this is good or bad for XAI without more numbers. What if they bought their compute way over market price and sell it at a loss?