Very frustrating that criticism shifts to adding an extra comma to his wealth. Zero change will happen over maximum allowed commas.
The bigger issue is the complete fraud of SpaceX valuation. Terafab is not something they have compentence for, and space datacenters are simply more expensive, and cannot replace GPUs after 5 years obsolescence. Being angry at Musk’s wealth is pointless. Distracts from the direct fraud and theft of Nasdaq Index shareholders, and those who don’t time their dumping when indexes are forced to buy.
Well, they brought Intel on for the Terrafab so depending on how involved they are, the knowledge they need is there. SpaceX and Tesla know how to build big vertically integrated factories as well, so the combo could work.
The satellites very well could be cheaper given land costs, power issues, water issue, cooling issues, NIMBY issues etc etc. They already mass produce satellites cheaply, and these satellites are more simple than starlinks other than they have a lot more solar panels and a bigger radiator.
The issues with the satellites being successful completely hinges on starship succeeding and being rapidly reusable. If they cant get to that point, then the satellites will cost too much to launch.
The 5 year GPU obsolescence isnt an issue, they’re only going to have a 5ish year lifespan anyway before deorbiting.
If starship works, the data center dishes will work, but just because they work, doesnt mean there’s enough people out there to use the AI inference compute they want to launch. I think thats where its going to fail. Not enough people want to use the pricey AI, and these crazy valuations assume people will. Its the same problem at all the AI companies.
All that being said, the IPO was still crazy overpriced.
$30m launch costs seem a floor. Very high employee count. Higher maintenance than planes per flight, Expensive rockets. Very complex attachment/detachment from solar+radiators that would last longer than 5 years (apparently they scrap solar/radiators every 5 years because that is too complex)
AI 1 is 125kw. $2.5m would be building allocation. $625k-$1.5m would be solar power costs with battery. All in US. Cheaper elsewhere. lasts 30+ years = 6 launches. Nebraska farmland would be under $1000/year (4 acres) for 1mw of solar, batteries, datacenter. Desert land in NV $100/year.
At well over 2x the cost for a space datacenter, you need to charge 2x the rate to terrestrial competition. But its much more than 2x more expensive. 150kw of space solar is $4.5m. 18x more expensive yearly, though even at 1/10th cost, 1.8x more expensive yearly. Radiators, shielding another $600k, so adding to best case solar, 3.3x more expensive/year. Comparing launch costs to building to get just 2x price difference, needs $1m/satellite(with solar/radiators) launch costs which is the just fuel costs.
I didn’t include the minimal land lease requirement for fully green earth datacenter. An unsustainable aspect of current space law is that SpaceX gets free squatter’s rights for first to take valuable Sun Synchronous Orbit, including using the most convenient wasteful overuse of prime solar facing angles that denies space to all others. In it’s current wasteful design, if a 1m wide window was taxed at $1000, they would pay $70k/year in taxes to UN/global UBI fund. If they redesigned to a 1x500m solar array, they’d pay $1000 in taxes, but leave room for them and others to fill the rest of SSO space. The value of that space fully saturated could fund $47B/year of something earth people need.
I think you’re estimates for launch costs are off. Starship is targeting $50-$200/kg or lower if they are successful. In the short term, you’re $30 million is probably more correct or maybe even to low, but they will very likely lose money while they iterate on reusability and get the costs down to their expected levels.
It’s going to take a lot of launches to get Starship to the point of reusability where the $50-200/kg comes true, but they need to launch something to help recoup those costs which is what starlink/ai will be. I’m not convinced the AI dishes will generate as much revenue as they think though and that it will still be a loss.
The current spec’d ship can theoretically do 100T+ but that’ll be expanded, but that puts all in launch costs and staffing and refurbishment etc at around $5-20 million per launch. You don’t sell it for $50/kg if you can’t eventually support the whole business at $50/kg.
Assuming they can fully utilize the 100T with these dishes that’s ~50 DC Satellites per launch as currently designed, or ~6250Kw compute and 7500Kw of solar per launch. (edit: for reference, its going to be 60 starlink v3, so if physical size isn’t a limiting factor, just 10 less dishes)
I don’t know where you’re getting $4.5 million for 150kw of solar. Space will be more expensive, but it’s $1.1-$1.5w in the USA right now at large commercial scale on land, and SpaceX will be making their own so there’s no profit margin on that. You’re putting it at 20x-27x the cost of land solar after someone takes a profit.
I have no idea what radiators will cost, but they figured it out for Starlink v3, this will just be a lot bigger. (edit: i do expect it to cost less than the solar array though, it should be simpler)
I think we might get a lot more details on what things could actually cost, if we can get any leaked or real info about what a Starlink V3 satellite costs, which maybe now that they’re public, that will come out once they start launching? They might obfuscate the per dish cost though by grouping it with other things?
I like your idea of some future taxing of usage in space. As it gets more crowded, more money will need to be spent on monitoring it, and coordinating things all of which is an ongoing cost, and taxing it yearly/per dish to help fund research and such would be great.
Starship is targeting $50-$200/kg or lower if they are successful.
$30m is $200/kg for full starship capacity (150T). AI1 is volume constrained.
where you’re getting $4.5 million for 150kw of solar. Space will be more expensive, but it’s $1.1-$1.5w in the USA right now at large commercial scale on land
$4.5m is standard space GaS solar cost. SpaceX does claim $3000/w. $450k. 1MW solar in US is under $1m. Add 16 hour battery 2mwh = $200k. At $1m space equipment before GPUs, is $6m equivalent terrestrial costs due to 5 year replacement. Actually 5x the cost. You can also cut earth costs by over half by installing in China or Mongolia with the same latency to US as from space.
Very frustrating that criticism shifts to adding an extra comma to his wealth. Zero change will happen over maximum allowed commas.
The bigger issue is the complete fraud of SpaceX valuation. Terafab is not something they have compentence for, and space datacenters are simply more expensive, and cannot replace GPUs after 5 years obsolescence. Being angry at Musk’s wealth is pointless. Distracts from the direct fraud and theft of Nasdaq Index shareholders, and those who don’t time their dumping when indexes are forced to buy.
Well, they brought Intel on for the Terrafab so depending on how involved they are, the knowledge they need is there. SpaceX and Tesla know how to build big vertically integrated factories as well, so the combo could work.
The satellites very well could be cheaper given land costs, power issues, water issue, cooling issues, NIMBY issues etc etc. They already mass produce satellites cheaply, and these satellites are more simple than starlinks other than they have a lot more solar panels and a bigger radiator.
The issues with the satellites being successful completely hinges on starship succeeding and being rapidly reusable. If they cant get to that point, then the satellites will cost too much to launch.
The 5 year GPU obsolescence isnt an issue, they’re only going to have a 5ish year lifespan anyway before deorbiting.
If starship works, the data center dishes will work, but just because they work, doesnt mean there’s enough people out there to use the AI inference compute they want to launch. I think thats where its going to fail. Not enough people want to use the pricey AI, and these crazy valuations assume people will. Its the same problem at all the AI companies.
All that being said, the IPO was still crazy overpriced.
$30m launch costs seem a floor. Very high employee count. Higher maintenance than planes per flight, Expensive rockets. Very complex attachment/detachment from solar+radiators that would last longer than 5 years (apparently they scrap solar/radiators every 5 years because that is too complex)
AI 1 is 125kw. $2.5m would be building allocation. $625k-$1.5m would be solar power costs with battery. All in US. Cheaper elsewhere. lasts 30+ years = 6 launches. Nebraska farmland would be under $1000/year (4 acres) for 1mw of solar, batteries, datacenter. Desert land in NV $100/year.
At well over 2x the cost for a space datacenter, you need to charge 2x the rate to terrestrial competition. But its much more than 2x more expensive. 150kw of space solar is $4.5m. 18x more expensive yearly, though even at 1/10th cost, 1.8x more expensive yearly. Radiators, shielding another $600k, so adding to best case solar, 3.3x more expensive/year. Comparing launch costs to building to get just 2x price difference, needs $1m/satellite(with solar/radiators) launch costs which is the just fuel costs.
I didn’t include the minimal land lease requirement for fully green earth datacenter. An unsustainable aspect of current space law is that SpaceX gets free squatter’s rights for first to take valuable Sun Synchronous Orbit, including using the most convenient wasteful overuse of prime solar facing angles that denies space to all others. In it’s current wasteful design, if a 1m wide window was taxed at $1000, they would pay $70k/year in taxes to UN/global UBI fund. If they redesigned to a 1x500m solar array, they’d pay $1000 in taxes, but leave room for them and others to fill the rest of SSO space. The value of that space fully saturated could fund $47B/year of something earth people need.
I think you’re estimates for launch costs are off. Starship is targeting $50-$200/kg or lower if they are successful. In the short term, you’re $30 million is probably more correct or maybe even to low, but they will very likely lose money while they iterate on reusability and get the costs down to their expected levels.
It’s going to take a lot of launches to get Starship to the point of reusability where the $50-200/kg comes true, but they need to launch something to help recoup those costs which is what starlink/ai will be. I’m not convinced the AI dishes will generate as much revenue as they think though and that it will still be a loss.
The current spec’d ship can theoretically do 100T+ but that’ll be expanded, but that puts all in launch costs and staffing and refurbishment etc at around $5-20 million per launch. You don’t sell it for $50/kg if you can’t eventually support the whole business at $50/kg.
Assuming they can fully utilize the 100T with these dishes that’s ~50 DC Satellites per launch as currently designed, or ~6250Kw compute and 7500Kw of solar per launch. (edit: for reference, its going to be 60 starlink v3, so if physical size isn’t a limiting factor, just 10 less dishes)
I don’t know where you’re getting $4.5 million for 150kw of solar. Space will be more expensive, but it’s $1.1-$1.5w in the USA right now at large commercial scale on land, and SpaceX will be making their own so there’s no profit margin on that. You’re putting it at 20x-27x the cost of land solar after someone takes a profit.
I have no idea what radiators will cost, but they figured it out for Starlink v3, this will just be a lot bigger. (edit: i do expect it to cost less than the solar array though, it should be simpler)
I think we might get a lot more details on what things could actually cost, if we can get any leaked or real info about what a Starlink V3 satellite costs, which maybe now that they’re public, that will come out once they start launching? They might obfuscate the per dish cost though by grouping it with other things?
I like your idea of some future taxing of usage in space. As it gets more crowded, more money will need to be spent on monitoring it, and coordinating things all of which is an ongoing cost, and taxing it yearly/per dish to help fund research and such would be great.
$30m is $200/kg for full starship capacity (150T). AI1 is volume constrained.
$4.5m is standard space GaS solar cost. SpaceX does claim $3000/w. $450k. 1MW solar in US is under $1m. Add 16 hour battery 2mwh = $200k. At $1m space equipment before GPUs, is $6m equivalent terrestrial costs due to 5 year replacement. Actually 5x the cost. You can also cut earth costs by over half by installing in China or Mongolia with the same latency to US as from space.