SpaceX AI Constellation
Thread Starter
Ecce Homo! Loquitur...
Joined: Jul 2000
Aviation Qualifications: Spotter
Posts: 24,662
Likes: 7,348
From: Peripatetic
SpaceX AI Constellation
Undiudtedly linked to their recent joint venture with xAI*.
https://uk.pcmag.com/networking/1628...ta-center-push
SpaceX Eyes 1 Million Satellites For Orbital Data Center Push
Not just thousands. In an FCC filing, the company mentions deploying up to a staggering 'one million satellites' in orbits ranging from 500 kilometers to 2,000km.
* https://www.msn.com/en-us/money/comp...ai/ar-AA1VmhES
BREAKING: SpaceX is requesting to launch and operate a constellation of 1 million satellites with unprecedented computing capacity (orbital data centers) to power advanced AI, according to a new FCC filing.
SpaceX: "Launching a million satellites that operate as orbital data centers is a first step towards becoming a Kardashev II-level civilizationone that can harness the Sun's full power-while supporting Al-driven applications for billions of people today and ensuring humanity's multi-planetary future amongst the stars."
In the SpaceX filing:
SpaceX aims to deploy a system of up to one million satellites to operate within narrow orbital shells spanning up to 50 km each (leaving sufficient room to deconflict against other systems with comparable ambitions).
System will operate between 500 km and 2,000 km altitude and 30 degrees and sun-synchronous orbit inclinations.
SpaceX plans to design and operate different versions of satellite hardware to optimize operations across orbital shells.
System will rely nearly exclusively on high-bandwidth optical links for communications. These optical links will route traffic within the network and to satellites in the Starlink constellation, via its high capacity (petabit) and high reliability laser mesh, which in turn will transmit traffic to authorized earth stations on the ground.
SpaceX added: "With Starship's ability to deliver unprecedented tonnage to orbit for AI compute, the capacity for intelligence processing in space could surpass the electricity consumption of the entire U.S. economy, without the immense cost and disruption of rebuilding Earth's strained electrical grid to support the explosive demand for data centers. In turn, satellites that function as solar-powered orbital data centers are the most cost-effective, energy-efficient, and environmentally sound way to build infrastructure to meet accelerating demand for Al-enabled goods and services.
With the inherent efficiencies of deploying solar powered data centers and launch cost rapidly decreasing due to the development of the Starship launch vehicle, SpaceX will be able to cost- effectively scale up its constellation as demand increases and compute evolves.
For instance, launching 1 million tonnes per year of satellites generating 100 kW of compute power per tonne would add 100 gigawatts of AI compute capacity annually, with minimal ongoing operational or maintenance needs."
SpaceX: "Launching a million satellites that operate as orbital data centers is a first step towards becoming a Kardashev II-level civilizationone that can harness the Sun's full power-while supporting Al-driven applications for billions of people today and ensuring humanity's multi-planetary future amongst the stars."
In the SpaceX filing:
SpaceX aims to deploy a system of up to one million satellites to operate within narrow orbital shells spanning up to 50 km each (leaving sufficient room to deconflict against other systems with comparable ambitions).
System will operate between 500 km and 2,000 km altitude and 30 degrees and sun-synchronous orbit inclinations.
SpaceX plans to design and operate different versions of satellite hardware to optimize operations across orbital shells.
System will rely nearly exclusively on high-bandwidth optical links for communications. These optical links will route traffic within the network and to satellites in the Starlink constellation, via its high capacity (petabit) and high reliability laser mesh, which in turn will transmit traffic to authorized earth stations on the ground.
SpaceX added: "With Starship's ability to deliver unprecedented tonnage to orbit for AI compute, the capacity for intelligence processing in space could surpass the electricity consumption of the entire U.S. economy, without the immense cost and disruption of rebuilding Earth's strained electrical grid to support the explosive demand for data centers. In turn, satellites that function as solar-powered orbital data centers are the most cost-effective, energy-efficient, and environmentally sound way to build infrastructure to meet accelerating demand for Al-enabled goods and services.
With the inherent efficiencies of deploying solar powered data centers and launch cost rapidly decreasing due to the development of the Starship launch vehicle, SpaceX will be able to cost- effectively scale up its constellation as demand increases and compute evolves.
For instance, launching 1 million tonnes per year of satellites generating 100 kW of compute power per tonne would add 100 gigawatts of AI compute capacity annually, with minimal ongoing operational or maintenance needs."
SpaceX Eyes 1 Million Satellites For Orbital Data Center Push
Not just thousands. In an FCC filing, the company mentions deploying up to a staggering 'one million satellites' in orbits ranging from 500 kilometers to 2,000km.
* https://www.msn.com/en-us/money/comp...ai/ar-AA1VmhES
Thread Starter
Ecce Homo! Loquitur...
Joined: Jul 2000
Aviation Qualifications: Spotter
Posts: 24,662
Likes: 7,348
From: Peripatetic
..
This is not just hype for a potential fundraise. Filing with the FCC with plans for 1M AI Inference sats in SSO is in line with the 100GW statements Elon has made several times over the last 90 days.
Weve done a lot of math (100s of hours at this point) around this concept specifically, if you want to dig in when you have an hour or two.
AI Chips for space (most recent): https://research.33fg.com/analysis/e...-gpus-in-space
Radiative cooling math: https://research.33fg.com/analysis/d...e-data-centers
Location for the constellation(SSO): https://research.33fg.com/analysis/w...y-benchmarking
Energy cost for Orbital compute: https://research.33fg.com/analysis/o...-earth-by-2030
Feel free to ask Grok to read this for you. They are ungated and it does a great job with the math.
Weve done a lot of math (100s of hours at this point) around this concept specifically, if you want to dig in when you have an hour or two.
AI Chips for space (most recent): https://research.33fg.com/analysis/e...-gpus-in-space
Radiative cooling math: https://research.33fg.com/analysis/d...e-data-centers
Location for the constellation(SSO): https://research.33fg.com/analysis/w...y-benchmarking
Energy cost for Orbital compute: https://research.33fg.com/analysis/o...-earth-by-2030
Feel free to ask Grok to read this for you. They are ungated and it does a great job with the math.
Joined: Jul 2013
Aviation Qualifications: Non-Aircrew
Posts: 5,679
Likes: 3,341
From: Everett, WA
Given the well known risks that a satellite collision presents to the large number of satellites in low earth orbit (doesn't matter much if it's intentional or accidental - once there are a few collisions, the resultant chain reaction will make those orbital altitudes unusable for many years)
I'd rather see him apply those really smart people in Space X to the problem of orbital debris.
I'd rather see him apply those really smart people in Space X to the problem of orbital debris.
Thread Starter
Ecce Homo! Loquitur...
Joined: Jul 2000
Aviation Qualifications: Spotter
Posts: 24,662
Likes: 7,348
From: Peripatetic
Given the well known risks that a satellite collision presents to the large number of satellites in low earth orbit (doesn't matter much if it's intentional or accidental - once there are a few collisions, the resultant chain reaction will make those orbital altitudes unusable for many years)
I'd rather see him apply those really smart people in Space X to the problem of orbital debris.
I'd rather see him apply those really smart people in Space X to the problem of orbital debris.
SpaceX has developed a novel Space Situational Awareness (SSA) system, called Stargaze, that significantly enhances the safety and sustainability of satellite operations in low Earth orbit (LEO), and its screening data will be made available to the broader satellite operator community free of charge in the coming weeks.
Practices—such as leaving rocket bodies in LEO, operators maneuvering their satellites without sharing trajectory predictions or coordinating with other active satellites, and countries conducting anti-satellite tests—have heightened the risk of collision, necessitating improvements in space-traffic coordination. Conventional methods typically observe objects only a limited number of times per day, causing large uncertainties in orbital predictions, further compounded by volatile space weather.
Stargaze delivers a several-order-of-magnitude increase in detection capability compared to conventional ground-based systems. Stargaze uses data collected from nearly 30,000 star trackers, each of which makes continuous observations of nearby objects, resulting in approximately 30 million transits detected daily across the fleet.
Thesystem autonomously detects observations of orbiting objects and are then aggregated to generate accurate orbit estimates and predictions of position and velocity for all detected objects in near real-time. These predictions integrate into a space-traffic management platform that identifies potential close approaches between objects in space and generates Conjunction Data Messages (CDMs).
To fully realize the utility of such frequent observations, SpaceX developed this system to provide conjunction screening results within minutes, compared to the current industry standard of several hours…..
Thread Starter
Ecce Homo! Loquitur...
Joined: Jul 2000
Aviation Qualifications: Spotter
Posts: 24,662
Likes: 7,348
From: Peripatetic
Recruiting......
Elon Musk:
We are hiring for many critical engineering roles to develop the technologies for AI satellites in space at our facilities in Austin and Seattle.
Solar, process, automation, manufacturing, mechanical, electrical, optics, software... come build space data centers with great engineers at @SpaceX @Starlink
We are also hiring elite engineers at our new 230 MeV cyclotron facility in Florida, where we are bringing single-event radiation testing in house to accelerate development across all SpaceX vehicles.
Solar, process, automation, manufacturing, mechanical, electrical, optics, software... come build space data centers with great engineers at @SpaceX @Starlink
We are also hiring elite engineers at our new 230 MeV cyclotron facility in Florida, where we are bringing single-event radiation testing in house to accelerate development across all SpaceX vehicles.
Not many companies have cyclotrons!
Thread Starter
Ecce Homo! Loquitur...
Joined: Jul 2000
Aviation Qualifications: Spotter
Posts: 24,662
Likes: 7,348
From: Peripatetic
...............
What will a 100 kW/ton orbital compute satellite look like?
We’ve been building a satellite mass-budget model to show how compute satellites will differ architecturally from Starlink.
Satcom/Starlink baselines allocate ~35 % of dry mass to phased-array antennas, gimbals, and other continuous Earth-pointing mechanisms required for RF routing.
Compute satellites will have no downlink; data moves via laser links to the Starlink constellation. The freed mass is completely reallocated to solar arrays and radiators, shifting us from an RF-routing architecture to a thermodynamic-optimised one.
The triangular bottleneck in orbital compute satellites is the three-way interdependence of solar generation, thermal rejection, and compute capability: any single subsystem can only scale as far as the other two allow, so true power-density gains require all three to advance in lockstep, otherwise one becomes the binding constraint.
Compute sats have no need for ultra-precise Earth-pointing, so solar arrays can be far lighter and flexible, using passive gravity-gradient and centrifugal tensioning in dawn-dusk SSO.
Operating the compute ASICs at ~370 K, exactly as Elon highlighted, exploits the T⁴ scaling of blackbody radiation, boosting net heat rejection significantly, and thereby slashing the required radiator area (and therefore mass) dramatically. We believe radiators must deploy two-sided and sit in the shade of the solar arrays for optimal cold-space view factors, in order to hit 100kW/ton.
At first glance, thermal rejection becomes one of the largest subsystems (~34 % of dry mass). Yet this dominance only appears because projected PV efficiency gains and higher chip temperature enable 7Χ higher power throughput, greater than the expected headroom in radiator efficiency gains. The triangular bottleneck closes cleanly at 100.2 W/kg system power density, almost exactly the line in the sand Elon and SpaceX have drawn.
As finance folks who've had the privilege of learning from leading space-industry engineers, we offer this conceptual take with a grain of salt: these 100 kW/ton satellites would likely look quite different from Starlink satellites...
We imagine compute satellites will have a compact central bus housing only the dense compute payload and laser links, surrounded by large, light, thin-film solar arrays deployed radially like wings or sails, kept taut by passive gravity-gradient and centrifugal tensioning in dawn-dusk SSO. Paired with them will be two-sided deployable radiator fins, deliberately positioned in the permanent shade of the arrays for optimal cold-space view factors.
We are keen to learn more though so please share any suggestions.
Read the full analysis here for all our modelling and charts:
https://research.33fg.com/analysis/t...-power-density
We’ve been building a satellite mass-budget model to show how compute satellites will differ architecturally from Starlink.
Satcom/Starlink baselines allocate ~35 % of dry mass to phased-array antennas, gimbals, and other continuous Earth-pointing mechanisms required for RF routing.
Compute satellites will have no downlink; data moves via laser links to the Starlink constellation. The freed mass is completely reallocated to solar arrays and radiators, shifting us from an RF-routing architecture to a thermodynamic-optimised one.
The triangular bottleneck in orbital compute satellites is the three-way interdependence of solar generation, thermal rejection, and compute capability: any single subsystem can only scale as far as the other two allow, so true power-density gains require all three to advance in lockstep, otherwise one becomes the binding constraint.
Compute sats have no need for ultra-precise Earth-pointing, so solar arrays can be far lighter and flexible, using passive gravity-gradient and centrifugal tensioning in dawn-dusk SSO.
Operating the compute ASICs at ~370 K, exactly as Elon highlighted, exploits the T⁴ scaling of blackbody radiation, boosting net heat rejection significantly, and thereby slashing the required radiator area (and therefore mass) dramatically. We believe radiators must deploy two-sided and sit in the shade of the solar arrays for optimal cold-space view factors, in order to hit 100kW/ton.
At first glance, thermal rejection becomes one of the largest subsystems (~34 % of dry mass). Yet this dominance only appears because projected PV efficiency gains and higher chip temperature enable 7Χ higher power throughput, greater than the expected headroom in radiator efficiency gains. The triangular bottleneck closes cleanly at 100.2 W/kg system power density, almost exactly the line in the sand Elon and SpaceX have drawn.
As finance folks who've had the privilege of learning from leading space-industry engineers, we offer this conceptual take with a grain of salt: these 100 kW/ton satellites would likely look quite different from Starlink satellites...
We imagine compute satellites will have a compact central bus housing only the dense compute payload and laser links, surrounded by large, light, thin-film solar arrays deployed radially like wings or sails, kept taut by passive gravity-gradient and centrifugal tensioning in dawn-dusk SSO. Paired with them will be two-sided deployable radiator fins, deliberately positioned in the permanent shade of the arrays for optimal cold-space view factors.
We are keen to learn more though so please share any suggestions.
Read the full analysis here for all our modelling and charts:
https://research.33fg.com/analysis/t...-power-density
Thread Starter
Ecce Homo! Loquitur...
Joined: Jul 2000
Aviation Qualifications: Spotter
Posts: 24,662
Likes: 7,348
From: Peripatetic
Video
Nvidia CEO Jensen Huang announced today that the company is working on a new chip/computer for orbital data-centers called Nvidia Vera Rubin Space-1.
"It's going to start data-centers out in space. Of course, in space there's no conduction, no convection, there's just radiation, so we have to figure out how to cool these systems out in space, but we got lots of great engineers working on it."
"It's going to start data-centers out in space. Of course, in space there's no conduction, no convection, there's just radiation, so we have to figure out how to cool these systems out in space, but we got lots of great engineers working on it."
Thread Starter
Ecce Homo! Loquitur...
Joined: Jul 2000
Aviation Qualifications: Spotter
Posts: 24,662
Likes: 7,348
From: Peripatetic
Video
Musk:
Video
Announcing TERAFAB: the next step towards becoming a galactic civilization
SpaceXAI + Tesla TERAFAB Project Goal is a trillion watts of compute/year Most must necessarily go to space, as US electricity is only 0.5TW
ELON MUSK: “To reach a petawatt, we’d build an electromagnetic mass driver on the Moon with robots like Optimus and humans. With no atmosphere and 1/6 gravity, payloads can reach escape velocity without rockets, drastically cutting costs. That could scale compute 1000Χ beyond a terawatt—I hope we live to see it.”
This would roughly be the scale of the Tesla/SpaceX TERAFAB compared to Tesla’s Giga Texas factory, which is already one of the largest buildings in the world. We are talking about an unprecedented amount of scale.
Musk: “Yeah, 100M sq ft is the right order of magnitude”
Musk: “Yeah, 100M sq ft is the right order of magnitude”
Thread Starter
Ecce Homo! Loquitur...
Joined: Jul 2000
Aviation Qualifications: Spotter
Posts: 24,662
Likes: 7,348
From: Peripatetic
That's just TerFb, not SpaceX or any of his other businesses.
Not sure at what stage Musk becomes the first trillionaire - but it can't be long....
Not sure at what stage Musk becomes the first trillionaire - but it can't be long....
What is the Present Value of TERAFAB?
TSMC and Samsung make about $2 billion in revenue per GW of AI chips they produce. Nvidia makes about $10 billion and AMD makes about $6 billion per GW of AI chips/systems they sell.
Let's assume that Tesla's TERAFAB economics are similar to TSMC and Samsung - so $2 billion per GW of AI chips. Scale this up to 1TW and we're looking at about $2 trillion of revenue.
Now let's assume that Tesla is able to do this in 2035 - nine years from now - and that this $2 trillion per year doesn't grow.
Let's assume a 50% profit margin. Let's also assume a 50% chance of failure - so there's a 50% chance of zero revenue & profits. If we use a 15% discount rate, what's the present value of this?
Roughly about $1 trillion.
TSMC and Samsung make about $2 billion in revenue per GW of AI chips they produce. Nvidia makes about $10 billion and AMD makes about $6 billion per GW of AI chips/systems they sell.
Let's assume that Tesla's TERAFAB economics are similar to TSMC and Samsung - so $2 billion per GW of AI chips. Scale this up to 1TW and we're looking at about $2 trillion of revenue.
Now let's assume that Tesla is able to do this in 2035 - nine years from now - and that this $2 trillion per year doesn't grow.
Let's assume a 50% profit margin. Let's also assume a 50% chance of failure - so there's a 50% chance of zero revenue & profits. If we use a 15% discount rate, what's the present value of this?
Roughly about $1 trillion.
Thread Starter
Ecce Homo! Loquitur...
Joined: Jul 2000
Aviation Qualifications: Spotter
Posts: 24,662
Likes: 7,348
From: Peripatetic
A billion here, a billion there, sooner or later you're talking real money.....
SPACEX / TERAFAB:
The County of Grimes, Texas, will be home to SpaceX's Multiphase, next gen, vertically integrated semiconductor manufacturing and advanced computing fabrication facility.
Estimated capital for initial phase is $55 billion. Estimated total investment of $119 billion.
The agreement covers the SpaceX Reinvestment Zone No. 1 2026-001 located at Gibbons Creek Reservoir and surrounding areas.
A public hearing will be held June 3rd.
The County of Grimes, Texas, will be home to SpaceX's Multiphase, next gen, vertically integrated semiconductor manufacturing and advanced computing fabrication facility.
Estimated capital for initial phase is $55 billion. Estimated total investment of $119 billion.
The agreement covers the SpaceX Reinvestment Zone No. 1 2026-001 located at Gibbons Creek Reservoir and surrounding areas.
A public hearing will be held June 3rd.
Thread Starter
Ecce Homo! Loquitur...
Joined: Jul 2000
Aviation Qualifications: Spotter
Posts: 24,662
Likes: 7,348
From: Peripatetic
Slightly rearranged for clarity.
https://x.com/SawyerMerritt/status/2...611420273?s=20
https://x.com/SpaceX/status/2064099405758906727?s=20
SpaceX has just officially unveiled its AI1 satellite, the first generation of its AI satellite.
Overall Specs:
150 kW peak compute payload
120 kW average compute payload
70 kW per ton
Compute provider interchangeable
Architecture:
Centralized compute module
Large deployable solar arrays
Deployable liquid-radiator thermal management system
AI-focused compute satellite design ("AI1 satellite")
Dimensions:
Wingspan: 70 meters
Deployed height: 20 meters
Thermal System:
110 m² deployable liquid radiators
Redundant pumping loops
Integrated micrometeoroid shielding
Solar Power System:
150 kW solar array
250 W/m²
SpaceX-manufactured solar technology from Bastrop, Texas
Elon: "The AI satellite is much simpler than a Starlink satellite. The AI satellite is essentially a lot of solar cells, you still need some laser links, but you don't have all of the super complex antennas that you have on a Starlink satellite.
"The easier one to design for is the AI satellite. It's bigger. A lot of this is technology we've already made with the Starlink V3 satellites."
https://x.com/SpaceX/status/2064099405758906727?s=20
SpaceX has just officially unveiled its AI1 satellite, the first generation of its AI satellite.
Overall Specs:
150 kW peak compute payload
120 kW average compute payload
70 kW per ton
Compute provider interchangeable
Architecture:
Centralized compute module
Large deployable solar arrays
Deployable liquid-radiator thermal management system
AI-focused compute satellite design ("AI1 satellite")
Dimensions:
Wingspan: 70 meters
Deployed height: 20 meters
Thermal System:
110 m² deployable liquid radiators
Redundant pumping loops
Integrated micrometeoroid shielding
Solar Power System:
150 kW solar array
250 W/m²
SpaceX-manufactured solar technology from Bastrop, Texas
Elon: "The AI satellite is much simpler than a Starlink satellite. The AI satellite is essentially a lot of solar cells, you still need some laser links, but you don't have all of the super complex antennas that you have on a Starlink satellite.
"The easier one to design for is the AI satellite. It's bigger. A lot of this is technology we've already made with the Starlink V3 satellites."
