Elon and SpaceX are targeting 1000 starship launches in 2028 and 3000 in 2029 and 8000 in 2030. Elon talked about rough general goals for AI data center in space 1 GW next year, 10 GW in 2028, 100 GW in 2029 the plan and rough goals they are working towards. This is aggressive because the first GW will need 100-200 STarship launches. And they need 100 STarship launches in 2027 for communications and 500 launches for comm (internet and DTC).
New starlink terminals, stated a few hundred million will be made which translate to minimum TAM target of 300-400 million high speed internet customers. $200 billion per year in subscriptions.
I added the math . A few hundred million starlink terminals, $150-200 billion per year from high speed internet if it is 400 million customers.
Starlink terminals production and customer sales determines how much $500-1000 per year customers they have for Starlink. Adding about 1 million per month and going to 1.5 million per month. Going to 3 million per month and then 6 million per month would enable 36 million new customers in 2027 and then 72 million in 2028 or 2029. 100 million total customers in 2028 would be $50 billion per year in Starlink revenue.
They are looking to launch Rubin and TPUs and Trainium and other chips.
Watch @ElonMusk provide a technical update on SpaceX’s capability to manufacture, launch, and operate AI satellites at scale → https://t.co/PSCyWrNsOg pic.twitter.com/vhtr46uax7
— SpaceX (@SpaceX) June 8, 2026
Elon talked about rough general goals for AI data center in space 1 GW next year, 10 GW in 2028, 100 GW in 2029 the plan and rough goals they are working towards. This is aggressive because the first GW will need 100-200 STarship launches. And they need 100 STarship launches in 2027 for communications and 500 launches for comm (internet and DTC). Each GW of AI in space would be about $50 billion per year in revenue if it is renting out Rubin chips.
Brian Wang is a Futurist Thought Leader and a popular Science blogger with 1 million readers per month. His blog Nextbigfuture.com is ranked #1 Science News Blog. It covers many disruptive technology and trends including Space, Robotics, Artificial Intelligence, Medicine, Anti-aging Biotechnology, and Nanotechnology.
Known for identifying cutting edge technologies, he is currently a Co-Founder of a startup and fundraiser for high potential early-stage companies. He is the Head of Research for Allocations for deep technology investments and an Angel Investor at Space Angels.
A frequent speaker at corporations, he has been a TEDx speaker, a Singularity University speaker and guest at numerous interviews for radio and podcasts. He is open to public speaking and advising engagements.
Looks like these satellites will have about $5m in GPUs and $500k in non GPU cost. That’s about $3.33 per watt in non-GPU cost. Mix in launch costs of $100 kg/orbit ($15m/launch) to $200 kg and SpaceX has a 2X advantage in non GPU costs vs terrestrial.
Plug your own numbers in here: https://andrewmccalip.com/space-datacenters
Then the questino becomes, how many satellites and launches to get to 100 GW in orbit. Its 667k satellites in orbit, and 75 satellites/launch equal 8.9k launches. If launched evenly over 5 years, that’s 4.9 launches per day. With their 5 launch towers either under construction or active, that’s 1 launch per day per tower.
It’s all do-able.
Getting 100GW of chips is actually the hardest part. Terrafab will take time.
But with non-GPU costs at 50% of terrestrial, and $2b/month/per GW (Anthropic/Google deals), 100GW can be rented for $2.4/trillion per year. The numbers are nutty.
I think SpaceX at $2.4 trillion per year revenue in 10-20 years is entirely possible – just on AI data center rentals. This ignores all their other opportunities.
In my lifetime, SpaceX could become a $100t valuation company.
The Kessler syndrome is the biggest showstopper.
One incident in any orbit takes out everything in less than a year. All it takes is a freak collision, a solar storm taking out a bunch of sats, sabotage, whatever. After such an incident, space is inaccessible “forever” unless we invent some cleaning method. This will be very difficult since we can’t launch anything.
SpaceX probably has an emergency protocol so they can delete Starlink with a simple command. That will remove most of the mass in LEO and not add that to the cascade. It will also help keeping the lowest orbits open in case of a Kessler syndrome, which means they can keep launching. Instead of going bankrupt, they can help saving humanitys access to space by launching cleaning hardware for a few decades.
I can imagine it will be a really bad day at the office having to press than “rm -r /*”-button for Starlink in a hurry when there is a lack of reliable information.
the starlink satellites are in very low orbit. If all satellites lost power at the same time the orbits will decay and the satellites will burn up in the atmosphere is in a couple of years. at most. The low orbits starlink uses are self cleaning. In fact each starlink satellite uses ion thruster to stay in orbit.
“Elon and SpaceX are targeting 1000 starship launches in 2028.”
That is about 1.5-2.5 years from now. We will be lucky if we get 50 of them in 2028.
This year we had 1 Starship launch and it is June. In 2025 there were 5 of them.
Should have 5 operational towers by 2028, and full re-usability though maybe not rapid re-usability, and two Starship production lines going. 50 launches is less than 1 launch per month per tower. Seems doable. 1000 launches seems unlikely. That’s about 1 every 2 days.
I was also reading that in 2026 we will have multiple Starship launches towards Mars with thousands of Teslabots on them,… It seemed doable?
What heat problem for Space data centers? Every spacecraft has to manage heat. The ISS has about 600 m^2 of radiator area for 120kw of solar. This deck shows SpaceAI compute Sats as having about 110 m^2 radiators for 150kw of solar. The ISS both generates heat from other activities and has to operate at lower less efficient temperatures because of the humans onboard. The AI chips are designed to run hot to optimize heat radiator efficiency.
Both energy and cooling for AI compute in Space can be designed for mass produced engineered solutions that scale perfectly. Sunlight and thermal radiation are very constant and predictable in Space but on Earth are a specific new design problem for every location.
It’s so bizarre that, of all things, people focus on heat dissipation as a deal-breaker for this. This is the one part of the whole system that can TRIVIALLY be calculated and simulated ahead of time. This is like 19th century physics and some geometry. It’s not hard.
Anyone that thinks that no one at SpaceXAI has bothered to calculate their heat dissipation, is just massively stupid and reddit-brained.
It reminds me of people who say that Mars shouldn’t be terraformed because it would lose its atmosphere after a million years.
So wait what did I miss? Do they think they’ve resolved the heat problem for space data centers? O_o Because that would be amazing!
You missed a basic education in physics, as well as common sense.