Elon Musk wrote on X on 31 October 2026 that “Orbital compute deployed per month probably exceeds terrestrial around the end of 2028 and then far exceeds it in 2029”. In the same post he wrote: “Terrestrial compute will be peanuts compared to orbital compute”.
SpaceX’s annual rollout target in its S-1 IPO filing, Trending Topics reported, is 100 gigawatts of compute a year into orbit. That is 5 times the 20 GW of SI capacity Musk said SpaceX could potentially have online by the end of 2027, and about 7.7 times the 13 GW below which he would be surprised, by our arithmetic.
Short explainer · Musk on orbital compute
A vertical news Short summarising this story. Figures come from Musk’s posts, a Trending Topics report on SpaceX’s S-1 and our 20 October report; our arithmetic is labelled.
Business Pill · Neoclouds
A short explainer on neoclouds and who owns AI compute. It teaches the general idea only and says nothing about orbital compute.
The key insight: As we read the posts and the S-1 report, Musk’s prediction rests on launch capacity: SpaceX’s reported target of 100 GW of compute a year into orbit needs about one million metric tons a year, about 452 times the roughly 2,213 tons it reached in 2025, by our arithmetic.
The post was a reply to another account and had about 385,000 views when we read it. Musk gave no quantities in it. He compared orbital compute deployed per month with terrestrial compute, and put the crossover “around the end of 2028”.
It is a prediction, worded with “probably”. About 26 months separate the post from the end of 2028, by our arithmetic.
Musk’s end-2027 range for SpaceX’s SI capacity (surprised if less than 13 GW; 20 GW possible but unlikely, from the 20 October 2026 call) against the S-1 target of 100 GW of compute sent to orbit each year, as reported by Trending Topics. Ratios are our arithmetic.
What Was Posted on X
@SPCX100T Terrestrial compute will be peanuts compared to orbital compute.
Orbital compute deployed per month probably exceeds terrestrial around the end of 2028 and then far exceeds it in 2029.
Elon Musk’s post of 31 October 2026, a reply to another account: “Orbital compute deployed per month probably exceeds terrestrial around the end of 2028 and then far exceeds it in 2029.” This piece quotes it and sets it against SpaceX’s reported S-1 target.
Elon Musk just made a huge prediction: new orbital compute deployed each month surpasses new compute deployed on Earth around the end of 2028, and far exceeds it in 2029
This isn’t a hopeful guess. SpaceX’s own plan is built to hit it
Mass to orbit is the engine. SpaceX launched about 2,200 tons to orbit in 2025, more than 80% of everything humanity sent up
Now its official path takes that to roughly 167,000 tons in 2028, about 740,000 tons in 2029, and over 2 million tons a year by 2032. Elon said SpaceX is projected to hit “a million tons to orbit per year in about three years”
Starship V3 is designed for 100 tons per launch, with 200 tons expected as soon as V4. Launch cost per kilogram has already fallen from about $18,500 to $2,700 on Falcon 9, and Starship targets a cut of more than 99%
Every ton becomes compute. SpaceX at a million tons a year of SI satellites and 100 kW per ton adds 100 gigawatts of orbital SI computing capacity every year, with near-continuous solar power and radiative cooling in space. No grid queues, no land fights, no cooling towers
For scale, SpaceX’s entire ground fleet is expected to pass ~3 GW this year
That’s why Elon is so confident. Earth’s data centers are capped by power plants and permits. Orbit is capped only by how fast you can launch, and SpaceX is the only one launching at this scale
A post by @XFreeze quoting Musk, with a year-by-year path of tons to orbit. The piece attributes its 167,000-ton and 740,000-ton figures to XFreeze, not to SpaceX.
The Plan Behind It
SpaceX’s S-1 filing set “100 gigawatts of compute power per year into orbit” as its annual rollout target, Trending Topics reported on 21 May 2026. The plan involves “approximately one million metric tons of payload annually”, carried by “thousands of launches per year”, with first deployments “as early as 2028”.
The same report says SpaceX assumes “satellites carrying more than 100 kW of compute power per metric ton”, and that SpaceX reached “around 2,213 metric tons of mass to orbit” in 2025. A million tons a year is about 452 times that, by our arithmetic. The report adds that Starship V3 is designed for 100 tons, with later generations meant to double it.
The 2027 Comparison
On SpaceX’s 20 October third-quarter call, Musk said SpaceX could potentially have 20 GW of SI capacity online by the end of 2027, though he called that unlikely, and that he would be surprised if it deployed less than 13 GW of compute capacity, as we reported. The call figures did not split that capacity between the ground and orbit. SpaceX is also building Colossus 2, which Musk said will hold about 1.2 million Blackwell GPUs by the end of this year.
A post by @XFreeze that quoted Musk gives a year-by-year path of about 167,000 tons to orbit in 2028 and about 740,000 tons in 2029. We did not find those two figures in SpaceX’s filing coverage, so they are XFreeze’s, not SpaceX’s.
The Business Lever: Launch Capacity
As we read the Trending Topics report, the S-1 plan ties orbital compute to tons launched, at more than 100 kW per metric ton, carried by thousands of launches a year. Google is testing the same idea with TPUs in orbit through Project Suncatcher.
The Structural Read
As we read Musk’s post, the comparison is about the rate of new compute each month, not the total installed, and it is worded as a probability.
As we read our 20 October report, Musk’s 13 to 20 GW range for the end of 2027 is SpaceX’s SI capacity; the call figures did not split it between the ground and orbit.
As we read the Trending Topics report, the S-1 plan converts tons into compute at more than 100 kW per metric ton, which is how one million tons a year maps to about 100 GW a year.
“Terrestrial compute will be peanuts compared to orbital compute.”
Three Implications
FOR SPACEX INVESTORS Trending Topics reported that the S-1 puts first orbital compute deployments as early as 2028.
FOR TERRESTRIAL DATA-CENTRE BUILDERS Musk’s post puts the monthly crossover around the end of 2028; our Colossus 2 report has SpaceX at about 1.2 million Blackwell GPUs on the ground by the end of this year, per Musk.
FOR RIVAL LAUNCHERS AND CLOUDSGoogle is testing TPUs in orbit through Project Suncatcher, our earlier report shows.
The framework says “The structural bottleneck isn’t computing, it’s electrons.”
As we read it, the S-1 plan as reported by Trending Topics measures orbital compute in gigawatts, the same unit the framework uses for the electricity limit. The framework is the owner’s argument and does not discuss SpaceX’s orbital plan.
Musk did not define “compute” in the post: gigawatts, chips or operations. The S-1 figures here come from a report on the filing, not from the filing text, which we did not read. Trending Topics called orbital data centres an “unsolved physics problem” in its headline.
This Week In The AI Supercycle
Anthropic’s Reality Check, The Capital Squeeze, & The New World Order
This week’s issue: for two years the AI build-out was priced as a growthstory. This week it started being priced as a credit story.
Musk says monthly orbital compute additions probably pass terrestrial ones around the end of 2028. SpaceX’s S-1 target of 100 GW a year into orbit, as reported by Trending Topics, is 5 to 7.7 times the 13 to 20 GW of SI capacity he gave for the end of 2027, by our arithmetic.
A note on sourcing. Facts come from Elon Musk’s post on X, a post by @XFreeze quoting it, Trending Topics’ 21 May 2026 report on SpaceX’s S-1 filing, and our 20 October report on SpaceX’s third-quarter call, all read on 31 October 2026. Ratios and day counts are our arithmetic. We did not contact SpaceX.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
