Orbital solar data centers' cost-per-watt of compute is $51.10/W, far above $15.85/W for terrestrial sites; SpaceX is uniquely positioned to try orbital
It might not be rational But It might be physically possibleBefore we get nerd sniped by the shiny engineering details, ask the only question that matters.
The key issue is not simply access to solar power, but whether launch, hardware, and operations can overcome a large cost gap. A later broader examination of the space-data-center case similarly centered the trade-offs around power supply and costs.
First-order effects
At $51.10/W versus $15.85/W on Earth, orbital compute carries a little more than three times the cited cost per watt, making terrestrial deployments the economic default on this comparison.
SpaceX's distinctive position is strategic rather than proof of viability: its ability to attempt the model separates it from companies that can discuss orbital compute but lack the same apparent path to operating it.
The cost spread raises the bar for Blue Origin and other potential entrants: they would need a clearer advantage in deployment, energy availability, or system integration to justify pursuing a higher-cost compute platform.
Third-order effects
If orbital compute advances, the sector is likely to be shaped by vertically integrated launch-and-infrastructure operators rather than by data-center developers alone; lower launch and operations costs would be the critical gating variable.
The comparison reinforces a broader possibility that AI capacity constraints will be addressed first through terrestrial energy and site innovation, with orbital systems remaining a specialized option unless their full-system economics change materially.
The trend: AI infrastructure is increasingly a contest over dependable powered capacity, with space-based compute competing against faster-to-deploy terrestrial energy solutions.
Solar-powered AI compute in space sounds cool, but after running the numbers, it's probably cheaper to mine a small piece of the sun and power hyperscale data centers on the moon.
Really interesting! Love the conversation around this. Few differences between ours: - Mine does $/PFLOP instead of $/W - Takes into account Moore's Law, exhibits compounding cost reduction - Takes into account Musk's law of launch costs collapsing - Uses dedicated radiators
great article everything is awesome when they r backed with logic however, once we make satellite data centers work economically, the amount of innovation that paved the way will most likely outweigh the cost entirely e.g by cutting launch cost, improving solar cell
will see which one is more expensive, gravity or bureaucratic inertia in 10 years, assuming first mover advantage is significant. maybe elon's doge experience changed his cost/value model.
The motivation for data centers in space isn't primarily economic. It is sovereignty. Beaming AI support to a handheld device from space is a memetic weapon of godlike magnitude. Space based hard power is still a necessity but you'd win the infowar outright.
Excellent post, thanks for doing the work 🤩 Ya, give musk a huge stack 💰 in SpaceX options grants. I linger in the “...If it's an existential humanity play, that's cool too, but it's a slightly different game.” 👍
The unfortunate truth here for Andrew: We now know that if we yap hard enough, he will be forced to build another beautiful, first-principles model of a frontier technology. [image]
“If you've built an empire, the best possible use of it is to burn its capital like a torch and light up a corner of the future. Pay for the iteration loops. Build the cathedrals. This is how we advance civilization.”
“No one cares about your Loro Piana. If you've built an empire, the best possible use of it is to burn its capital like a torch and light up a corner of the future ... Build the cathedrals. This is how we advance civilization.” @andrewmccalip Fantastic model, Andrew, towards
Wow incredible example of a casual interactive essay. Of course AI makes this kind of essay way easier to write, wow, what a brilliant use of AI. Magic Ink is finally becoming available to everyone.
People are putting out so much superb analysis along with tools to work out the feasibility of space based data centers. I played around with the variables in the calculator below. If you assume Starship can get launch costs below $100kg and the satellite power costs drop by
Solid snapshot analysis but I think it omits the main point. The stated assumptions include, “No adjustments for permitting or regulatory delay.” But that's the whole reason** to build in space. My analysis also shows it's not economically viable today, and it is the regulatory
I think this is a solid and comprehensive model, I appreciate you open sourcing it Andrew. Solar PV maxing out at 24% efficiency is something I'll have to adjust in my own model but just a minor change I would make in this all around great project. I also agree that SpaceX is the
This is a good analysis, it agrees with others that a space token is currently 2-3x more expensive than a ground token. But it misses a broader point. As long as SpaceX can turn a profit on their space tokens, they have effectively unlimited demand for launch and own a parallel
Building Datacenters... Terrestrial sites have 3.5GW heat sink ceilings, and so many NIMBYs that permit authorities are starting to block them. Space is 100 miles from NIMBYs but your heat sink is behind a vacuum, and you need to lift everything out of a gravity well. I wonder
ok time to chime in. @andrewmccalip this model is miles better than anything else I've seen on here. kudos for the transparency. That being said, in the pursuit of a more accurate analysis, there are a few fundamental oversights in the model. After fixes (default case) ->
“If you've built an empire, the best possible use of it is to burn its capital like a torch and light up a corner of the future.” This goes so hard. @_dylanwheeler hold me accountable to this.
✨DATACENTERS IN SPACE!! ✨ Here's the maths and costs. Not impossible. SpaceX/Teska/xAI/Elon is the only conglomerate that can do it. 4x as expensive as earth-based. Knock-on effects from a literal sci-fi space-future might make bank.
Came for rational first-principles modeling of the Shiny New Thing. Stayed for the philosophical bangers. “We should be actively goading billionaires into irrational, high-variance projects. ... The best possible use is to burn capital and light up a corner of the future.” [image…
Great fighting words: “Space is hard, and shipping flight systems is a credibility filter. I'm annoyed at everyone else. The conversation is full of confident claims built on one cherry-picked fact and zero arithmetic.” — @andrewmccalip
Data centers in space. It might not be economically rational. But it might be physically possible. I'm trying to bring some quantitative structure to a conversation that's been mostly big-number vibes. https://andrewmccalip.com/... [image]
From a quick skim this is about the best thing I've read about orbital data centers yet. Suppose orbital data centers happen as described here. What happens to near-term heavy-lift launch pricing? #linklog
Good blog on cost/physics of Orbital data centers under optimistic assumptions. — Result: it's feasible, need to come down in cost by a factor of four to even approach being price competitive
Economics of Orbital vs Terrestrial Data Centers by Andrew McCalip — It might not be rational — But — It might be physically possible — andrewmccalip.com/space-datace... [image]