SpaceX Reveals Orbital AI Data Center Plan, Deployment Targeted for 2028
Taylor Wilson
SpaceX has publicly detailed Starmind, a constellation of orbital AI data-center satellites slated for 2028 deployment — its first concrete plan to move AI compute from the ground into space, sidestepping rising power costs, water limits, and local political resistance.
Why move a data center into orbit?
Ground-based data centers face a triple bottleneck: rising electricity costs, water-use restrictions, and local political pushback against new facilities.
This means → the rate at which AI compute can scale is being slowed by the rate at which buildings can get approved and powered.
SpaceX's answer: change the playing field. In orbit, solar energy is free, the vacuum of space handles cooling, and no municipal permits are required.
What exactly is Starmind?
SpaceX released a video describing Starmind — a constellation (a network of satellites working in coordination) of orbital AI compute satellites, with deployment planned from 2028.
Musk said the AI satellite is "actually much simpler than a Starlink satellite." In plain terms = the core is a GPU compute board inside a satellite shell, without the complex phased-array antennas Starlink needs.
This is SpaceX's first public disclosure of a specific plan — with a timeline — to extend AI compute into space.
What gives SpaceX the credibility to attempt this?
Starlink currently operates roughly 10,000 satellites, serves over 10 million users, generates billions of dollars in annual revenue, and runs at a profit margin above 60%.
This means → SpaceX has already proven it can manufacture, launch, and operate satellite constellations at scale — that industrial capability transfers directly to AI satellites.
SpaceX also disclosed it has deorbited 260 satellites in recent months — close to Amazon's entire in-orbit fleet of over 300. Put simply = SpaceX's discards alone nearly match a major competitor's full constellation.
Does it already have an AI compute business?
SpaceX completed its merger with Musk's AI company xAI in February this year and already holds ground-based AI compute assets.
It currently leases compute capacity to Google and Anthropic at a scale of billions of dollars per year.
This reflects that SpaceX is not starting from zero — it is already a player in the AI compute market. The orbital data center is about moving an existing business one layer up.
What is the biggest open question?
Whether orbital AI compute can launch on schedule by 2028 is the first test — between the video and actual orbit, engineering validation and launch scheduling remain hard gates.
The commercial viability is equally uncertain: whether per-unit cost of space-based compute can compete with ground facilities is not yet publicly documented.
In plain terms = the concept is compelling and the timeline is tight. The ultimate question is whether SpaceX can convert its advantage in satellite manufacturing costs into a price advantage in selling compute.
Content is for reference only, not financial advice.