Data centers are running into major operating issues as the need for huge amounts of power and water push supplies to the limit. Companies like SpaceX and Google are betting that operating data centers in space could be part of the solution, but the technology needs a lot more work before it can scale beyond a niche, according to analysis from consultancy Altman Solon.
As data center expansion continues with huge construction projects around the world, power queues now stretch up to seven years, and water used for cooling is becoming scarce in many regions. Local communities are increasingly pushing back against new facilities, meaning developers have to factor in significant legal and reputational risks.
Is space the new data frontier?
Altman Solon estimates that European data centers alone could consume around 108 terawatt-hours of electricity by 2030, more than the entirety of the Netherlands currently uses in an entire year. Long-term demand for computing power is expected to outpace the available electricity supply, and that gap is driving companies to look beyond the ground entirely.
That is why launching data centers into space is such a strong potential solution. Orbiting around Earth, these data centers would make use of nearly unlimited solar power, would not need water to cool, could offer in-orbit processing of data for other space use cases, and would avoid the significant carbon footprint associated with terrestrial data centers.
Space data centers would also come with some serious downsides, however: Radiation damage to electronics is a major issue that has yet to be solved, solar power would dip to virtually zero while orbiting through the shadow of the Earth, and making any repairs would be virtually impossible. Experts say scaling space data centers is not realistic in the short-term.
Real moves towards making it a reality
Despite that, the tech industry is still trying to make it happen. This technology has quickly gone from a far-out concept to a reality, in the form of experimental pilot programs. Google’s Project Suncatcher, for example, envisions clusters of solar-powered satellites carrying their own TPU chips. Google is sending a tiny AI data center satellite into space with a SpaceX Falcon 9 rocket on October 1 as a test to better understand how it responds to the harsh conditions.
Blue Origin has a similar effort underway called Project Sunrise, which would rely on more than 51,000 satellites and connect to Earth through the company’s TeraWave network. Meanwhile, SpaceX filed an FCC filing in January to deploy up to one million data center satellites. In March 2026 alone, Elon Musk proposed a chip factory called TeraFab, a project valued between $20 billion and $25 billion meant to produce radiation-hardened chips for satellites.
A new startup, Starcloud, raised $170 million and reached a valuation of over $1 billion, making it one of the fastest companies in Y Combinator’s history to do so. These moves followed SpaceX’s filing with federal regulators to eventually deploy up to a million data center satellites, along with similar projects from Google and Blue Origin.
Why space looks appealing
Orbiting data centers would not replace massive terrestrial facilities outright, but instead form a new, networked layer of computing power. In addition to abundant solar energy, there would also be no need for land permits or lengthy zoning battles either, which alone can account for up to 15% of the cost of building a facility on Earth.
Starcloud expects its approach to cut carbon dioxide emissions by a factor of ten compared with a traditional data center, even after accounting for the environmental cost of a rocket launch.
"Chips in orbit can be operated until true physical end-of-life, avoiding the premature refresh cycles common in terrestrial facilities," adds Anthony Milovantsev, partner with Altman Solon.
"Launch economics could also be cross-subsidized by piggybacking on other space infrastructure, such as Starlink’s growing constellation deployment, treating compute as a marginal addition rather than a standalone launch cost."
Real obstacles remain
The challenges are substantial. Getting rid of heat is a major issue since a modest three-megawatt facility would require a radiator system roughly the size of a football field. Equipment in orbit is also exposed to radiation damage and cannot easily be repaired if something breaks. Every fix for these problems, such as adding shielding or backup batteries, adds weight, which then drives up launch costs further, creating a difficult cycle for engineers to solve.
There is also a wider, systemic problem with launching more and more satellites into space: The growing threat of a collision. There are tens of thousands of satellites in space and the threat of a catastrophic collision grows every year. The SpaceX Starlink fleet routinely performs collision-avoidance maneuvers, but most other satellites do not have any sort of maneuvering capabilities, and there are thousands of old, inoperable satellites adding to the huge amount of space junk in orbit.
Experts say that one major collision could knock out essential telecommunications and navigation tech we rely on for a wide range of everyday activities. Collisions create large fields of debris that can cause more damage to other orbiting satellites, in what can end up being a compounding chain reaction, sometimes called the Kessler Effect.
The cost threshold that matters
Analysts widely agree that launch costs need to fall to around $200 per kilogram or below for orbital data centers to become commercially competitive. Progress has been dramatic, with costs falling by a factor of 20 to 30 over the past three decades.
Today, a rocket like the Falcon 9 costs around $3,000 per kilogram, while older rockets can cost $10,000 to $20,000 per kilogram. SpaceX hopes its Starship rocket will eventually bring costs below $100 per kilogram, and Google has cited that same $200 per kilogram threshold as the point where a shift becomes economically realistic.
The earliest uses will likely involve processing satellite data directly in orbit and sending only useful conclusions back to the ground. Off-planet storage is also being explored to protect data from disasters on Earth. Most analysts expect meaningful commercial viability sometime between 2029 and 2035, assuming launch prices keep dropping.
"The key uncertainty is not technical feasibility – that has been demonstrated at least at small scale. It is launch economics and increasing scale," says Justin Hotchkiss, associate partner at Altman Solon.
"If Starship continues the historical trend of launch cost reduction and delivers on its cost targets, the timeline accelerates. If not, orbital compute remains a defense and space-data niche. Either way, the implications are material enough to warrant consideration today."