Starcloud Wants to Put Data Centers in Orbit

Data Centers in Orbit are moving from conference-stage pitch to factory floor, as a small American startup raises fresh funding to mass-produce satellites that process information above the clouds. Starcloud, founded by former engineers from a satellite internet venture and a European aerospace firm, has now attracted backing from some of the biggest names in chips and networking, betting that the hardest problems facing earthbound computing can be solved by leaving the ground altogether. The pitch is simple to state and hard to execute: sunlight is constant in space, heat can be radiated into the void, and no one needs to find land, water or a grid connection up there.

According to Data Center Dynamics, the latest round was led by a New York investment firm and joined by chipmaker Nvidia, networking giant Cisco Investments, and several venture firms, with a Seoul-based investment arm participating as the only Korean backer. The round values the company at well over two billion dollars and takes its total backing to approaching half a billion dollars since its founding two years ago. That is a striking vote of confidence for a company that still employs only about two dozen people, and it signals that investors now treat orbital computing as a serious infrastructure bet rather than a curiosity.

Why space looks attractive to chip makers

Earthbound data centers face three bottlenecks that have grown tighter as artificial intelligence demand has surged: securing a grid connection, finding suitable land, and obtaining cooling water. Waiting lists for power can stretch for years in some regions, neighbours object to new facilities, and drought-prone areas ration the water that cooling systems consume. In orbit, a satellite in a dawn-to-dusk path can collect sunlight almost continuously, without clouds, night or weather, and waste heat can be shed by radiators into deep space instead of into rivers or cooling towers. The physics is unforgiving in its own way, since every watt must be radiated away and radiators add size and mass, but the constraints are engineering problems rather than planning disputes.

Starcloud has already put hardware in orbit to test the idea. Its first demonstrator, a satellite weighing roughly as much as a large adult, launched late last year carrying a high-end graphics processor, and the team used it to run a lightweight open language model and to analyse radar observations for spotting ships in distress and wildfires. Those early tasks were chosen deliberately: they show that useful work can be done where the data is collected, rather than beaming raw observations to Earth for processing. If that approach scales, satellites that today mostly observe the planet could also interpret what they see, sending down answers instead of floods of raw imagery.

What the new money will build

The fresh capital is earmarked for production, not just prototypes. The company is building out a factory in Washington state to manufacture its third-generation spacecraft, a far larger computing module than the demonstrator, with solar arrays spanning an area comparable to several tennis courts. The long-term ambition is bolder still: a constellation numbering in the tens of thousands of satellites, delivering capacity measured in gigawatts, a scale that would rival a sizeable share of the world's data center fleet. Critics will note the gap between that ambition and a single small demonstrator, and the company still needs regulatory approval for its constellation plans, but the factory is the step that turns a roadmap into a supply chain.

Competition is arriving from several directions at once, which is itself a sign that the sector is maturing. A Chinese computing constellation began regular in-orbit services in late summer, serving dozens of institutions, while a search giant is preparing a test satellite carrying its own custom processors. A rocket company has also filed plans for its own orbital computing system, and lawmakers have begun debating how to evaluate such systems for national security. Starcloud's edge, according to its backers, is focus: it is building only the computing layer, partnering with chip suppliers rather than trying to own rockets, launch pads and ground stations as well. That narrow focus also concentrates its risk, since a delay in chip supply or launch availability hits the whole plan at once.

What readers should watch next

Data Centers in Orbit will not replace the warehouse-sized facilities outside major cities any time soon, and anyone who suggests otherwise is selling a story rather than a schedule. The near-term test is narrower: can the next-generation satellite operate reliably for months, keep its processors cool, and deliver results that customers will pay for? A successful demonstration on that scale would answer the engineering question and leave the business question, namely whether orbital capacity can compete on price with land-based facilities that keep getting cheaper and more efficient. A failure, or a long delay, would cool investor enthusiasm for the whole sector, not just one company.

There are also questions that no funding announcement answers. Space debris grows with every constellation, and tens of thousands of additional satellites would sharpen concerns about collisions and about who gets to use crowded orbits. Energy accounting matters too: launching thousands of tonnes of hardware on rockets carries its own carbon cost, and that bill must be weighed against the water and grid savings claimed for orbital operation. Regulators, astronomers and rival operators will all want a say before any giant constellation is approved, and that process moves at a very different speed from venture funding.

For now, the story is best read as a marker of where computing ambition is heading. A decade ago, the idea of running advanced processors in orbit sounded like science fiction; today it has a factory, a demonstrator, a roster of blue-chip backers and a queue of competitors. Whether Starcloud becomes a major infrastructure provider or a footnote in a larger shift, the bet it represents is now too large to ignore: that the next great data center location may not be a suburb with cheap power, but a thin, sunlit band above the atmosphere. For related technology coverage, see how Japan's Mars moon mission is bringing samples home, why travellers remain wary of handing trips to AI agents, and how overnight recordings are flagging disease risk early.