Starcloud has reached a valuation of $1.1 billion in its latest funding round, making it one of the fastest startups to achieve unicorn status after its time at Y Combinator. The Series A round was led by Benchmark and EQT Ventures and closed just seventeen months after the company’s demo day. This investment underscores the growing interest in moving data centers to orbit, a shift motivated by resource limitations and political hurdles slowing terrestrial development. However, this business model relies on technology that is not yet proven and requires very significant capital.
To date, Starcloud has raised a total of $200 million. The company launched its first satellite, equipped with an Nvidia H100 GPU, in November 2025. A more powerful version named Starcloud 2 is scheduled for launch later this year. It will carry multiple GPUs, including an Nvidia Blackwell chip and an AWS server blade, along with a bitcoin mining computer.
Looking further ahead, the company plans to develop a dedicated data center spacecraft called Starcloud 3. This spacecraft is designed to launch on SpaceX’s Starship rocket. It would be a three-ton vehicle producing 200 kilowatts of power, built to fit within the “pez dispenser” system SpaceX uses to deploy Starlink satellites. CEO Philip Johnston believes this could be the first orbital data center cost-competitive with Earth-based facilities, targeting power costs around five cents per kilowatt-hour, but this depends on launch costs falling to approximately $500 per kilogram.
A significant challenge is that Starship is not yet operational. Johnston anticipates commercial access opening in 2028 or 2029. This timeline highlights a central reality for all major space data center projects: advanced space computing will remain prohibitively expensive until a new generation of rockets achieves frequent, low-cost launches, a scenario that may not materialize until the 2030s. Johnston stated that if Starship is delayed, the company will continue launching smaller versions on SpaceX’s Falcon 9 rocket, but acknowledges they will not be competitive on energy costs until Starship flies regularly.
Johnston outlines two potential business models. The first is selling processing power to other spacecraft in orbit; their first satellite already analyzes data for Capella Space’s radar satellites. The second, for the future when launch costs decrease, involves powerful distributed orbital data centers pulling workloads from terrestrial counterparts.
This industry is in its earliest stages. When Nvidia CEO Jensen Huang recently unveiled the company’s Vera Rubin Space-1 chip modules, he did not mention that none have been produced or shared with partners. The number of advanced GPUs in orbit is currently in the dozens, compared to the nearly 4 million Nvidia sold to terrestrial data centers in 2025. Furthermore, SpaceX’s Starlink network, the largest in orbit with 10,000 satellites, generates roughly 200 megawatts of power. In contrast, over 25 gigawatts of data center power capacity are under construction on the ground in the U.S. alone.
Johnston contends Starcloud is ahead of competitors, having deployed the first terrestrial-grade GPU in orbit. That GPU was used to train an AI model and run a version of Gemini, which the company claims as a first. Beyond performance, Johnston says the mission provided crucial data on operating powerful chips in space. He admits an H100 may not be the ideal space chip, but it was chosen to prove that state-of-the-art terrestrial hardware could function there. This knowledge, hard-won after another GPU failed during launch, will guide future designs.
Substantial technical hurdles remain, including efficient power generation and cooling the intense heat from the chips. The Starcloud 2 spacecraft will feature the largest deployable radiator ever flown on a private satellite, and Johnston expects at least two more iterations of that design will launch.
Synchronization presents another major challenge. The largest data center workloads, like AI model training, require hundreds or thousands of GPUs working in tandem. Achieving this in space would demand either enormous single spacecraft or extremely reliable, high-power laser links between multiple satellites flying in formation. Most companies in this field expect these complex training workloads will follow long after simpler inference tasks are established in orbit.
Other companies developing space data center businesses include Aetherflux, Google’s Project Suncatcher, and Aethero, which launched Nvidia’s first space-based Jetson GPU in 2025. A significant potential competitor is SpaceX itself, which has sought government permission to operate a million satellites for distributed computing in space.
Competing directly with SpaceX is a formidable prospect, but Johnston sees room for multiple players. He suggests SpaceX is building for a different primary use case, likely focused on serving its own Grok and Tesla workloads. While SpaceX may eventually offer third-party cloud services, Johnston believes it is unlikely to become what Starcloud aims to be: a dedicated energy and infrastructure provider in space.