Starcloud raises 170 million dollars for orbital data centers

Starcloud, a startup selling “compute in orbit,” has raised $170 million in a Series A round led by Benchmark and EQT Ventures, valuing the company at about $1.1 billion, according to TechCrunch. The company has already launched a satellite carrying an Nvidia H100 GPU and says a second spacecraft with multiple GPUs—including an Nvidia Blackwell chip and an AWS server blade—will follow later this year. Its next step is a larger “data center spacecraft” designed around SpaceX’s Starship payload system.

The pitch is less about science fiction than about the growing friction of building data centers on Earth. Power interconnect queues, local permitting fights, and grid-upgrade disputes are slowing new capacity just as AI training and inference demand more electricity and more GPUs. Starcloud’s CEO Philip Johnston argues that orbit can eventually deliver power at around five cents per kilowatt-hour—if Starship launch prices fall to roughly $500 per kilogram and if Starship flies frequently enough to make deployments routine. That conditional matters: the company concedes it cannot be cost-competitive on energy until Starship is operating at high cadence, which it expects only in 2028–2029, and possibly later.

What Starcloud is actually selling today is not a replacement for hyperscale cloud. The near-term business model is servicing other spacecraft: processing sensor data in space instead of downlinking raw data to Earth, which can be bottlenecked by bandwidth and ground-station availability. TechCrunch notes Starcloud’s first satellite analyzes data from Capella Space’s radar satellites, an example where “compute near the sensor” can be valuable even if it is expensive.

The longer-term vision—moving terrestrial workloads off-planet—runs into physics and reliability. Radiation, launch vibration, and thermal management are not minor engineering footnotes; Starcloud itself says an Nvidia A6000 failed during launch, and that running an H100 in space was partly about learning what breaks. Latency also does not disappear: orbit is close, but not local, and many enterprise workloads are built around predictable network performance, not intermittent links and orbital windows.

The comparison with Earth-based buildouts is stark. TechCrunch cites estimates that US data centers with more than 25 gigawatts of power are under construction, while even SpaceX’s 10,000-satellite Starlink constellation produces on the order of 200 megawatts. Starcloud’s planned “Starcloud 3” spacecraft is described as a 200-kilowatt, three-ton platform—meaning the gap between orbital prototypes and terrestrial scale is measured in orders of magnitude, not product iterations.

Starcloud’s funding round is therefore a bet on two timelines moving in lockstep: cheaper launches arriving before terrestrial constraints ease, and orbital hardware becoming dependable before customers demand data-center-grade uptime. For now, the company’s roadmap still depends on a rocket that is not yet flying commercially.

Starcloud says its first H100 satellite launched in November 2025 and that its first cost-competitive orbital data center hinges on Starship access in 2028 or 2029.