Starcloud Raises $250M for Orbital AI Data Centers
Starcloud raised a $250 million Series A extension on August 21, 2026, at a $2.3 billion valuation to develop AI data centers in orbit, according to TechCrunch, GeekWire, and SiliconANGLE. Manhattan West led the financing, bringing the Redmond, Washington startup's total funding to $450 million. Reporting identifies manufacturing expansion, Nvidia collaboration, and launch-capacity procurement as uses for the new capital.
Starcloud raised a $250 million Series A extension on August 21, 2026, at a $2.3 billion post-money valuation to support its effort to deploy AI compute infrastructure in orbit. TechCrunch, GeekWire, SiliconANGLE, and Dealroom report that Manhattan West led the round, bringing the Redmond, Washington startup's total capital raised to $450 million since its 2024 founding.
The investor group includes existing backers Benchmark, EQT, Soma, NFX, and 776, along with new investors Nvidia, Cisco Investments, Cedar Capital, Goanna Capital, and Standard Capital, according to GeekWire. The funding extends Starcloud's March Series A round.
CEO and co-founder Philip Johnston told TechCrunch that the company needs to secure substantial launch capacity. "We can see what's coming, we're going to need to book an enormous amount of launch," Johnston said. He also told the publication that Starcloud had requested FCC permission to operate 88,000 spacecraft.
Manufacturing, GPUs, and launch access
GeekWire reports that Starcloud is building production lines for its Starcloud-3 spacecraft at a new 100,000-square-foot manufacturing facility in Woodinville, Washington. Unite.AI, citing the company announcement, reported that the proceeds are allocated to manufacturing capacity, engineering work with Nvidia, and procurement of future launch allocations.
Nvidia's participation builds on an earlier in-orbit hardware test. GeekWire and SiliconANGLE report that Starcloud launched an Nvidia H100 GPU aboard a satellite in November 2025. Starcloud reported using that hardware to train a model called NanoGPT; SiliconANGLE characterized the run as the first orbital AI-training run.
Johnston said in the company announcement, as quoted by Unite.AI: "Last November we put the first NVIDIA H100 in orbit. Today this fresh capital empowers us to build the infrastructure to launch many more of NVIDIA's most advanced GPUs into space."
According to GeekWire, future satellites are intended to use Nvidia's Space-1 Vera Rubin Module, which Nvidia has said provides 25 times the in-space compute capability of an H100. That is a vendor performance claim rather than an independently reported on-orbit result.
An unproven infrastructure thesis
Public reporting frames launch access as a central constraint. TechCrunch reports that Starcloud is pursuing a future Starcloud-3 spacecraft intended to fly on SpaceX's Starship, while Johnston cited tightening launch availability and SpaceX's reported Falcon 9 retirement timeline as reasons to book capacity. The same report notes that Starship remains unproven for this role and that competing launch systems are not yet flying regularly or are still in development.
SiliconANGLE reports that Starcloud has described a sequence of spacecraft generations, including Starcloud-2, which it said would carry AI chips, storage, and a backup module for training and inference. The outlet reported a 2027 launch target for that vehicle, while Unite.AI reported a later-2026 target based on a March release. Starcloud's current public timeline therefore appears inconsistent across reports.
Starcloud has claimed that orbital systems can use solar power and reduce the need for terrestrial cooling infrastructure. Those propositions remain dependent on satellite thermal management, reliability, communications bandwidth, launch economics, and regulatory approvals at the scale proposed. Companies pursuing comparable space-compute concepts face a very different systems-engineering problem from terrestrial AI infrastructure, where rack deployment and grid interconnection, rather than launch manifests and orbital operations, are usually the binding constraints.
For ML infrastructure teams, the immediate significance is not a new generally available compute platform. It is the size of the financing and Nvidia's direct participation in a still experimental category whose commercial viability depends on hardware qualification, launch cadence, and cost per usable GPU-hour.
Key Points
- 1Starcloud's $250 million extension values orbital AI compute at $2.3 billion, giving an unproven infrastructure category substantial new capital.
- 2Nvidia joined the round after Starcloud's H100 orbital test, while future spacecraft are reported to target Vera Rubin-based compute.
- 3Comparable orbital-compute projects depend on launch availability, thermal engineering, bandwidth, and regulatory approval before they can compete with terrestrial GPU clusters.
Scoring Rationale
The financing is unusually large for an experimental AI infrastructure company and includes Nvidia, making it notable for practitioners tracking long-horizon compute supply. The sources describe an unproven category with substantial technical and launch dependencies.
Sources
Public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems