Starcloud Raises $250M at $2.3B Valuation as Nvidia Joins the Orbital Data Center Bet

Starcloud has raised $250 million to accelerate its orbital data center constellation, bringing Nvidia into the financing alongside lead investor Manhattan West and a new group of strategic and institutional backers. The Series A extension doubles the Redmond, Washington startup's valuation to $2.3 billion and brings total capital raised since its 2024 founding to $450 million. This is no longer a pitch about putting a GPU in space. It is a bet that orbital compute can become infrastructure — and that the company building the first recognizable version should be capitalized like one.

The financing arrives with unusually concrete evidence behind it. Starcloud's 60-kilogram Starcloud-1, launched by SpaceX in November, became the first satellite to run Nvidia's H100 GPU in low Earth orbit. The company says it used that spacecraft to train an AI model and run a version of Google's Gemini in space. Its next step is a 450-kilogram Starcloud-2 rideshare mission planned for January, designed to generate about eight kilowatts — roughly 100 times Starcloud-1's power. The long-term target is Starcloud-3, a three-ton, 200-kilowatt-class spacecraft aimed at AI inference and training workloads uplinked from Earth.

The Round at a Glance

What the New Capital Actually Funds

Starcloud says the additional capital will support manufacturing capacity, rocket-launch procurement, and engineering work with Nvidia. Those three uses reveal where the company's bottleneck has moved. Starcloud-1 proved that an orbital spacecraft can host a high-end terrestrial GPU; the next challenge is no longer a laboratory demonstration but a production and deployment system. Manufacturing must move from one-off spacecraft to repeatable vehicles, launches must be reserved years ahead, and the compute payload must be integrated with the power, thermal, radiation, and communications systems that turn a GPU into a usable cloud service.

Nvidia Is More Than an Investor Here

Nvidia's participation changes the interpretation of the round. A financial investor is underwriting Starcloud's market and execution; Nvidia is also helping shape the compute stack that makes the market possible. In March, Nvidia unveiled the Space-1 Vera Rubin Module, a system intended to bring more powerful AI processing to satellites and future orbital data centers. The company says the module could deliver up to 25 times more AI compute than the H100. Starcloud's partnership gives that module an early orbital proving ground and gives Starcloud access to the vendor whose hardware defines the AI infrastructure cycle on Earth.

That is strategic alignment rather than simple sponsorship. Nvidia needs new demand for increasingly capable accelerators; Starcloud needs a roadmap that keeps its orbital computers relevant as terrestrial GPU generations advance. Cisco's participation adds a second layer: networking and data movement. An orbital data center is not a satellite with a chip bolted on. It is a distributed network node that must accept workloads, move results, synchronize with customers, and operate despite narrow launch windows and intermittent ground links. The investor list is beginning to resemble an infrastructure stack.

The Roadmap Is a Climb in Orders of Magnitude

Starcloud's roadmap is legible because each vehicle represents a different business question. Starcloud-1 asked whether a high-performance GPU could operate in orbit and whether a model could be trained there. Starcloud-2 asks whether the company can scale spacecraft mass and power enough to host meaningful workloads. Starcloud-3 asks whether orbital compute can become a service with the density and capacity of a data center rather than an impressive satellite experiment.

The order-of-magnitude jumps are the opportunity and the risk. A 60-kilogram demonstration and a three-ton production platform share a brand but not an engineering program. Power generation, heat rejection, radiation tolerance, launch integration, ground networking, and operations all become different problems at each step. The new capital buys Starcloud the ability to attack those problems in sequence; it does not make the sequence disappear.

The Launch Constraint Hiding Inside the Funding Story

The most important sentence in the announcement may be the one about launch availability. SpaceNews reports that Falcon 9 rideshare reservations are unavailable beyond late 2028 or early 2029 as SpaceX prepares to transition toward Starship. That creates a strange constraint for an orbital data center company: the capital may be ready, the GPUs may be ready, and the customers may be ready, but the manifest is not. Starcloud's plans rely heavily on Starship, the much larger reusable rocket SpaceX is developing, while SpaceX itself is pursuing up to one million orbital data center satellites after its acquisition of xAI.

This is both supplier risk and market validation. A launch bottleneck proves there is enough demand to consume capacity, but it also means every orbital compute startup is exposed to the same schedule. Starcloud has to procure launches early, develop hardware compatible with more than one vehicle where possible, and ensure each spacecraft has enough commercial life to justify the wait. A constellation plan measured in thousands of satellites is meaningless if the first hundred cannot reach orbit on a predictable cadence.

The Customer Thesis

Starcloud aims to run cloud workloads in space for early customers such as Crusoe, an AI infrastructure provider. The service proposition is not that every workload belongs in orbit. It is that some workloads benefit from a different location: satellite imagery can be processed near the sensor, large AI training jobs can draw on continuous solar power, and data that is expensive or slow to downlink can be reduced to useful insights before it reaches Earth. The customer pays for compute capacity and data movement, not for the novelty of a satellite.

• Earth-observation operators that want to turn raw imagery into decisions before downlink
• AI infrastructure providers seeking additional power and compute capacity outside constrained terrestrial grids
• Defense and intelligence customers that value resilient, geographically distributed processing
• Spacecraft operators that need shared compute instead of placing a full accelerator stack on every satellite
• Model developers testing workloads where orbital latency, coverage, or data custody creates an advantage

The BlacKnight Take

Nvidia's participation makes Starcloud's round a sector signal, not just a company financing. The AI infrastructure market is beginning to treat orbit as a possible new deployment layer, and Starcloud has the strongest early demonstration: an H100 actually ran in space, a larger mission is on the roadmap, and the capital now exists to build beyond prototypes. But the valuation is underwriting a chain of unproven conversions — from GPU to spacecraft, spacecraft to constellation, constellation to service, and service to margins.

The cleanest way to track the thesis is not the 88,000-satellite filing or the $2.3 billion mark. It is the staircase: Starcloud-2 launches on schedule, delivers materially more useful compute than Starcloud-1, converts a named customer into recurring workload revenue, and reserves enough launch capacity to build a fleet. If those milestones arrive, Nvidia's check will look like the opening move in a new infrastructure category. If launch cadence or power economics break the staircase, this round will be remembered as the moment the market priced the dream before it priced the service.

Frequently Asked Questions