Starcloud's $250M round doubles its valuation, but orbital d

Starcloud has raised $250 million at a $2.3 billion valuation to build orbital data center infrastructure, with Nvidia and Cisco Investments joining as new investors. The company announced the round on Friday, with Manhattan West leading the deal and existing backers Benchmark, EQT, and Soma participating. Starcloud frames the capital as fuel for a much larger ambition: 88,000 satellites and 20 gigawatts of orbital compute that would move AI workloads off power-constrained terrestrial grids.

The valuation is the real story inside the headline. Starcloud was worth $1.1 billion in March, when it raised $170 million, so the new price tag more than doubles the company in five months. Cumulative capital since the 2024 founding now stands at $450 million. Some of that acceleration is market-wide: SpaceX's blockbuster IPO has pulled investors back into space infrastructure, and orbital data centers are the newest version of that trade.

The underlying pitch is an energy argument. Terrestrial AI buildouts are hitting limits on grid capacity, land, and water, and Starcloud's answer is to generate power where the compute sits. Manhattan West partner Lauren Selig has joined the board as an observer, a small detail that shows how seriously the lead investor takes the position.

The investor mix deserves attention. Manhattan West led the round, Nvidia and Cisco Investments came in as new names, and Benchmark, EQT, and Soma returned. A chipmaker, a networking company, and a growth investor sharing one cap table points to where the value is expected to accrue: in the silicon, the connectivity, and the operating platform. Having the chip supplier and the network vendor inside the cap table before the first commercial mission is an unusual position for a startup at this stage. The timing is the reverse of the usual vendor playbook, where suppliers wait for revenue proof before taking equity.

What the money buys

The use of funds is concrete. Starcloud says the round will expand manufacturing capacity, deepen the engineering collaboration with Nvidia, and fund procurement for newer products. The centerpiece is a 100,000-square-foot production facility in Woodinville, Washington, built for the next-generation Starcloud-3 spacecraft. That is a commitment to serial production. A company planning to fly 88,000 satellites has to build them on Earth first, and that factory is where the plan becomes physical.

None of this is cheap. Space hardware is manufactured in small batches, tested against extremes, and launched on schedules set by someone else. Starcloud's $450 million cumulative raise is real money, but it is a down payment on a program that will need steady access to capital until the first commercial missions produce revenue. That is one reason the valuation doubling matters as much as the check itself: it resets the terms for every future raise.

Why Nvidia is backing orbital data centers

The strategic core of the round is Nvidia. Starcloud and Nvidia are co-developing the Space-1 Vera Rubin Module, a radiation-hardened version of Nvidia's next-generation AI silicon built to survive extreme orbital conditions. Nvidia has committed roughly $25 million to the program. That is a small check that secures a large position: if orbital compute becomes a real market, Nvidia's chip is already the default platform in it.

The hardware story has receipts. Starcloud-1 flew in November 2025 and put the first data center-grade Nvidia GPU in orbit. The company says it has also run an H100 in orbit and trained a model on that hardware. Two eight-kilowatt Starcloud-2 satellites are booked on 2027 rideshare flights, where they will run inference for customers that include U.S. government agencies. The workload choice is deliberate: near-term missions do inference, which fits what orbital hardware can realistically deliver, while the Vera Rubin Space module is developed for later spacecraft.

Put the capacity target in perspective. Twenty gigawatts is the scale of a multi-year global buildout for a hyperscaler, and Starcloud wants that capacity in orbit instead of on a constrained grid. Even a fraction of that number would make the company one of the larger compute operators anywhere, with the added quirk that its hardware would be circling the planet at orbital velocity.

The government angle is worth pausing on. Agencies contracting for orbital inference suggests workloads where data cannot easily leave the collection point or where terrestrial connectivity is the weak link. For a company with no operating constellation, anchor government demand is also the credibility that later attracts commercial customers.

The launch bottleneck

Here is the gap between the valuation and the physics. Two eight-kilowatt satellites amount to a validation program, far from a working data center. The 88,000-spacecraft constellation and 20 gigawatts of capacity still depend on an FCC application that has not been approved. And the delivery mechanism is the binding question: rideshare slots can carry Starcloud through its 2027 milestones, but no existing rideshare cadence can place 88,000 satellites in orbit in any relevant timeframe.

The math of 88,000 satellites depends on Starship flying commercial missions on a regular cadence. That timeline sits outside Starcloud's control. It hangs on SpaceX's flight rate, payload economics, and regulatory rhythm. Every orbital data center plan I have read in the past two years contains the same sentence, which is why this round reads less like a bet on Starcloud's execution and more like a bet that the launch market matures underneath it.

I read the $2.3 billion as option value. A company whose deployed footprint is one GPU satellite, two booked rideshares, and a factory under construction cannot justify that price on current revenue. Manhattan West, Nvidia, and Cisco are paying for the right to be early in a market that does not exist yet, and the doubling of the valuation in five months is the market's way of saying that the option is getting more expensive.

The skeptical case deserves a hearing. Orbital compute faces open questions: power generation at scale in orbit, thermal management, ground latency, launch cost per kilogram, and a constellation license the FCC has not granted. Plenty of space infrastructure bets have died on these exact problems. But the competitive facts are specific: Starcloud has flown the only data center-grade Nvidia GPU in orbit, trained a model there, and now holds a co-development relationship with the chipmaker itself. That is a concrete moat, and launch risk does not erase it.

The hardest technical question is power. Every watt of compute in orbit requires solar collection, storage through eclipse periods, and thermal rejection in a vacuum, and none of those systems has been demonstrated at the scale Starcloud is describing. The company's answer is that space sidesteps grid constraints and land costs entirely, but the engineering burden of delivering 20 gigawatts of conditioned power in orbit is a problem of a different order.

Consider what Nvidia gets for its $25 million. If orbital compute matures, the Space-1 Vera Rubin Module becomes the reference platform for a market with no installed base today. If it does not, the exposure is trivial for a company of Nvidia's size. That asymmetry is why the chipmaker's participation matters more than the check size. For Nvidia, the $25 million is a cheap option on a future market. For Starcloud, it converts a vendor into a co-developer.

Watch the milestones rather than the headlines. The FCC decision on the 88,000-spacecraft application, the Starship flight rate over the next twelve months, and the two 2027 rideshare missions will each tell you more than the valuation multiple. Those three data points will test whether the rideshares actually fly, whether a regulator moves on a constellation of this size, and whether Starship's commercial cadence starts to look like a schedule instead of a promise. Cisco's participation is a quieter signal that the orbital play is about networking as much as compute, because a fleet that size needs a space-to-ground data path to be worth anything.

Why this matters

This round locks Nvidia into orbital data center hardware and gives Starcloud the capital to move from validation toward production. For anyone planning AI infrastructure, the energy bottleneck is pushing compute into places that were unthinkable two years ago, and launch capacity now decides which of these bets pays off.

✔Human Verified

Researched and cross-referenced against primary sources by the Bytevyte editorial team. This article was generated with the assistance of artificial intelligence and reviewed by the Bytevyte editorial team.

What the money buys

Why Nvidia is backing orbital data centers

The launch bottleneck

Why this matters

Related Articles