Starcloud raised a $250 million extension to its Series A round, bringing total funding to $450 million at a $2.3 billion valuation — more than double its prior valuation. The round was led by Manhattan West, with existing investors Benchmark, EQT, Soma, NFX, and 776 participating alongside new investors including Nvidia, Cisco Investments, Cedar Capital, Goanna Capital, and Standard Capital. Starcloud, founded in 2024 and based in Washington state, builds satellites carrying high-performance GPUs, with the goal of eventually handling AI compute workloads currently run in terrestrial data centers.
What Starcloud has actually flown
This isn't purely speculative: Starcloud flew its Starcloud-1 satellite in November 2025, carrying what the company describes as the first data-center-class GPU deployed in orbit — an Nvidia H100 — and used it to train an AI model in space. The satellite is a roughly 60-kilogram platform, about the size of a small refrigerator. That's a real, independently reported milestone, distinct from the company's larger, unbuilt ambitions.
The plan the new funding supports
The additional capital is stated to fund a larger manufacturing facility and continued development of Starcloud-3, the company's next and largest planned orbital data center spacecraft, intended to launch on SpaceX's Starship. The company's publicly stated long-term roadmap describes scaling from small demonstrators like Starcloud-1 toward gigawatt-class orbital data centers.
The company's own numbers show the core physics problem
Cooling in a vacuum works only by radiation, not convection or conduction — there's no air or water to carry heat away, only the ability to radiate it into space. Per Starcloud's own published white paper, a two-sided radiator held at around 20°C emits roughly 633 watts per square meter — over 1,000 times slower than water-cooling equivalent chips on Earth. That means a modest 1-megawatt orbital facility would need roughly 1,600 square meters of radiator — about the size of a hockey rink — just to stay cool. Running radiators hotter (around 60°C instead of 20°C) roughly halves the required area, but pushes chip temperatures toward their operating limits, trading hardware lifespan for less mass. Independent estimates of a full gigawatt-scale facility, solar arrays included, put the footprint at around 1.6 square miles and total mass above 113 million kilograms — described in that reporting as comparable to an aircraft carrier in orbit.
Other constraints beyond cooling
Reporting on the broader orbital-data-center concept also flags rocket launch costs as a major constraint on building facilities at this scale, the need to replace GPUs roughly every five to six years as they age or become obsolete, and exposure to space debris across a much larger physical surface area than a conventional satellite.
What to expect next
- Watch for Starcloud-3's actual launch and performance data, since Starcloud-1 already validated that a data-center GPU can survive and operate in orbit — the open question is whether cooling and power systems scale the way the roadmap assumes.
- Watch whether Nvidia's investment comes with hardware or technical collaboration terms beyond capital, given Nvidia's GPUs are the specific product being tested in orbit.
- Watch how the cooling-area math is addressed at larger scales — the 1,600-square-meter radiator figure for just 1 megawatt is the single most concrete number for judging whether gigawatt-class claims are physically grounded or aspirational.
References: Starcloud — funding announcement · Starcloud-1 mission page · cooling-physics figures relayed from WebSearch summaries of the World Economic Forum, SatNews, and Scientific American's coverage of Starcloud's own published white paper — this environment could not independently fetch those articles or the white paper itself (egress restrictions) · related coverage: Frontier Arcade: trends & predictions