How Starcloud turned AI’s power problem into a billion-dollar space startup pitch

Starcloud just raised $170 million at a $1.1 billion valuation, a sign that investors are willing to fund some very strange-sounding ideas if those ideas sit close enough to the AI boom. In Starcloud’s case, the pitch is almost delightfully audacious: put data centers in orbit, where solar power is nearly continuous, land is not a constraint, and Earth’s grid bottlenecks matter less. Reuters reported that the round was led by Benchmark and EQT Ventures, and that the company now has $200 million in total funding.

What makes the company more than a novelty is the problem it is trying to solve. AI systems are chewing through electricity, cooling capacity, and land fast enough that data-center constraints have become one of the defining business stories of this cycle. Reuters said Starcloud is betting that orbital infrastructure can relieve some of that pressure, and that its long-term ambition is an 88,000-satellite constellation for AI workloads. That is not a side project. That is a proposal to move part of the computing stack off-planet.

The company has also been careful to make the vision feel less science-fictional than it first sounds. Reuters reported that Starcloud is already working with Nvidia, Amazon Web Services, and Google Cloud, and that in November 2025 it launched a satellite carrying an Nvidia H100 chip to demonstrate AI training and inference in orbit. A second launch featuring AWS Outposts is planned for October 2026. Those details matter because they turn the company from a slide deck into something more tangible: a startup with live hardware, real partners, and a roadmap that extends beyond mood-board futurism.

Still, the smartest part of the story may be the timing rather than the technology. Starcloud is raising money at a moment when the AI industry’s physical limits are becoming impossible to ignore. The company’s pitch works because it is not saying, “Wouldn’t space be cool?” It is saying, “What happens when Earth stops being convenient?” That is a more serious question now than it was even a year ago, especially as hyperscalers scramble for power and communities push back on ever-larger terrestrial data-center footprints. Reuters described the round as part of a broader surge in investor appetite for space infrastructure bets precisely because AI demand is straining land-based capacity.

That does not mean the economics are easy. They are not. Reuters reported in February that AWS CEO Matt Garman called orbital data centers “pretty far” from reality and said current launch costs make the concept “just not economical.” His skepticism is not trivial, especially since AWS is both a potential partner in these systems and one of the most experienced operators of data infrastructure on Earth.

And the doubts have only grown louder.

In a Reuters analysis published this week, experts argued that SpaceX’s much larger orbital data-center ambitions could run into the same kind of trap that sank Microsoft’s old undersea-data-center experiment: the hardware may work technically, but still fail commercially because it is too costly, too rigid, and too hard to upgrade. Reuters reported that Microsoft’s Project Natick met its technical targets but was abandoned because of weak demand and bad economics, while analysts said space would be even more expensive. One research note cited by Reuters estimated Musk’s million-satellite dream could run into the trillions of dollars.

That is what makes Starcloud such a good startup story. It sits right on the seam between an obvious market need and a very non-obvious solution. If launch costs really do fall enough by 2028 or 2029, as Starcloud CEO Philip Johnston told Reuters, then orbital infrastructure could start to look less absurd and more like a niche but real extension of the AI economy. If they do not, the company may end up as one more beautifully funded monument to a problem that was real but a solution that was too early.

The broader lesson is that the next wave of startup winners may look weird at first because the bottlenecks are getting weird too. When chips are scarce, power is tight, and local communities do not want endless new server farms in their backyard, founders start looking for relief valves in strange places. Starcloud’s billion-dollar valuation says investors are willing to bankroll one of the strangest. Whether that ends up looking visionary or expensive depends less on the poetry of the pitch than on a much duller question with cosmic consequences: can orbital compute ever beat Earth on cost?