Starcloud Raises $250M to Put AI Compute in Orbit
Backed by Nvidia and Cisco, the hardware startup is betting that the next frontier for AI infrastructure isn't on the ground. A $2.3 billion valuation says investors agree.
The assumption has always been that AI compute lives on the ground — inside hyperscale campuses drawing from the grid, cooled by water, constrained by real estate. Starcloud is building the case that orbit is the better answer.
On August 21, 2026, the AI hardware startup announced a $250 million funding round that values the company at approximately $2.3 billion post-money. The round was led by Manhattan West, with strategic participation from Nvidia and Cisco Investments — two names whose presence in a cap table carries specific weight. Nvidia supplies the silicon that runs frontier models; Cisco builds the networking fabric that moves data between them. Neither writes checks into infrastructure plays without a thesis about where compute is actually going.
What Starcloud Is Actually Building
Starcloud's stated focus is orbital AI data centers — compute infrastructure positioned in Earth orbit rather than in terrestrial facilities. The company frames its platform as purpose-built for high-intensity AI workloads, specifically the kind of frontier model training and inference that pushes conventional data center economics to their limits.
The distinction it draws is deliberate: specialized AI infrastructure, not general-purpose cloud. That positioning matters because it defines the competitive lane. Starcloud isn't trying to displace AWS or Azure for routine enterprise workloads. It's targeting the narrow but rapidly expanding tier of compute demand that existing infrastructure struggles to serve efficiently — the workloads where power density, thermal constraints, and terrestrial permitting become binding constraints.
Orbital infrastructure sidesteps some of those constraints by design. The physics of space — abundant solar power, passive radiative cooling, absence of land-use politics — present real engineering advantages for dense compute. The question has always been whether the cost and complexity of getting hardware to orbit could be made to pencil out. Starcloud's $250 million raise suggests at least some institutional capital believes the math is closing.
Why Nvidia and Cisco Are in the Round
Strategic investors rarely participate in infrastructure rounds purely for financial return. Nvidia's involvement signals something about where the company sees AI compute demand moving — and what kind of infrastructure it wants its chips running inside. If orbital data centers become a real compute tier, Nvidia needs a platform relationship there. The same logic applies to Cisco: networking between orbital nodes, and between orbit and ground stations, is a new problem set that its hardware and software could be built to solve.
The presence of both companies in this round transforms it from a venture bet into something closer to an ecosystem-formation play. Starcloud gets more than capital — it gets supply chain credibility and potential preferential access to hardware that is, in many segments of the market, still constrained.
Manhattan West leading the round adds a different signal: institutional financial capital, not just strategic, is treating orbital AI infrastructure as an asset class worth anchoring.
The Broader Bet on Non-Terrestrial Compute
This is reported as one of the largest AI infrastructure rounds disclosed in August 2026, which puts it in context. Investor appetite for AI infrastructure has not softened — it has diversified. The early rounds in this cycle went to terrestrial GPU clusters and purpose-built data center developers. Capital is now moving further out the risk curve, toward architectures that don't yet have a proven operational track record at scale.
That's a meaningful shift. It suggests that the straightforward terrestrial plays are seen as increasingly competed, and that differentiated infrastructure — orbital, modular, physics-constrained in different ways — is where differentiated returns might still exist.
For founders and operators building on AI infrastructure, Starcloud's round is a signal worth tracking. The compute layer is not settling. The geography of where AI workloads actually run — and who controls that layer — is still being contested. A $2.3 billion bet that orbit is part of the answer is no longer a science fiction premise. It's a funded company with Nvidia and Cisco on the cap table, and a clock running on whether the engineering can catch the valuation.