$250M Series A Extension Values Starcloud at $2.3B

For venture investors, Starcloud's round is a case study in hardware startups compressing the gap between moonshot demo and mega-round. Ten months after its first orbital AI training run, the company extended its Series A at a $2.3 billion valuation with Manhattan West leading and strategic capital from Nvidia and Cisco Investments — a signal that deep-tech infrastructure is still commanding premium pricing when a startup controls a credible technical wedge.

On August 22, 2026, Starcloud Inc. said it has raised $250 million in new funding at a $2.3 billion valuation, extending a Series A round first announced in March. Manhattan West led the deal, with participation from more than a dozen backers including Nvidia Corp. and Cisco Investments. The round is a landmark for a hardware startup that only about 10 months earlier put an Nvidia graphics card on a small satellite and completed what it calls the first-ever orbital AI training run. That trajectory — from a single orbital GPU demo to a multibillion-dollar valuation within roughly a year — explains why the announcement matters beyond the dollar amount.

“

said it has raised $250 million in new funding at a $2.3 billion valuation, extending a Series A round first announced in March.

Starcloud's pitch targets the physical constraints of AI infrastructure. Terrestrial data centers face rising power demand, cooling complexity, and land and permitting bottlenecks. Starcloud argues orbital clusters remove much of that burden. The company said satellites do not require the elaborate cooling systems used on the ground. Solar arrays in space can be oriented to face the sun continuously, whereas ground-based solar output falls at night and is degraded by storms and cloud cover. Those advantages, if they hold at scale, could make orbit a competitive compute environment for certain workloads.

The roadmap is specific. After the first orbital training run, Starcloud plans Starcloud-2 for 2027. That system would carry onboard AI chips, storage equipment and a data backup module, and would run both training and inference workloads. Today satellites transmit raw sensor data to the ground; Starcloud says Starcloud-2 will analyze sensory measurements in orbit and stream results, which can be transmitted faster than raw datasets. A later generation, Starcloud-3, would contain graphics cards, cooling and heat dissipation gear with a 200-kilowatt power footprint. Two additional generations are contemplated. Ultimately, the company says it hopes to launch 88,000 more satellites with a combined 20 gigawatts of computing capacity.

The in-orbit inference capability is particularly relevant for satellite operators that face limited downlink bandwidth. Moving analysis to the satellite and transmitting only derived results changes the bandwidth equation and reduces latency for time-sensitive decisions. It also aligns with a broader shift toward edge AI, where inference occurs closer to data sources. If successful, Starcloud's approach could create a new class of space-based AI infrastructure that complements, rather than replaces, terrestrial data centers.

The strategic significance is amplified by Nvidia and Cisco Investments. Nvidia's participation links the leading AI accelerator maker to an emerging off-planet compute architecture, while Cisco's corporate investment arm suggests interest in networking and data movement between orbit and ground. For a startup with no revenue disclosed in the source material, a $2.3 billion valuation reflects investor conviction in the scarcity value of orbital AI compute and the technical milestone of an actual training run in space.