Orbital AI: Seattle-area startup Starcloud hits $1.1B valuation to build space-based data centers

Starcloud raised $170M in new funding at a $1.1B valuation, reaching unicorn status 17 months after Y Combinator demo day and bringing total capital raised to $200M. The Series A was led by Benchmark (with EQT participation) and an extension co-led by both firms; Benchmark GP Chetan Puttagunta will join the board. The startup successfully demonstrated Starcloud-1 in orbit with an Nvidia H100 and plans Starcloud-2 later this year with ~100x the power and an Nvidia Blackwell B200 to run customer workloads. Management expects space-based economics to become favorable in 3–5 years but anticipates <1% of new compute capacity in orbit near-term.

Analysis

Space-based compute changes the demand curve for high-performance accelerators and upstream subsystems rather than replacing terrestrial clouds overnight. Expect a persistent, high-margin niche for specialized inferencing and training workloads where latency to specific satellites or secure processing matters, creating a multi-year runway for premium GPU volumes but with steep engineering hurdles that raise per-unit integration cost by tens of percent versus rack servers.

Second-order winners are suppliers of space-qualified power conversion, thermal control and radiation mitigation — these are low-volume, high-margin franchises whose revenue growth will lag headline chip demand by 6–18 months as qualification cycles complete. Conversely, commodity terrestrial datacenter suppliers face a bifurcation: continued bulk demand for colocated capacity but slower incremental builds in politically sensitive regions, compressing incremental pricing power in those geographies.

Key risks sit in three buckets: (1) economics — launch and ops costs must compress meaningfully relative to terrestrial power+connectivity to move beyond lab pilots; (2) regulatory/ spectrum/insurance friction that can delay commercial scale by multiple years; (3) technical durability — single-point failure modes and refresh cadence for space-hardened accelerators could blow up TCO assumptions. Catalysts to watch are customer SLAs delivered in commercial trials, multi-satellite scheduling software maturity, and any visible drop in launch cost per kg or per GPU node that persists beyond a single promotional manifest.

Key Decisions for Investors

• Long NVDA (6–12 months): buy a modestly sized call spread (delta ~0.35–0.5) to capture upside from an expanded accelerator TAM into edge/space use-cases; cap premium paid to limit downside to defined loss. R/R: asymmetric upside if adoption evidence continues; downside = premium decay if momentum stalls.
• Long MPWR (12–18 months): accumulate shares as a thematic play on space-qualified power management and thermal control demand; trim into 30–40% rallies. R/R: steady multi-quarter revenue re-rating if qualification wins translate to production; risk is slow order ramp and high certification costs.
• Long EQT (12–24 months) small position: exposure to infrastructure operators that can redeploy capital into diversified compute assets (including hybrid space/ground strategies). R/R: dividend + optionality from new service lines; risk = slow monetization and capex absorbtion.
• Pair trade — Long NVDA / Small Short MSFT (3–12 months): small size short to express dispersion between specialized accelerator capture (NVDA) and large, diversified cloud operator execution risk on novel infra. R/R: directional gain if market re-weights toward accelerators; risk = MSFT’s core cloud growth outperforms and squeezes the pair.