Starcloud raises $170 million Series Ato build data centers in space

Starcloud's Series A values the company at $1.1B after raising a total of $200M in funding led by Benchmark and EQT Ventures. The firm launched its first satellite with an Nvidia H100 GPU in Nov 2025 and plans Starcloud-2 (multiple GPUs including an Nvidia Blackwell and an AWS server blade plus a bitcoin miner) later this year and a Starship-launched Starcloud-3 (200 kW, 3-ton) designed to reach ~$0.05/kW·hr if launch costs fall to ~$500/kg. Key risks: heavy capex, unproven space-operating tech (power, cooling, synchronization), and dependency on Starship becoming operational at scale (commercial access expected 2028–2029), with broader competitiveness potentially delayed into the 2030s; competitors include SpaceX and other space-data-center projects.

Analysis

The build-out of an orbital compute layer is an option on lower-cost, high-cadence launch and on solving hard engineering constraints (thermal, power, inter-satellite bandwidth). Those are multi-year to decade-scale bets; treat any near-term milestones as informational rather than de-risking events because the unit economics hinge on repeated, cheap mass launches and reliable formation-flying links.

Second-order winners are component-level suppliers that can capture higher ASPs for radiation-tolerant power, thermal management, and optical interconnects — not incumbent hyperscalers. That shifts margin capture away from data-center real estate owners toward specialized hardware vendors and fabless semiconductor suppliers that can command engineering premiums and long lead times.

From a competitive standpoint, the biggest structural risk to cloud incumbents is fragmentation of the compute stack: if a new class of workloads (latency-tolerant, energy-seeking training/inference) migrates off-planet at scale, hyperscalers will either become customers, partners, or squeezed by vertically integrated space operators. The timeline for any material revenue reallocation is measured in years; invest choices should reflect optionality, not immediate revenue replacement.

Key Decisions for Investors

• Long NVDA via a 12-month call spread (buy 12-month ATM call, sell one 30-40% OTM call) sized 1-2% of portfolio to express durable GPU pricing power from new demand verticals; target 25-40% upside, max loss = premium paid.
• Relative trade: Long NVDA / Short AMZN (equal notional) for 12-24 months to capture secular margin capture in silicon vs. hyperscaler capex risk; size small (1% net market exposure), take profits if NVDA outperforms by 20%, stop-loss if the pair moves against by 12%.
• Buy a tactically small put spread on CWK (6-12 month) to hedge exposure to data-center real-estate obsolescence risk as a hedge against public hyperscaler capex reallocation; limit downside cost to premium and size at 0.5-1% portfolio.
• Allocate a 0.5-1% long option position in an optics/space-hardware pure-play (private or small-cap where available) as a lottery ticket on bandwidth/thermal IP monetization; expect multi-year time horizon and binary outcomes — treat as exploration capital.