Cowboy raises $275 million to build rockets with orbital data center upper stages

Cowboy Space raised $275 million in Series B funding at a $2 billion valuation, bringing total capital raised to about $365 million. The startup plans to build launch vehicles whose upper stages become orbital AI data centers, with first demonstrations expected later this year and a one-megawatt data center launch targeted by end-2028. The round, led by Index Ventures with participation from several major backers, signals strong investor appetite for orbital compute and space-based power infrastructure.

Analysis

The market is starting to price a new layer of demand that is not really “space” so much as long-duration infrastructure optionality: if even a fraction of these orbital compute concepts move from demo to procurement, the near-term winners are not the startups but the enabling layer — high-end accelerators, radiation-tolerant components, thermal systems, optical interconnect, and launch/mission services. That makes the second-order beneficiary set broader than the article implies: NVIDIA benefits from early-design wins, while any terrestrial data-center capex spent on power, cooling, and land gets partially displaced if orbital economics improve. The more important signal is that hyperscalers and AI labs are now willing to experiment with off-Earth capacity, which shifts the conversation from “science project” to “supply-chain bottleneck hedge.”

The risk is that this becomes a capital-formation story faster than an operating model story. The physics demo can be compelling while the unit economics remain fragile: launch cadence, on-orbit failure rates, radiation degradation, insurance, and regulatory delays can easily push commercial revenue 3–5 years out even if demos succeed in months. That delay matters because markets are currently discounting an addressable market curve that assumes execution across launch, licensing, and power delivery all work in parallel; any one of those breaking will compress valuations hard.

The most interesting contrarian view is that this may be bullish for compute demand but bearish for near-term terrestrial infrastructure expectations. If orbital compute is real, the first-order effect is not that data centers “move to space” broadly; it is that the most power-constrained frontier workloads get ring-fenced, while terrestrial campuses still need backup generation, grid interconnect, and cooling spend. That argues for a trade that favors picks-and-shovels exposure over the headline story: the startups can be overvalued while the enabling hardware and a subset of defense/space contractors quietly gain multi-year content.

Key Decisions for Investors

• Go long NVDA on a 6-12 month horizon as a call option on early design wins in orbital AI compute; use pullbacks to add, with upside from incremental platform lock-in and low downside if the theme stalls.
• Long META / short a basket of terrestrial data-center infrastructure proxies over 3-6 months if orbital power narratives intensify; the pair expresses a hedge against future capex displacement while preserving AI demand exposure.
• Initiate a small long SAIC position for 6-18 months only as a speculative government/space-program read-through; upside is limited but could benefit from defense-adjacent space infrastructure spend if the market broadens beyond pure venture.
• Buy downside protection on overhyped private-space comparables via public-market hedges where possible; the risk/reward favors shorting enthusiasm indirectly rather than the underlying AI buildout, which may remain structurally bullish.
• If available in liquid wrappers, pair long NVDA / short high-multiple listed data-center REIT or infrastructure names for 3-9 months to express ‘compute wins, real estate loses’ while avoiding pure binary venture risk.