**Recent technical demonstrations have advanced orbital AI compute but fall short of market thresholds for a qualifying "AI data center," keeping near-term odds low.** Starcloud’s November 2025 Starcloud-1 mission placed a single Nvidia H100 GPU in low Earth orbit, enabling the first in-space LLM training and Gemini inference runs, yet it carried far fewer than the required 100 data-center-grade accelerators. SpaceX unveiled its AI1 satellite design in June 2026 (~150 kW peak compute, roughly one server rack) and filed FCC plans for up to one million orbital AI satellites targeting initial deployment as early as 2028, while Google’s Project Suncatcher and Blue Origin’s Project Sunrise advanced prototype and constellation filings focused on solar-powered TPU/GPU clusters in sun-synchronous orbits. **Terrestrial constraints—power queues, water scarcity, and grid limits—drive interest, but economic and engineering hurdles remain steep.** Launch costs, radiation hardening, thermal management, and inter-satellite laser links must scale dramatically; most analysts view meaningful capacity (tens of MW+) as unlikely before the late 2020s even with Starship maturation. Upcoming catalysts include Google/Planet prototype launches targeted for early 2027, Starcloud’s planned follow-on missions, and FCC decisions on large constellations, all of which could shift sentiment if they demonstrate rack-scale or larger qualifying systems. Traders price strong long-term feasibility against near-term execution risk, reflecting aggregated capital-at-risk consensus on ambitious timelines.