Major technology firms and startups are accelerating efforts to deploy orbital AI compute amid terrestrial power and cooling constraints for large language models. Starcloud's November 2025 launch of Starcloud-1 demonstrated the first Nvidia H100 GPU operating in orbit, successfully training a small LLM and running inference on models like Gemma, though it falls short of data-center scale requiring at least 100 accelerators. SpaceX filed FCC plans for up to one million AI1 satellites targeting deployment as early as 2028, while Blue Origin submitted Project Sunrise for 51,600 satellites and Google advanced Project Suncatcher prototypes with TPUs slated for early 2027 launches alongside Planet. Pending regulatory approvals, launch cost reductions, and satellite constellation builds remain key hurdles, with trader-implied odds reflecting skepticism around near-term qualification despite ongoing hardware tests and funding rounds.