Major energy constraints on Earth for training and running large language models have accelerated interest in orbital AI data centers, where continuous solar power and deep-space radiative cooling eliminate terrestrial bottlenecks. SpaceX leads with plans for initial AI compute demonstrations by late 2027 using Nvidia-based satellites, following its FCC filing for up to one million units and “as early as 2028” deployment target. Starcloud achieved the first orbital H100 GPU flight in November 2025 and trained an LLM in space shortly after, while Google’s Project Suncatcher and Blue Origin’s Project Sunrise advance prototype launches slated for early 2027. Regulatory approvals, launch cadence, and demonstrated radiation-hardened performance remain key swing factors for market resolution.