SoftBank, a major backer of OpenAI's Stargate project, commits $500 billion over four years to build terrestrial data centers. Its CEO, Masayoshi Son, dismisses orbital alternatives as too slow and costly for the current AI race, according to Fortune. The $500 billion investment anchors the immediate future of AI compute firmly on Earth, signaling that orbital data centers will likely remain a niche or long-term speculative venture.

This terrestrial focus stands in stark contrast to Elon Musk's vision for orbital data centers, a concept that directly benefits SpaceX's launch services. Musk's plan involves a constellation of satellites requiring frequent replacement, thereby creating consistent demand for SpaceX, according to TechCrunch. The fundamental divergence reveals a tension between ambitious technological visions and the pragmatic, cost-driven realities currently shaping AI infrastructure.

The Current State of AI Compute

  • SoftBank focuses on building data centers on Earth, backing OpenAI's Stargate project with an initial $19 billion. The total planned investment reaches $500 billion over four years, according to Fortune. The $500 billion commitment establishes a clear industry consensus for terrestrial infrastructure. The scale of investment suggests a belief that current orbital solutions cannot meet the immediate demands for AI compute power, prioritizing speed and capacity over potential long-term space advantages.
  • Starcloud partnered with NVIDIA to launch a single satellite containing an NVIDIA H100 GPU into orbit, according to Spiceworks. The launch of a single satellite marks an initial, symbolic foray into space-based compute. While technically impressive, a single GPU in orbit offers negligible compute capacity compared to terrestrial farms, highlighting the vast gap between proof-of-concept and scalable deployment.
  • SpaceX continues renting out terrestrial compute power to major AI players like Google and Anthropic, according to TechCrunch. SpaceX's activity of renting out terrestrial compute power underscores a reliance on the very market it aims to disrupt. The dual strategy suggests SpaceX acknowledges the current limitations of orbital compute, even as it promotes its future potential.