SpaceX's confidential S-1 pre-IPO filing contradicts CEO statements by warning that orbital AI data centres involve unproven technologies and may never achieve commercial viability. Just three months after claiming at the World Economic Forum that space-based AI was a 'no-brainer' achievable within two to three years, the company's legal disclosures highlight severe engineering constraints: heat dissipation in vacuum requires enormous radiator surfaces, scaling solar power to hyperscale levels is orders of magnitude beyond anything built in space, and GPU hardware obsolescence makes in-orbit servicing prohibitively complex. Despite this, competitors including Google (Project TPU satellites), Blue Origin, and startup unicorn formerly known as StarCloud are all pursuing orbital compute. SpaceX is targeting a $1.75 trillion IPO valuation partly on the strength of this initiative, creating a tension between promotional narrative and legal risk disclosure. Meanwhile, terrestrial alternatives like nuclear-powered data centres and undersea deployments are attracting far more capital and progressing faster.
Table of contents
The physics of the problemThe competitive landscape in orbitThe terrestrial alternativesThe IPO contextSort: