Should we worry that a rocket company now controls who gets to build AI?
SpaceX collects over $2.3 billion per month renting GPU access to Anthropic, Google, and now Reflection AI — making it arguably the world’s largest AI infrastructure provider. When the physical bottleneck for building frontier AI is controlled by a single company that also launches satellites and runs a social network, what does that concentration of leverage mean for competition, open-source AI, and who ultimately shapes the technology?
Commentaires (1)
Today’s episode of Minds, Bodies, and Terawatts (June 23) dug into SpaceX’s $6.3 billion compute deal with Reflection AI and what it reveals about power consolidating around physical infrastructure rather than algorithms. The episode explored a fascinating tension: Reflection is an open-source lab that gives its model weights away for free, yet it needs the most expensive compute facility on Earth to train them. That paradox suggests ‘open’ AI may still funnel through very narrow gates. Give the full episode a listen and let us know — is infrastructure concentration the bottleneck nobody’s talking about?
Related reading on unscarcity.ai:
Envie d'aller plus loin ?
Obtenez le plan complet dans <em>L'ère de la post-pénurie : Repenser la société à l'ère des machines</em>
