--- format: "aidr-story-markdown/v1" id: "2514db179bc5e604a94296f853285ea95b77538d566ef5f905302b551c9e0fde" canonical_url: "https://aidr.today/2514db17?lang=en" title: "SpaceX Plans 1 Million Starmind Satellites for First AI Compute Formations" lang: "en" requested_lang: "en" available_langs: ["en","vi"] translation_fallback: null fallback_fields: [] published_at: "2026-10-06T16:37:33.000Z" category: "Infra" topics: ["llm"] source_urls: ["https://huggingnews.com/ai/spacex-plans-1-million-starmind-satellites-for-first-ai-compute-formatio-5ed6b9ac","https://marketbrief.now/ai/spacex-plans-1-million-starmind-satellites-for-first-ai-compute-formatio-5ed6b9ac"] summary: "SpaceX is developing a method to group satellites into tight clusters to enable coherent AI compute, with approximately 10 satellites per cluster and 10 terabits of bandwidth between them, as part of a planned constellation of 1 million Starmind satellites. The architecture involves a hybrid model of larger 250kW satellites and tight clusters operating in sub 750m proximity, a shift from the kilometer-scale spacing historically used for Starlink. Analysis suggests the setup could support coherence for up to 720 GPUs, more than double the approximately 300 GPUs in previously tested DeepSeek clusters." --- # SpaceX Plans 1 Million Starmind Satellites for First AI Compute Formations > [Open the canonical story]() **Published:** 2026-10-06T16:37:33.000Z **Category:** Infra **Topics:** llm ## Summary SpaceX is developing a method to group satellites into tight clusters to enable coherent AI compute, with approximately 10 satellites per cluster and 10 terabits of bandwidth between them, as part of a planned constellation of 1 million Starmind satellites\. The architecture involves a hybrid model of larger 250kW satellites and tight clusters operating in sub 750m proximity, a shift from the kilometer\-scale spacing historically used for Starlink\. Analysis suggests the setup could support coherence for up to 720 GPUs, more than double the approximately 300 GPUs in previously tested DeepSeek clusters\. ## Sources - [Story source]() - [Story source]()