ทวีตที่ปักหมุด
SpaceX $SPCX just disclosed Starlink V3 capacity in its S-1. The market is reading the wrong number.
The headline is a >10x jump per satellite (1 Tbps vs V2’s 96 Gbps). But the real number to internalize isn’t 10x.
Constellation throughput compounds two curves: satellite count × per-satellite capacity. 10x fleet × 10x capacity = 100x total constellation throughput.
This is what re-rates the Starlink TAM, not the per-satellite headline.
This is the communications layer reaching dominance.
Terrestrial fiber and cellular towers are capex-prohibitive across emerging markets. Light in fiber travels ~31% slower than in vacuum, so orbital mesh already beats subsea cable on transcontinental latency. At ~100 Pbps constellation capacity, the marginal cost curve for global bandwidth collapses.
Post xAI merger, the compute layer gets its physical substrate.
100k V3/V4/V5 satellites isn’t just broadband. It’s the orbital backbone for AI compute distribution — Grok inference, training data pipelines, edge compute — all vertically integrated inside one entity.
Communications + compute in the same orbital fabric. No competitor is close.
Elon Musk@elonmusk
There will also ultimately be >100k V3/V4/V5 satellites for Starlink broadband and direct to cellphone connectivity. If growth continues, Starlink will one day carry the majority of Internet traffic. At that point, it is the Internet and everything else just connects to Starlink.
English
