The SPACEX AI Revenue Case
xAI already #2 in AI cloud rental, ahead of Google and Amazon on all cloud revenue
Current run rate: $26B/year in signed contracts (up from under $1B), mostly profitOnly using 40% of Colossus 2; fully rented out, run rate goes to $64B/year Full Colossus 1 2 rental would exceed Microsoft, Amazon, and Google cloud combined
Rubin chips (500K GPUs, ~1 GW): worth ~$100B/year in rental revenueIf xAI replicates Anthropic’s model capability, inference value 10x, pushing toward $300b to 400B/year.
@elonmusk @ARKInvest @MilkRoadAI @chamath
Jensen Huang said 1 GW of data center is worth $300b to 400B/year in revenue. Complete AI data centers are worth more than a pile of components alone.
xAI can build new data centers in ~1 year vs. 4 years for competitors
Cursor acquisition closing gap with Anthropic’s Claude Code; Grok training 10x faster; models expected to converge
Starlink and Launch Moats
Starlink: 12M customers now, targeting 20M by year-end
Version 5 dish: thinner, cheaper, router built in, USB-C, battery-powered, no separate GPS neededDownload: up to 1-2 Gbps; upload: symmetrical gigabit, fully competitive with fiber Factory scaling from 1M dishes/month to 50K/day (~1.5M/month) in 2-3 months Elon flagged “a few hundred million dishes” as the deployment target, implying ~$200B/year in broadband revenue at $500/customer/year
Starlink Mobile: 650 satellites, video and text working now; V3 satellites bring 5G direct-to-cellCapacity scales 1,000x as satellite versions improve (V3 to V5)
SpaceX launches 75-90% of everything into orbit; dominance measured in decades
Starship V3 expected operational by July; each launch carries ~30x the capacity of 10 Falcon 9sTarget: 1,000 Starship launches in 2028, 4,000 in 2029, 14,000 by 2032 Even at 700K tons in 2030 (2-3 years late), still 1,000x ahead of nearest competitor
133 Starship launches needed per gigawatt of space-based AI 1 GW with Rubin chips worth $100b to 300B/year
Gigafactory expansion: 11M sq ft total complex, ~10x the size of Giga Austin. no competitor makes more than 100 satellites/year
Revenue Picture and IPO Framing
Four existing, revenue-generating businesses (not speculative). Launch and satellite manufacturing: dominant monopoly Starlink broadband: 12M customers, scaling fast AI infrastructure (xAI): $26B run rate, capacity to reach $64B per year just selling what exits. Consumer/digital advertising $3B/year now, path to $600B via X platform and Starlink Mobile on-ramp
Additional near-term streams Golden Dome (up to $1.2T over 20 years per CBO), Tesla supercharger rack co-location
IPO valued at ~$1.75T. Near-term technology catalysts include V3 Starship launches and V3 satellite deployment. Successful V3 launch in July would crystallize the roadmap skeptics currently push to 2028
At 100 GW of launch capacity and Feynman-era chips: potential revenue in the multi-trillion range, exceeding the combined EBITDA of today’s top 20 companies