- SpaceX allocated $15.8 billion to AI-compute infrastructure in Q2, equal to 86% of its $18.4 billion in capex, after spending $7.7 billion in Q1 and $749 million a year earlier.
- SpaceX ended June with 1.4 gigawatts of installed compute capacity and targets more than 2 gigawatts by year-end 2026 and roughly 10 gigawatts by the end of 2027, all using Nvidia hardware under the exclusivity commitment.
- SpaceX has customer agreements covering about 325,000 Nvidia GPUs for Anthropic and 110,000 Nvidia GPUs for Google, with delivery obligations that SpaceX has not yet fully purchased.
- The single-supplier strategy removes competitive bidding on price and delivery timing as Nvidia reports 75% gross margins and supply constraints for its Vera Rubin architecture.
Quotes
“going forward, we've decided to build exclusively on Nvidia because we think the Vera Rubin architecture is the best architecture. We think it's the best AI computer, and we greatly value our close cooperation and partnership on many levels with Nvidia. So, we're exclusive to Nvidia.”
“our understanding with NVIDIA is that we will receive a very significant percent of their GPUs next year.”
“less than a one-year payback”